University of Southern California Dissertations and Theses

Auditory brainstem responses (ABR): quality estimation of auditory brainstem responsses by means of various techniques

(USC Thesis Other)

Auditory brainstem responses (ABR): quality estimation of auditory brainstem responsses by means of various techniques

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content AUDITORY BRAINSTEM RESPONSES (ABR): QUALITY ESTIMATION OF AUDITORY BRAINSTEM RESPONSES BY MEANS OF VARIOUS TECHNIQUES by Maria latrou A Thesis Presented to the FACULTY OF THE SCHOOL OF ENGINEERING UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree MASTER OF SCIENCE IN BIOMEDICAL ENGINEERING August 1995 Copyright 1995 Maria latrou This thesis, written by MARIA IATROU under the guidance of Faculty Committee and approved by a ll its members, has been presented to and accepted by the School of Engineering in p artia l fu lfillm en t of the re quirements fo r the degree of MASTER OF SCIENCE IN BIOMEDICAL ENGINEERING Date. Faculty Committee Chairman ACKNOWLEDGEMENTS The experiments were conducted at the Eiectrophysioiogy Lab of House Ear Institute, under the supervision of Dr. Manuel Don, to whom I am grateful for his guidance and suggestions. I would also like to thank Dr. Curtis Ponton for his significant assistance, and Ann Masuda who very patiently taught me how to conduct the experiments. I express my gratitude to Dr. Vasilis Marmarelis for establishing the collaboration between the University of Southern California's Department of Biomedical Engineering and the House Ear Institute, as well as for his support throughout this project. I would also like to thank Dr. Jean-MichaeJ Maarek and Dr. Michael Khoo for their suggestions. iii TABLE OF CONTENTS ACKNOWLEDGMENTS ii LIST OF FIGURES v I. INTRODUCTION 1 A. OVERVIEW OF AUDITORY BRAINSTEM RSPONSES (ABR) 1 1. Anatomic Origins and Generators of Each Wave Component 6 B. BACKGROUND NOISE AND SNR 10 1. Averaging 11 2. Filtering 11 3. Artifact Rejection 12 4. Stimulus Parameters 12 5. Electrode Placement 12 6. Relaxation and Sedation 13 C. A QUALITY CONTROL CRITERION FOR ABR 13 D. AIM OF THE PRESENT STUDY 14 II. METHODS 16 A. EXPERIMENTAL DESCRIPTION 16 1. Subjects 16 2. Stimuli 16 3. Subject Preparation 17 iv B. DATA PROCESSING AND ANALYSIS 20 1. Averaging: Standard and Weighted 20 i. Theoretical Background 20 ii. Application in this Study 22 2. Quality Estimation of ABRs 23 i. Theoretical Background 23 ii. Application in this Study 25 3. Crosscorrelation 26 III. RESULTS 27 1. Averaging 27 2. Background Noise 29 3. Minimum “Quality’’ Estimation 33 4. Minimum “Quality" Estimation for Combination of Recording Channels 37 5. Crosscorrelation 50 IV. DISCUSSION 51 V. REFERENCES 56 VI. APPENDIX “A” 57 Table 1. List of Abbreviations 58 Table 2. Database for Estimating the Quality Criterion 59 Table 3. Crosscorelation Between Channels 73 V LIST OF FIGURES Fig. 1. Representation of Major AER Waveform 2 Fig. 2. ABR Waveforms from Original Studies 3 Fig. 3. Schematic Representation of Auditory 4 Brainstem (Front View). Fig. 4. Schematic Representations of Auditory 5 Brainstem (Side View). Fig. 5. Cross-section of the Ear 7 Fig. 6. Auditory Brainstem 9 Fig. 7 & 8. Placement of Electrodes 18 Fig. 9 & 10. Fsp for Standard and Bayesian Noise 27 Averaging Fig. 11 & 12. Noise Characteristics vs Number of 28 Measurements Fig. 13 & 14. Background Noise Characteristic Under 30 the Various Test Arrangements Fig. 15 & 16. Background Noise Characteristics Under 31 the Various Test Arrangements Fig. 17 & 18. Background Noise Characteristics Under 32 the Various Test Arrangements Fig. 19 & 20. Number of Blocks Required for Fsp to Reach 34 the Criterion of 3.1, with Subject Asleep and Stim. Levels at 38 and 43 dB. vi Fig. 21 & 22. Fig. 23 & 24. Fig. 25 & 26. Fig. 27 & 28. Fig. 29 & 30. Fig. 31 & 32. Fig. 33 & 34. Fig. 35 & 36. Fig. 37 & 38. Fig. 39 & 40. Number of Blocks Required for Fsp to Reach 35 the Criterion of 3.1, with Subject Asleep/Awake and Stim. Levels at 53 and 38 dB. Number of Blocks Required for Fsp to Reach 36 the Criterion of 3.1, with Subject Awake and Stim. Levels at 43 and 53 dB. Number of Blocks Required to Reach the 37 Criterion of 3.1 with 99% Confidence Level for Individual and Combinations of Channels Number of Blocks Required to Reach the 40 Criterionof 3.1 with 99% Confidence Level for Individual and Combinations of Channels Number of Blocks Required to Reach the 41 Criterionof 3.1 with 99% Confidence Level for Individual and Combinations of Channels Number of Blocks Required to Reach the 43 Criterion of 3.1 with 99% Confidence Level for Individual and Combinations of Channels Number of Blocks Required to Reach the 44 Criterion of 3.1 with 99% Confidence Level for Individual and Combinations of Channels Number of Blocks Required to Reach the 46 Criterion of 3.1 with 99% Confidence Level for Individual and Combinations of Channels Number of Blocks Required to Reach the 47 Criterionof 3.1 with 99% Confidence Level for Individualand Combinations of Channels Number of Blocks Required to Reach the 49 Criterionof 3.1 with 99% Confidence Level Fig. 41 & 42. vii for Individualand Combinations of Channels Number of Blocks Required to Reach the 53 Criterion of 3.1 with 99% Confidence Level for Individual and Combinations of Channels 1 INTRODUCTION A. OVERVIEW OF AUDITORY BRAINSTEM RESPONSES (ABR) An auditory evoked response (AER) is activity within the auditory system (the ear, the auditory nerve, or auditory regions of the brain) that is stimulated by acoustic stimuli. The intensity of the stimuli may be high to low. Sounds with greater intensity evoke larger auditory brain responses. The sounds are presented to a person by an acoustic transducer and the evoked brain activity is picked up by surface electrodes placed on the scalp. Since the electrodes are located relatively far from the generators of the responses, it is difficult or even impossible to specify the source of these responses within the auditory system. It is usually possible to determine the region of the auditory system responsible for the response by analyzing its pattern and calculating the time interval between its occurrence and the onset of the stimulus. The auditory evoked responses (AER) occur within the first second after the stimulus. In Fig. 1 representative waveforms of major AER are shown. The AERs are voltages of the order of microvolts (pVolts). Activity arising from the higher regions of the auditoty system tends to be larger in amplitude than activity generated from lower egions like the ear, the auditory nerve, and the brainstem , because it involves more neural units and the electrodes are relatively close. The Auditory Brainstem Response (ABR) is the response pattern of an AER that occurs a few milliseconds (ms) after the stimulus onset and lasts several milliseconds. For high stimulus intensities the observation window for an ABR extends from 1 ms to 11 ms postimulus (Don and Elberling, 1993). Don Jewett was the first investigator who observed the ABR in the 1960s, although at first he dismissed these observations as "nothing but artifacts.” Electrocochleography (ECochG) AP Auditory Middle Latency Response [AMLR) B * [10ms] Auditory P300 Response Auditory Brainstem Response (ABR) Auditory Late Response (ALR) [50ms] Fig. 1. Representation of Major AER Waveforms 3 ABR waveforms are presented in Fig.2 from the first paper on ABRs by Jewett and Williston in 1971. The major peaks in the waveforms noted with Roman numerals are the components of an ABR. iv : . ii m -■'J \ vr i j i .-v v i i v -l- Vv JB _ •£ * * * - A B > v • .' * 1 . :.-v V- * / w"\ *. DJ PL SK V. D S H W I V • III \ VI VII B-i -*« ' . * / % c i y \ J E J - v v W . JB LY V JY . JS 5 M SEC A . « a • * X 5M S E C HS JEB Fig. 2. ABR Waveforms from Original Studies Thus, ABR waveforms comprise of seven waves I, II, III, IV, V, VI, VII. ABR latency and amplitude measures reflect different physiologic processes which may interact and different waves reflect functionally separable substrate systems. Within each substrate system, latency is postulated as a function of fiber transmission time plus time to synaptic threshold of a target cell 4 population, while peak amplitude is postulated as a function of graded postsynaptic potential summation. Information about the specific anatomic origins of the ABR is conflicting for later components (waves III, IV, V, and VI) than for earlier components (waves I and II). MGN Pineal CIC CP LLD. Brachium Conjuctivum LLV Brachium Pontis VII Facial Colliculus Restiform Body VIII AS AN AVCN-^ P V C N - — . DCN SOC Obex /^uuiiuiy Radiation LCfadil Cunni 1mm Fig. 3. Schematic Representation of Auditory Brainstem (Front View). Fig. 4. Schematic Representations of Auditory Brainstem (Side View). Figures 3 and 4 are schematic representations of the auditory brainstem and Table 1, in appendix A, includes the definitions of the abbreviations used in these figures ("HandBook of auditory evoked responses”). Therefore, except for waves I and II, ABR components presumably have multiple generators. There is contribution of activity from more than one anatomic structure to a single wave. Conversely, the same anatomic structure can contribute to more than one peak. Auditory information is not simply passed-on sequentially from one relay station to another. Complexity of auditory brainstem anatomy and timing of activity arising from different structures are certainly important factors in the anatomy versus ABR wave component relationship. It is reasonable to expect that structures that are located close to one another in the auditory brainstem, or that are activated at or about the same time after the stimulus, each contribute in varied amounts to the wave. With this spatial and temporal summation of auditory brainstem activity, a wave component probably could arise from mostly one structure even though more than one structure was activated by the stimulus. Also, only a subset of neural units within any anatomic region may actually contribute to the ABR. Anatomic origins and generators of each wave component WAVE I. The ABR wave I component is the farfield representation of the action potential (AP) in the distal portion of the eighth nerve that is, afferent activity of the eighth-nerve fibers as they leave the cochlea and enter the internal auditory canal (Fig. 5, "Bases of auditory brain-stem evoked responses”). Cartlage Mastoid Cemicircular Cells Malleus Canals / I Incus A Vestibule Vestbular N. Facial N. Cochlear N. Internal " A u d ito ry Canal Cochlea External Auditory Canal Round Window Cross Section of Eustachian Tube Fig. 5. Cross-Section o f the Ear WAVE II. Wave II is generated by the proximal eighth nerve as it enters the brainstem (Figures 3,4). A proximal eighth-nerve generator site for wave II is supported by the relationship between the latency of waves I and II and the relatively slow conduction time expected for the auditory nerve (10-20 m/sec) which is on the average 25 mm long with a diameter of approximately 2-49|im in adults. 8 WAVE III. Wave III arises from second order neuron activity (beyond the eighth nerve) in or near the cochlear nucleus, whereas the negative trough arises from the trapezoid body (Fig.6, “An introduction to the physiology of hearing”). WAVE IV. Wave IV often appears as a leading shoulder on wave V. Intracranial investigations imply that the wave IV arises from pontine third-order neurons mostly located in the superior olivary complex , but probably with contributions also from cochlear nucleus and nucleus of lateral lemniscus. WAVE V. Wave V is the component analyzed most often in clinical application of the ABR. The positive voltage wave V is related to the termination of lateral lemniscus fibers as they enter the inferior colliculus (contralateral to the stimulated ear), while the large, broad negative trough following wave V is attributed to dendritic potentials within the inferior colliculus. WAVE VI AND VII. These peaks are attributed either to continued synchronous firing of neurons in the inferior colliculus or to a thalamic (medial geniculate body) origin. ABR is influenced by subject age, gender, and body temperature. It is not seriously affected by subject state of arousal or most drugs including sedatives and anesthetic agents; this is a major clinical advantage. ABR is used for newborn infant auditory screening, estimation of auditory sensitivity in very 9 young or difficult-to-test children, neurodiagnosis of eighth nerve or auditory brainstem dysfunction and monitoring the eighth nerve and auditory brainstem status intraoperatively during posterior fossa surgery. DAS RB N V IAS Auditory Nerve SOC SOC Fibres in TB I Fig. 6. Auditory Brainstem Its clinical use is limited by the fact that it doesn’t provide any information on the auditory system above the brainstem level, it is not a hearing test and finally with click stimulus it estimates hearing sensitivity in 1-4 KHz region and not for lower frequencies. There are two reasons for this. First, the response to the cochlear activation in the higher frequency regions has already occurred by the 10 time the traveling wave has covered the distance from the base (high frequency region) to the apex (low frequency region) and activated hair cells in this region. Second, the leading “front” of the traveling wave is less abrupt when it reaches the apical region and consequently not as effective in producing synchronous firing of many eighth-nerve afferent fibers over a concentrated portion of basilar membrane. In the last two decades, the use of auditory brain-stern responses for assessing peripheral auditory function has proliferated. The recordings of ABRs have been proven valuable in the audio-, oto-, and neurological clinics. However, for most applications, a major drawback of ABRs is their low amplitudes relative to the physiological background noise (BN) which requires the use of time consuming signal extraction techniques, such as averaging responses in the time domain. The poor signal-to-noise ratio (SNR) has been the major problem in identification of near-threshold responses and in reliable measurement of latency and amplitude of the components of the ABR. B. BACKGROUND NOISE AND SNR The physiological background noise is composed of electrical activity of both neural and muscular origins. Large amplitudes of background noise are often clearly observed in the recordings when a subject moves or is not relaxed. The largest the background activity is, the more the sweeps that are required to 1 1 reduce the noise in average. Testing may even be abandoned in such cases. Therefore, it is very essential for reliable ABRs to improve the ratio between the actual evoked potential (EP) and the background noise. The SNR can be improved by using different techniques, the description of which follows in brief. 1. Averaging The standard technique computes the time averaging of a series of postimulus time epochs. If the background noise is stationary, the averaging method reduces the level of BN in the final averaged wavefom by the square root of the number of sweeps. Besides the standard method for averaging another method called the “Bayesian weighted averaging” is also used. The latter method takes into consideration the noise variation and applies smaller weights to noisier waveforms before averaging. Consequently, this method is more efficient than the standard averaging in recovering the ABR from the background noise (Elberling, Don 1984). 2. Filtering The purpose of filtering is to reduce or remove those frequency components of the background noise that are present in the power spectrum of the EP with little or no energy. Analog as well as digital filtering is used for this purpose. However, because the BN and the EP have overlapping power spectra, filtering offers very little improvement on SNR. 12 3. Artifact rejection The averaging techniques are often combined with artifact rejection. Artifact rejection is excluding from averaging sweeps with signal amplitudes exceeding a preset rejection level,which eventually will improve the SNR. However, if the background noise remains higher than the preset rejection level over a long time the testing time becomes extremely long and the testing may be disrupted. 4. Stimulus parameters The EP can be enhanced by increasing the stimulus level, using clicks as stimuli or reduce the repetition rate of the incoming stimuli. However, click stimuli are less frequency specific and therefore could be inappropriate. Also, lower repetition rates mean longer testing time. All these factors have to be considered for optimal results. 5. Electrode placement The magnitude of the EP depends, in part, on the placement of the electrodes. Choice of specific electrode locations can enhance the magnitude of EP or its specific components. 6. Relaxation and sedation SNR is improved significantly when the subject is relaxed. Children are normally given a sedative to sleep, and newborns are tested during spontaneous sleep induced by feeding. C. A QUALITY CONTROL CRITERION FOR ABR As it has been mentioned previously, different techniques are employed in clinical use of ABRs in order to improve the sound-to-noise ratio (SNR). The primary technique involves the averaging of post-stimulus time epochs, sweeps. Normally, the test protocol prescribes a fixed number of stimuli (sweeps) as the basis of the averaging technique in order to obtain "comparable recording conditions”. However, the variation of the ABR amplitudes is large among “normal” adults and is further increased in “clinical" subjects. Also, depending in a large extent to the relaxation degree of the subject, the physiological background noise is also variable during and between test sessions. Therefore, such a test protocol can not ensure a given minimal “quality” of the averaged response or of the SNR. The control of the response quality is very important in clinical use because it improves the reliability of the tests and equally importantly ensures optimal use of the available testing time. 14 The quality of the responses are estimated by comparing the magnitude of the averaged recordings with that of the averaged background noise. Since, in the recordings the background noise and the response are mixed, statistical tools are used to approximate separately the quantities. The theoretical background of the method will be mentioned in a following section regarding the employed methods. In summary, this statistical method can be used “on line” in adaptive strategy to estimate the number of sweeps necessary to obtain the given minimum “quality” and it consists of evaluating a function that describes the SNR with a certain statistical precision. When the value of this function(Fsp) exceeds the value of 3.1, there is only 1% chance of false-positive indication of a response. By observing these values, an ABR is positively detected as soon as they reach the value of 3.1. However, if this criterion is not reached after a number of sweeps, e.g. 10000 sweeps, an ABR is not detected. The averaging of the responses and the calculation of the "quality” criterion are performed in blocks of 256 sweeps. It is important to mention that these recordings are taken from one pair of surface electrodes, namely one channel. D. AIM OF THE PRESENT STUDY As stressed above, the improvement of the SNR ratio is very crucial in recovering reliable ABRs from the background noise. The aim of the present 15 study is to demonstrate how the different techniques ensure better quality for the estimated ABRs and furthermore to explore the possibility to develop a new minimum "quality” criterion based on the method described in the previous section. The quality estimation of averaged auditory brainstem responses is currently using one channel recordings. The used quality criterion is determined statistically. Therefore, by using recordings from independent channels it is attempted to combine the quality criteria of each channel to determine a joint function which will estimate a minimum quality for the ABR recordings. The clinical purpose of that is time reduction in data acquisition, which is very important especially for children. For this study recordings from four channels were taken. The assumption for channel independence is based on previous studies at the Electrophysiology lab, of the House Ear Institute. 16 METHODS A. EXPERIMENTAL DESCRIPTION 1. Subjects Six subjects, 3 males and three females aged 24-40 years old served as subjects. Subjects were recruited from the staff at the House Ear Institute and from student bodies of local universities. All subjects were in good health and reported normal neurological status. Subjects had normal hearing as determined by pure tone thresholds at or less than 10 dB for frequencies between 500 and 4000 Hz and less than 15 dB for 6000 and 8000 Hz at the time that the ABR’s were collected. Pure tone audiometric testing was accomplished with a Grason-Stadler GSI 16 audiometer and Telephonic TDH 50P earphones in P/N 10 C 0 17-1 cushions. 2. Stimuli Rarefaction click stimuli were presented to the right ear for all subjects through TDH-50 p earphones. Clicks with duration of 100 fisec were repeated at time 17 intervals of 22 ms (approximately 45 clicks/sec) at 0, 38, 43, 53 dB peak-to- peak equivalent sound pressure level (p.p.e. spl) and for two conditions; quietly awake and asleep. The responses recorded from the electrodes were bandpass filtered between 0.1 and 3 KHz with 12 dB/octave slopes and amplified by Neuroscan Synamp Amplifier of two stages. The first stage amplified the signal 500 times and the second stage 100 times. The responses from the four channels were sampled at 25 KHz per channel. The recording epoch lasted for 13 ms (2 ms to 15 ms) after the onset of the stimulus. 10000 single sweeps were stored to the hard disk of the Neuroscan Electrophysiology recording system for off-line processing. The off-line processing involved standard and Bayessian averaging of the 10000 sweeps, the crosscorrelation between channels, evaluation of the averaged background noise and of the minimum “quality’' criterion in the Bayesian averaging , for each of the four channels and each of the four stimuli levels at both the asleep and the awake state. 3. Subject Preparation Electrodes were placed according to the International 10-20 System of Electrode Placement. This system is a procedure for the measured location of equally spaced electrode positions on the scalp, using identifiable scull landmarks as reference points. This system is based on proven relationship between a measured electrode site and underlying cortical structures and areas. The system is termed ‘'10-20” because electrodes are spaced either 10% or 20% of the total distance between a given pair of scull landmarks (Fig. 7). First the circumference of the head was measured, making sure that the tape goes through the frontal pole (Fp), the occiput (O) and T3 and T4 (midtemporal) points (Fig. 8 ). Then the locations of nasion and inion were marked with a non-toxic skin marking pencil and their distance was measured. The 10%, 30%, 50%, 70%, of the total inion-nasion distance were successively determined (Fig.8). These correspond to Fpz = frontal pole (midline), Fz = N a s io n Nasion Pruuricufar Point Left Side of Head inlon Fig. 7 & 8. Placement of Electrodes determined {Fig.8). These correspond to Fpz = frontal pole (midline), Fz = frontal zero (mid-frontal), Cz - vertex, Pz = Parietial zero (mid-Parietial), respectively. Fpz, Cz, Pz, Inion and also the right and left earlobes A2, A1 as well as the ipsilateral mastoid Ml (right) were scrubbed vigorously with an abrasive liquid substance. This removed the natural oils of the skin, thus improving interelectrode impedance. Surface electrodes were applied with conductive paste at these sites. A last electrode was added as a ground on the forehead. Finally, the interelectrode impedances were measured by using an Ohm meter. The measured resistance had to be lower than 3 KOhms, otherwise the corresponding electrode was removed and placed again after additional scrubbing of the scalp at this location. Subjects were then placed in a reclining chair in a sound treated, double walled sound room. ABR’s were obtained from four differential surface electrode channels: 1. Ml-Cz 2. Inion-Fpz 3. A2-Pz 4. A1-Fz Immediately after testing was completed electrodes were removed and cleaned. Electrode paste was carefully cleaned off subjects’ scalp. B. DATA PROCESSING AND ANALYSIS 20 1. Averaging: Standard and Weighted i. Theoretical Background The data for this study consisted of 10000 sweeps, four stimulus levels for each of the subjects at both sleeping and awake state. For each stimulus level and state the stored individual sweeps were processed to form two sets of averages: standard and weighted averages. The sweeps were averaged in blocks of 256 sweeps. For the averaging process it is assumed that the background noise BN(t) is an ergodic random process with Gaussian distribution within each block, i.e. the variance differs from block to block (Elberling.Don 1984). The postimulus time epoch S(t) consists of the evoked potential EP(t), which is a deterministic signal and the background noise BN(t). Normal averages were formed by the straightforward process of summing the individual sweeps and dividing by the number of sweeps summed. According to Elberling and Wahlgreen (1985) the estimated evoked potential after n sweeps is given by: EP„ = -» iS ,+ S 2 + - + S.) (1) n 21 Weighted averaging using Bayesian estimation principles were formed by weighting blocks of 256 sweeps inversely proportional to the value of the estimated background noise. Let Si and Vi indicate the waveform of the /th block and the estimated variance of the corresponding background noise. EP{ denotes the Bayesian estimate of the evoked potential after the /th block. After the nth block, we will have: V , V 2 K cn where C = — + — + •■■+— . " Vx V2 vn This last equation can be written in the following form, which enables comparison to be made with equation (1) of the standard averaging: » \ V2 V j Cn Thus, in the Bayes estimate the individual, /th, block is weighted by a factor n /( ( ^ • C J . It is apparent that when the background noise is stationary, i.e. the variance Vi is constant, the Bayes estimate will be identical to the estimate found by the standard averaging, and no improvement will be obtained. 22 ii. Application in this study The data for this study consisted of 10000 sweeps, four stimulus levels for each of the subjects at both sleeping and awake state. For each stimulus level and state the stored individual sweeps were processed to form two sets of averages: standard and weighted averages. The sweeps were averaged in blocks of 256 sweeps. For the averaging process it is assumed that the background noise BN(t) is an ergodic random process with Gaussian distribution within each block, i.e. the variance differs from block to block (Elberling.Don 1984). First the standard averaging was performed followed by the weighted averaging. For the latter averaging method the variance Vi of the BN for the /th block was estimated by computing sweep-to-sweep variance of a single time point in these sweeps (Elberling, Don). For a block of 256 sweeps, 256 values were used in computing the variance and estimating the background noise. With increasing number of sweeps the sample of single point values converges to the distribution of the real background noise. For the latter computation to hold it is necessary to assume that the BN is stationary within this block, otherwise the approximation of he variance Vi by taking values only on a single time point at every sweep would be false. 23 2. Quality estimation of ABRs i. Theoretical background The signal S(t) recorded over the post-stimulus epoch consists of the evoked potential EP(t), a deterministic signal, and the background noise, BN(t), an ergodic random process: S(t) = EP(t) + BN(t) (4) and after averaging over N sweeps S(t)N =EP(t)N + BN(t)w (5) Considering that the EP is deterministic equation (5) can be written as: S= EP + BN (6) The signal-to-noise ratio squared is defined as: > 2 _ VAR(EP) SNR2 = (7) VAR(BN) where VAR(EP) and VAR(BN) are the variances of the EP and the BN . By calculating the variance on both sides of (6) and by using (7) we get: VAR(S) = VAR(EP) + VAR(BN) + 2 • COV(EP, BN) (8) where COV(EP,BN) the covariance between EP, BN . 24 We could now find an estimation of the averaged background noise by collecting the noise values in one single point of each individual sweep. The variance, VAR(SP), of the single point sample will be a rather accurate measure of the variance of the true background noise VAR(BN) after a few hundred sweeps. Under the assumption of Gaussian stationary noise : VAR(SP) N = VAR{SP) (9) which is the estimated variance of the background noise. By dividing both sides of equation (8) with the above given estimated variance of the BN we form a ratio which follows an F-distribution and is defined as : F = — ffiS t = \SNR2 + 1 + R(EP, BN) • SNRI • YAR ( ^ ) (-| Q)r p VAR(SP) L v ' - I VAR(SP) K ' where R(EP,BN) = t(-,e correlation coefficient of EP and B N . VAR(BN) Since the two terms R(EP,BN) and VAR(BN) are statistically independent they can be treated separately with probabilities PR and Pv respectively. A given confidence limit for the Fs p determines for a known SNR the probability P jo in t of falsely recognizing background noise as EP. Therefore, Pj0 in t = PR • Pv . The statistical variation of R(EP, BN) follows t- distribution with Vi degrees of freedom, whereas the variance ratio follows the F-distribution, F(v!, v 2 ), with Vi and v2 the degrees of freedom of the numerator and the denominator, 25 respectively. The value of v2 is equal to the number of sweeps in each block (N=256), whereas Vi fluctuates because it depends on the noise. An effective value for Vi equal to 5 has been determined (Elberling, Don, 1984). Therefore, when the degrees of freedom are known, we can for given probabilities PR and Pv, calculate the corresponding values of R(EP,BN) and Since VAR(SP) P jo in t is given, a combination of PR and Pv can be selected for each SNR to give the maximum or minimum Fsp. When the signal-to-noise ratio is zero, as we can see from equation (10), the VARCBN) Fs p reflects the variance ra tio . This quantity can thus be evaluated VAR(SP) separately in the no stimulus condition,_where the EP is absent. It has been determined that for v -i equal to 5 the distribution of F( 5 < 2 5 6 ) has upper 99% fractife of 3.1. This means that a response detection criterion of Fs p = 3.1 gives a rate of failure positives of 1%. ii. Application in this study In this study, as in the clinical use of ABR recordings, a response was positively detected as soon as the Fs p -criterion was reached. An upper 99% (Fs p = 3.1) confidence level was used. The Fs p values were computed for every block of 256 sweeps (total of 39 blocks) for every channel separately. A data 26 base was formed for each subject, in which all the computed Fs p values were transferred and ordered in categories according to the corresponding channel, stimulus level, and state condition. The number of the block at which this minimum ‘tjuality"criterion was reached was reported in a new column for later use in comparison between channels, stimulus levels and state conditions. Furthermore, in the used program, the probabilities corresponding to each Fs p value were determined by postulating that the Fs p values followed F-distribution with degrees of freedom Vi = 5 and v2 = 256, i.e. F{5,256). Assuming that the recording channels were independent, the joint probabilities were computed, describing the possibility of all combinations of two channels to reach the set criterion. The block at which this criterion was reached was again noted in a new column for comparison between combination of channels or with the single channel cases. Finally, for each subject, each channel and each of their combination the mean value of the number of the block at which the criterion was reached was plotted as a function of the stimulus level. 3. Crosscorrelation The crosscoreliation functions between the four channel averaged ABRs were performed. The purpose of that was to demonstrate how strong or weak is the correlation between these channels and which of these correlate best. 27 RESULTS 1. Averaging In order to outline the difference between the standard averaging and the Bayesian estimation, the Fs p values and the background noise are plotted for 8 - 6 - 6 - F sp 4 - F s | 4 - 2- Ml "Fsp for Standard Averaging State = Awake, Stim. Level « 43 dB' Channel 1 Ml "Fsp for Bayesian Averaging State = Awake, Stim. Level = 43 dB' Channel 1 Block No Block No 40 3 0 20 20 40 Fig. 9 & 10. Fsp for Standard and Bayesian Noise Averaging each case. Figures 9 and 10 demonstrate how the Fs p values develop with increasing number of blocks for the standard and Bayesian averaging 28 respectively (data from subject 3). During the first 15 blocks, both estimates improve equally in "quality", and therefore the background noise can be regarded as stationary. Further averaging with the two methods demonstrates a difference between them, leading to a deterioration of the "quality” for the standard averaging and to its improvement for the Bayesian averaging. The reason for this latter effect is that the noise suddenly increased and became non stationary. ~ 8 0 - 6 0 - Nois ) (nV) 4 0 - s o — Noise (nV) Ml "Noise for Standard Averaging Slate=Awake, Stim. Level = 4 3 dB" Ml "Noise for Bayesian Averaging S tate = Aw ake, Stim. Level = 43 dB’ C hannel 1 20 - B lock N o. Block No. 30 40 10 20 30 40 20 10 Fig. 11 & 12. Noise Characteristics vs Number of Measurements In figures 11 and 12 the advantageous use of the Bayesian averaging is again demonstrated for the estimation of the background noise. 29 Fig. 11 shows for subject 3 the estimation of background noise with standard averaging. The estimation after the 15th block rises abruptly, indicating that the subject became very noisy in the preceding blocks. Following, the subject became quiet again. Fig. 12 shows the estimated background noise evaluated with the Bayesian method. The detrimental effect of episodic noise is now reduced and lower noise level is reached after 39 blocks of sweeps compared to the noise level reached after the 39th block with the standard averaging. 2. Background noise A possible solution to reducing the effect of episodic noise on the average is to form the Bayesian weighted averaging, the theoretical background of which was investigated previously. The effect of this averaging technique was investigated in 6 subjects. During stimulation at 38, 43, 53 dB p.p. e. spl, and also during no stimulation, N=10000 sweeps were collected from each individual. The weighted averaging of the background noise was performed in blocks of 256 sweeps for a total of 39 blocks. The state of the relaxation and thereby the level of the background noise varies considerably between subjects. In Fig.13 the averaged background noise is presented versus the number of blocks for subject 1, in the sleeping condition and during no stimulation for all 30 four channels. The background noise for subject 1 starts at low level for all channels. Channel 2 seems to be the noisier channel, whereas channel 3 starts from the lowest noise value, which is approximately equal to 13 nV. The averaged background noise drops significantly within the first five blocks for all channels. The noise values observed after the fifth block are below 9 nV. Line C hart Split by; State, Stim Stale= Asleep, S tim . Level = No Stim c h a n n e l I I -n - c h a n n e l 2 V . -V - channd3 Cn c h a n n e l4 channel 1 - 0 - channel 2 channel 3 -if- channel 4 Line Chart Split by: State, Stim. State= Asleep, Stim. Level = 43dB Fig. 13 & 14. Background Noise Characteristic Under the Various Test Arrangements Similarly, in Fig. 14 the averaged background noise is presented versus the number of blocks for subject 1, in the sleeping condition and during stimulation at 43 dB p.p.e. spl for all four channels. The observed values for the BN are close to the ones observed in Fig. 9 indicating that the subject is in a stable relaxation state. Channel 2 is again the noisier and channel 3 the quieter. The 31 BN as shown in Fig. 13, drops drastically at low noise levels within the five first blocks. As mentioned above, the subject’s state of relaxation strongly influences the level of background noise. In Fig. 15 the data for the BN are plotted against the number of blocks, when subject 1 is awake and during no stimulation. We observe that the highest value for BN is 100 nV and is recorded from channel 4. Line Chart, Split by: Slate, Stim. State = Awake, Stim, Level = No Stim. Line Chart, Split by: State, Stim. State = Awake, Stim. Level=43dB Noise m N o is e block no. block n o . Fig. 15 & 16. Background Noise Characteristics Under the Various Test Arrangements The background noise decreases with averaging, but more than five blocks of averaging are needed for all channels to drop below 9 nV. The reduction of the noise level at channel 4 is very fast, which indicates that the subject relaxed. In Fig. 16 the values of the BN are plotted against the number of blocks, when subject 1 is awake and during stimulation at 43 dB p.p.e. spl. 32 The subject is more relaxed than in Fig. 15, as it shows from the lower starting noise values. Channel 2 is picking up higher noise level, approximately 55 nV, whereas channel 3 is recording the lowest starting noise value. The noise level for the four channels drops below 9 nV after 15 blocks. Although, in this case it takes longer for the background noise to drop than in Fig. 10, within 10 blocks of averaging the BN is reduced largely. ^^SpHbjrState.Sta Slate - Asleep; Stim.Level= No Stim. Noise ■♦Channel! * Channel 2 -*■ Channel 3 Channel 4 Block No lOOi Noise Block No Fig. 17 & 18. Background Noise Characteristics Under the Various Test Arrangements In Fig. 17, BN data correspond to subject 2, while sleeping and during no stimulation. The noise levels are a little higher, than the corresponding of subject 1 in Fig. 14, with channel 2 being the noisiest channel. The averaged N decays fast and within 10 blocks all channels record noise values lower than 10 nV. 33 In Fig. 18, data for the BN of subject 2 are plotted under the exact conditions as in Fig. 16. for subject 1, in an attempt to see possible differences due to different subjects. Subject 2 presents a quite higher level of background noise, with worst recordings from channel 2. Data from channels 3 and 4 are almost identical to those in Fig. 16. The noise levels drop slower than in Fig. 16 for all channels. 3. Minimum “quality” estimation As stated under the theoretical background of "Quality estimation of ABRs", the quality of an evoked response recording is ensured by optimizing the SNR. It is possible to set minimum quality for confident detection of a response, and then to determine the numbers of sweeps necessary to reach this criterion of quality. This quality criterion is a measure of estimated variance of background noise, and is symbolized as Fs p , where “F” refers to a statistic description (the F distribution) and “sp” stands for “single point” sample. The Fs p is improved by Bayesian weighted averaging. To stress how the "quality" develops with increasing number of sweeps and demonstrate at which block the minimum criterion was reached, the Fs p values were plotted versus the number of blocks for subject 3, at both awake and asleep conditions and for 38, 43, and 63 p.p.e. spl stimulus levels. An upper confidence limit of 99% (Fs p = 3.1) was set. 34 Ml Lint C h a r t S p lit b y : S t a t e , S t im . I M ■ i<k« Fsp 30- M I Un*CtwtSDWBv:ililB,sUjn stale-asleep stfmtavef-43db - 0 - c h a n n e l 1: A s le e p , 43* c h a n n e l 2: A s le e p , 43 d b - D - c h a n n e l 3: A s le e p , 43 d b - V - c h a n n e l 4: A s le e p , 43 d b 0 | V T T - | | I >» I | 1 I I I 1 I I I I | I | I | | I | 1 | ■ 1 | | | ■ | , , ! ■ ° 5 10BMiK 2 0 2 5 3 0 3 S 4 0 (nrWrrrTTTTi i i i | i i l i T '< ■ ■ i I i ■ i ■ I ■ 0 5 10 15 20 25 30 M o c k n o . Fig. 19 & 20. Number of Blocks Required for Fsp to Reach the Criterion of 3.1, with Subject Asleep and Stim. Levels at 38 and 43 dB. In Fig. 19 the Fs p values for subject 3, while sleeping at stimulus level 38 dB p.p.e. spl and for all channels are plotted versus the number of blocks. It is apparent how Fs p improves as the number of blocks increases. Of all channels, channel 3 and channel 1 reach the criterion (Fs p = 3.1). The Fs p rises significantly faster for channel 3. After only six blocks of averaging the criterion is reached. For channel 1 the criterion is reached after the 22nd block. Recordings from channels 2 and 4 never reach it. In Fig. 20 the Fs p values for subject 3, while sleeping, at stimulus level 43 dB p.p.e. spl and for all channels are plotted versus the number of blocks. 35 Fsp values are larger than in Fig. 19 and they also improve with increasing number of blocks. Channel 3 shows higher values, that rise faster than those of the other channels and the minimum quality value is reached only after 3 blocks of averaging. Fs p values rise very slowly for the recordings from channel 2, and reach the criterion after 38 blocks. Channels 1,3, and 4 reach the set Fs p value of 3.1 within the 9 first blocks. 4.0- MI Line Chart Ml Un* Chart Spill By: italM lfm State = Awake, Stim. Level=38dB - 0 - channel 1: Asleep, 53 db channel 2; Asleep, S 3 db - O - channel 3: Asleep, 53 db channel 4: Asleep, S 3 db 30: c hannel 2: A w a k e , 33 d b -D - charnel 3: A w a k e , 3 8 * channel 4: A w a k e , 3 8 * 2.0- 1.5- 1.0- 0.5' block no. TT-! 40 0.0' block no. Fig. 21 & 22. Number of Blocks Required for Fsp to Reach the Criterion of 3.1, with Subject Asleep/Awake and Stim. Levels at 53 and 38 dB. In Fig. 21 data corresponding to subject 3, sleeping state, stimulus level 53 dB p.p.e. spl and for all channels are plotted against the number of blocks. Recordings from channel 3 show higher Fs p values, inclining faster than the rest of channels. Criterion is reached within the first five blocks. Fs p values rise very slowly for the recordings from channel 2, and never reach the minimum quality 36 criterion. Channels 1 and 3 reach the set Fs p value of 3.1 within the 7 first blocks. Figures 22, 23, 24 refer to the same subject, but to the awake state. Fig. 22 presents Fs p values for subject 3, while awake, at stimulus level 38 dB p.p.e. spl and for all channels. All values are lower than the corresponding ones in Fig.19 and only channels 1 and 3 reach the criterion, but after 36 blocks of averaging. In Fig. 23 the Fs p values corresponding to subject 3, while awake, at stimulus level 43 dB p.p.e. spl and for all channels are higher than those computed for the recordings at 38 dB p.p.e. spl (Fig22). Ml Una Chart Spill By: tlate,ttbn Fsp $taie>awake stim level-43 f t 9 4 channel 1: A w ake , 4 3 * h- ch annel 2: A w a k e , 4 3 db “ 0 “ channel 3: A w ake, 4 3 d b channel 4: A w ake, 4 3 d b b lo c k n o . I I I I | I I I 1 | I I I I | I I I I | I I I I | » I I I | I i I I | I I 1 I Fsp ■ 8 - 7 - 6 S ' 3 2 10 15 20 25 30 35 40 Ml Una Chart Spill By: Male, stim stale-awake stim level* 53db - 0 - channel 1 ; Awake, 53 db channel 2; Aw ake, 53 db - 0 - channel 3: Awake, 53 db channel 4: Awake, 53 db Block No. rn tihi 1 1 1 ii i |i"i i n i i ti|n i i| hi 1 1 1 n 1 1 5 10 15 20 25 30 35 40 Fig. 23 & 24. Number of Blocks Required for Fsp to Reach the Criterion of 3.1, with Subject Awake and Stim. Levels at 43 and 53 dB. 37 However, they are lower than those in Fig. 20 (asleep state). Channels 1 and 3 reach the set Fs p value of 3.1 within the 9 first blocks, channel 3 within 15 blocks, whereas channel 2 never reaches it. Finally, in Fig. 24 data corresponding to subject 3, awake state, stimulus level 53 dB p.p.e. spl and for all channels are plotted against the number of blocks. Channels 1 reaches the set Fs p value of 3.1 at the 15th block, channel 3 within the 16th block, and channels 1 and 4 after 32 blocks. 4. Minimum “Quality” Estimation for Combination of Recording Channels In the previous section, “Minimum “quality” estimation”, it was demonstrated how the "quality" develops with increasing number of sweeps and at which block the minimum criterion was reached for “one-channel" recordings. At present, in both clinical use and research, only recordings from one channel are used for “quality" estimation of averaged auditory brain-stem responses. In this study, it is attempted to combine the Fs p values of independent channels to determine a joint function, which will estimate a common minimum quality for their ABR recordings. Therefore, after computing the Fs p values for every channel separately, the corresponding joint probability of all possible pairs of channels was computed. The results of this computation are included under the 38 columns conf 12, conf 13, conf 14, conf 23, conf 24, conf 34 in Table 2, which is representative of the formed databases used in the analysis. The results in each of these columns correspond to the joint probability of the channel combinations 1 with 2,1 with 3, 1 with 4, 2 with 3,2 with 4, 3 with 4 respectively. In columns crit12, crit13, crit14, crit23, crit24, crit34 the Os correspond to a probability of positive detection for the ABR less than 0.99, which corresponds to the set criterion of Fs p = 3.1 and the 1s to a probability of positive detection for the ABR equal or more than 0.99. The number of block at which a 1 occurs for the first time at each stimulus level and state of relaxation is noted. At the end, the mean values over all subjects for these number of blocks are plotted against stimulus level, at each state of relaxation. Note: In the following plots, a value equal to 40 for the number of blocks is chosen to indicate that the set criterion was never reached during the averaging over the 39 blocks. Figures 25 through 32 refer to the asleep state. In Fig. 25 the results for all single channels are presented. Recordings from channel 2 reach the criterion slower than the recordings from the rest of the channels. More specifically, at 38 dB p.p.e. spl the criterion is never reached for channel 2. Among channels 1, 3, and 4, channel 3 reaches the criterion faster followed closely by channel 4. Recordings from channel 4 differ significantly from corresponding ones from channels 1 and 3, at the 43 dB p.p.e. spl. Channels 1 and 3 reach the confidence level within 11 blocks. At the highest stimulus level, 53 dB p.p.e. spl, channels 1,3,4 reach the set criterion within 9 blocks. At the no stimulus level none of the channels reaches it, as expected, since at this level the response we get is only due to the background noise. Results in section 3 seem in agreement with the present results. no. * blocks 2 0 chi: a s le e p -Q - ch 2: a s le e p c h 3: a s le e p -V - ch 4: a s le e p stim level dB no stim 3 8 4 3 33 45 4 0 3 5 3 0 b lo c k s n o . 20 1 5 1 0 5 0 Li ne Chart Split By: st at e 0.99 conf . level chi: a s l e e p -O - o itl2 : a s l e e p critl3: a s l e e p -9 - crit!4: a s l e e p s tim le v e l no s u m 3 8 4 3 5 3 Fig. 25 & 26. Number of Blocks Required to Reach the Criterion of 3.1 with 99% Confidence Level for Individual and Combinations of Channels In Fig. 26 the results for channel 1 are plotted for comparison in the same plot with the results from its combinations with the rest of the channels. Thus, crit12, crit13, crit14, represent the results from combination of channel 1 with 2, 3, 4 40 respectively. At the stimulus levels of 38, 43, 53 dB p.p.e. spl all combinations seem to reach the criterion faster than the single channel 1. Especially, crit13 and crit14 converge to it faster. At 43 dB p.p.e. spl they reach the Fs p - 3.1, within the first five blocks. Line Chart Split By: st at e 0.99 conf. level -§ - chl'aslecp 2! -0- critll* a s le e p DO* *v -W - criil3: a s ]e e p -9 - cril24: a s le e p ji Line Chart Split By: state 0.99 conf. level stim level dB 53 n o s tim no stim 38 43 53 Fig. 27 & 28. Number of Blocks Required to Reach the Criterion of 3.1 with 99% Confidence Level for Individual and Combinations of Channels Fig. 27 presents the channel 2 and its combinations with channels 1,3,4. The recordings from the single channel 2 reach the Fs p = 3.1 slowly. However, crit12, crit23, crit24 converge a lot faster to this value. At the 43 dB p.p.e. spl less than 10 blocks of averaging are needed and at 53 dB p.p.e. spl less than 6 blocks of averaging. At 38 dB p.p.e. spl all three combinations need more than 22 blocks of averaging. 41 In Fig. 28 the results for channel 3 are plotted for comparison in the same plot with the results from its combinations with the rest of the channels. Crit13, crit23, and crit34, represent the results from combination of channel 3 with 1, 2, and 4 respectively. At the stimulus levels of 38, 43 and 53 dB p.p.e. spl crit13 and crit34 demonstrate the fastest convergence to the criterion. Crit23 follows closely channel 3. At 53 dB p.p.e spl the results for channel 3, crit13, crit23, crit34 are almost identical. 45: 4 0 : 35: 30: 2* blocks no. 20- 1* lo j 5 0 Line Char t Split B y: state 0 . 9 9 conf. lev e l n o s t u n stim level dB 3 8 *♦ * ch 4: a s le e p -O - oitl4: a s le e p -9- ait24: a s le e p J j- crit34; a s le e p 45 40 3 5 3 0 2 5 blocks “ ■ 2 0 i 1 5 10 5 0 L ine Chut Spl it By: s ta te 0 . 9 9 c o o f. le v e l ciit!2: a s le e p -D~ critI3: a s le e p - 9- critl4: a s le e p - 9 - ait23: a s le e p crit24: a s le e p crit34: a s le e p 4 3 5 3 no s tu n stim level dB 3 8 4 3 5 3 Fig. 29 & 30. Number of Blocks Required to Reach the Criterion of 3.1 with 99% Confidence Level for Individual and Combinations of Channels In Fig. 29 results for channel 4, crit14, crit24, and crit34 are presented. At 38 dB p.p.e. spl crit34 reaches the Fs p = 3.1 faster and within 13 blocks, whereas 42 channel 4 and crit24 reach it after 26 blocks and crit14 after 20 blocks. At 43 dB p.p.e. spl crit14 reaches the criterion within 4 blocks, crit34 within 6 blocks, crit24 within 10 blocks and channel 4 within 24 blocks. At the highest stimulus level, crit14, crit24 and crit34 reach it within 5 blocks, whereas channel 4 within 9 blocks. Fig. 30 includes the results only from the combinations between the single channels in an attempt to show which combination is the best. Apparently, at 53 dB p.p.e. spl crit12, crit13, crit14, crit 23, crit 24, and crit34 are very close to each other, with crit. 14 the best among them. At 43 dB p.p.e spl crit13 and crit14 are identical and reach criterion within 3 blocks, crit34 and crit12 reach it within 7 blocks and at last crit23, crit24 reach it within 9 blocks. At 38 dB p.p.e. spl crit34 is the best, reaching the Fs p = 3.1 in 13 blocks. Crit13 reaches it within 18 blocks, crit14 within 20 blocks, crit12 and crit 23 within 24 blocks, and crit24 within 27 blocks. Over all stimulus levels it seems that crit13, crit14, crit34 reach the set criterion faster. In Fig.31 results for the single channels 1, 2, 3, 4 are depicted together with the results for crit13, crit14, crit34 which as it was shown in Fig. 26 are the combinations that reach the 99% confidence level faster. At 38 dB p.p.e. spl there is a substantial difference between the results from the single channels and those for crit13, crit14, crit34. The single channels reach the confidence 43 level slower than the depicted combinations. Crit34 is the best among all. At 43 dB p.p.e. spl crit14 and crit13 reach criterion within 3 blocks, crit34 within 6 blocks. Single channels 1 and 3 reach it within 11 blocks. 40 35 3(rf 25 bloc k no. 20 1 5 1 0 5 0 chi: a s l e e p -O - c h 2: a s l e e p c h 3: a s le e p -V - ch 4: a s le e p - + - critl3: a s le e p -M - criil4: a s l e e p - 0 - cri t34; a s l e e p Li ne Chart S p G t By: state 0.99 conf. l evel stim level dB n o stun 3 8 4 3 5 3 4 5 40 3 5 30 „ 2 5 bloc k no . 20 1 5 10 5 0 Li ne Chut Spilt By: st at e 0. 99 conf . level critl3: a s l e e p o itl4 : a s l e e p ait34: a s l e e p s d m level d B ■ 1 I n o s u m 3 8 4 3 I 5 3 Fig. 31 & 32. Number of Blocks Required to Reach the Criterion of 3.1 with 99% Confidence Level for Individual and Combinations of Channels Finally, at 53 dB p.p.e spl crit14, crit34, crit13, channel 1, channel 3 and channel 4 reach Fs p = 3.1 within 9 blocks. For this case the combinations don’t differ much from channels 1, 3 and 4. Reviewing this figure crit13, crit14, crit34 reach to the 99% confidence level faster than the single channels. Lastly, in Fig. 32 the results for crit13, crit14, crit34 are depicted. It has been concluded by observing the previous figures, that these combinations reach faster to the set criterion than the single channels and the rest of the combinations. At 38 dB 44 p.p.e. spl crit34 seems to converge to Fs p = 3.1 faster and within 12 blocks. At 43 dB p.p.e. spl crit14 and crit13 reach criterion within 3 blocks, crit. 34 within 6 blocks. At 53 dB p.p.e spl crit14, crit34, crit13 are almost identical. Crit14 is slightly faster. For all stimulus levels crit. 34 seems to be the most satisfactory, since it performs well at the lowest stimulus levels, the 38 dB p.p.e spl. lin t Ch a rt Split Bj; stale 0.9 9 oaf Jtvtl d f= S Line Chart Spli t By: state 0. 99 co n f. leve l df»5 5.00 ■ s ti m leve l d B 0.00 5 3 n o stim 4 3 n o s u m 3 8 Fig. 33 & 34. Number of Blocks Required to Reach the Criterion of 3.1 with 99% Confidence Level for Individual and Combinations of Channels Figures 33 through 40 refer to the awake state. In Fig. 33 the results for all single channels are presented. Recordings from channel 2 reach the criterion slower than the recordings from the rest of the channels. At 38 dB p.p.e. spl the criterion is never reached for channel 2. Among channels 1, 3, and 4, channel 45 1 reaches the criterion faster and within 32 blocks. At the 43 dB p.p.e. spl channel 1 reaches the confidence level within 14 blocks, channel 3 within 21 blocks, channel 4 within 29 blocks, and channel 2 within 39 blocks. At the highest stimulus level, 53 dB p.p.e. spl, channels 1 and 3 reach the set criterion within 12 blocks, channel 4 after 23 blocks, and channel 1 after 38 blocks. At the no stimulus level none of the channels reaches it, as expected, since at this level the response we get is only due to the background noise. Overall channel 1 performs best. Comparing Fig. 33 with Fig. 25 we notice that for the same channels and same stimulus levels the corresponding values are higher for the awake state. This was expected, since in the latter state more background noise it is introduced. In Fig. 34 the results for channel 1 are plotted for comparison in the same plot with the results from its combinations with the rest of the channels. At the stimulus levels of 38, 43, 53 dB p.p.e. spl all combinations seem to reach the criterion faster than the single channel 1. At 38 dB p.p.e. spl crit12, crit13, and crit14 reach the Fs p = 3.1, within the 23 blocks. At 43 dB p.p.e. sp! crit12, crit14 reach it within 12 blocks and crit13 within 8 blocks. At 53 dB p.p.e. spl, crit12, crit13, and crit14 reach it within 6 blocks. Again comparing the results at the stimulus levels of 38, 43, 53 dB p.p.e. with the corresponding in Fig. 26 , we see that for the awake state criterion is reached slower than in the sleeping state. 46 Fig. 35 presents the channel 2 and its combinations with channels 1,3,4. The recordings from the single channel 2 reach the Fs p = 3.1 slowly, whereas crit12, crit23, crit24 converge a lot faster to this value. At 38 dB p.p.e. spl crit24 reaches the set value after 26 blocks, crit23 after 23 blocks, and crit 12 within 19 blocks. At the 43 dB p.p.e. spl crit24 reaches the confidence level within 27 blocks, and crit12 within 12 blocks and crit23 within 14 blocks. At 53 dB p.p.e. spl less than 6 blocks of averaging are needed for crit23 and crit12, and crit24 needs 12 blocks. Line Chart Split By: state O .99coflf.level df=5 35- 30- - t - ch 3: a w a k e -O ' oitl3: a w a k e crit23: a w a k e -9 - ciit34: a w a k e Li ne Char t Split B j: st ate 0 .9 9 coat le v e l d f < * 5 stim level dB stim level dB 3 3 Fig. 35 & 36. Number of Blocks Required to Reach the Criterion of 3.1 with 99% Confidence Level for Individual and Combinations of Channels In Fig. 36 the results for channel 3 are plotted for comparison in the same plot with the results from its combinations with the rest of the channels. At the stimulus levels of 38, 43 and 53 dB p.p.e. spl crit13, crit34 and crit23 47 demonstrate the faster convergence to the criterion. At 38 dB p.p.e. spl crit34 converges to it after 25 blocks, crit23 and critl 3 converge to it after 22 blocks. At 43 dB p.p.e. spl crit23 reaches the preset “quality” after 13 blocks, crit34 within 10 blocks and c ritl3 within 7 blocks. At 53 dB p.p.e spl the results for critl 3, crit23, crit34 are almost identical. They reach criterion within 7 blocks. Fig. 37 presents the results for channel 4, c ritl4, crit24, and crit 34. At 38 dB p.p.e. spl c ritl4 reaches the Fs p = 3.1 faster and within 21 blocks, channel 4 and crit24 reach it after 22 blocks, and critl 4 and crit34 within 10 blocks. At 43 dB p.p.e.spl c ritl4 and crit34 reach the criterion within 11 blocks, crit24 and channel 45t Lbe Chart SpBtBj:itate 0.99 cocif. level dfcj m 25- 8 ® 20- cbl: a w a k e crit!3: a w a k e O - ch2:a w a k e cml4:awake ch 3: a w a k e -c - cri(23: a w a k e d o torn 3 3 4 3 5 3 3 8 4 3 5 3 stim level Fig. 37 & 38. Number of Blocks Required to Reach the Criterion of 3.1 with 99% Confidence Level for Individual and Combinations of Channels 4 within 28 blocks. At the highest stimulus level, critl 4 and crit34 reach it within 7 blocks, crit24 within 12 blocks and channel 4 within 22 blocks. C ritl 4 shows a better overall performance. Fig. 38 includes the results from all combinations between the single channels in an attempt to show which combination is the best. At 53 dB p.p.e. spl crit 12, critl 3, c ritl4, crit23, and crit34 are almost identical, with critl 3 the best among them. Crit24 shows the slowest convergence to the confidence level. At 43 dB p.p.e spl crit 12 reaches criterion within 12 blocks, c ritl4 within 11 blocks, crit34 reaches it after 11 blocks, critl 3 within 7 blocks, crit23 after 13 blocks and crit24 after 28 blocks. At 38 dB p.p.e. spl crit12 reaches the Fs p = 3.1 in 19 blocks, critl 4 reaches it within 21 blocks, critl 3 and crit23 within 23 blocks, crit34 after 24 blocks and crit24 within 26 blocks. Over all stimulus levels it seems that critl 3, critl 4, crit34 reach the set criterion faster. In Fig. 39 results for the single channels 1, 2, 3, 4 are depicted together with the results for c ritl3, crit14, crit34 and crit23. At 38 dB p.p.e. spf there is a substantial difference between the results from the single channels and those for critl 3, critl 4, crit34 and crit23. The single channels reach the confidence level slower than the depicted combinations. C ritl4 is the best among all. At 43 49 dB p.p.e. spl critl 3 reaches criterion within 7 blocks, crit34 within 10 blocks, critl 4 within 11 blocks and crit23 within 13 blocks. Channel 1 reaches the confidence level within 14 blocks, channel 3 within 19 blocks, channel 4 within 24 blocks, and channel 2 within 34 blocks. Finally, at 53 dB p.p.e spl crit. 14, crit. 34, crit. 13 reach it within 7 blocks, channel 1 reaches it within 14, channel 3 in 21 blocks, channel 4 within 29 blocks and channel 2 reaches it after 38 blocks. Consequently, critl 3, critl 4, crit34 and crit23 reach to the 99% confidence level faster than the single channels. line Chart Split B y : state ,30- ,30- ■ B 2 5 - Line Chart Split By: stale 0.9 9 c o n fJ e v 'e l df*5 1 0 - s tim level dB oo s tim Fig. 39 & 40. Number of Blocks Required to Reach the Criterion of 3.1 with 99% Confidence Level for Individual and Combinations of Channels Finally, Fig. 40 depicts the results for critl 3, critl 4, crit34 and crit23. It has been concluded by observing the previous figures, that these combinations reach 50 faster to the preset criterion than the single channels and the rest of the combinations. At 38 dB p.p.e. spl critl 4 seems to converge to Fs p = 3.1 faster and within 20 blocks. At 43 dB p.p.e. spl critl 3 reaches criterion within 7 blocks, crit34 within 10 blocks, critl 4 within 11 blocks and crit23 within 13 blocks. At 53 dB p.p.e spl c ritl4, crit34, c ritl3 are almost identical. C ritl3 is slightly faster. For all stimulus levels critl 3, critl 4, crit34 all perform almost as well. 5. Crosscorrelation The crosscorrelation values of the averaged ABR's between channels are included in Table 3 in Appendix A. Among all channels channel 1 and channel 3 have crosscorrelation coefficient above 0.8 which indicates that they are highly correlated. This result can be explained by the positions of their corresponding electrodes. Both sets of electrodes are picking up almost the same activity. Consequently, channels 1 and 3 can not be used as independent channels. 51 DISCUSSION The recording of the ongoing auditory brain stem activity is composed of the actual EP and the averaged background noise. Whether an ABR exists in the recording depends on the degree of stimulation. Its absence or presence can be detected by using the Fs p method, following other methods used for eliminating as much as possible the background noise, or equivalently improving the SNR. Since the frequency composition of the background noise and ABR overlap, filtering schemes can not improve significantly the SNR. However, by averaging the responses the noise is reduced. The longer the averaging the more the reduction in noise (Fig. 13- Fig. 18). Also the state of arousal can reduce the noise levels. During the sleeping state the background noise was reduced compared to the noise levels while the subject was awake (Fig. 13- Fig. 18). A high and constant noise level would prolong the time of testing , because more averaging would be required to ensure a confident visual or statistical detection of an evoked response. Although, standard averaging succeeds in decreasing the effect of stationary background noise, it can not compensate for episodic increase in noise. Such 52 an increase would require a lot of additional averaging to reduce the residual noise. The Bayesian averaging is employed in order to minimize the effect of non stationary noise. This technique reduces the residual noise compared to the standard averaging and, thus improves the SNR in the case of episodic noise bursts {Fig. 11, Fig. 12). Also, with the Bayesian weighting technique the desired detection criterion utilized by the method will be achieved faster than with the standard averaging (Fig. 9, Fig. 10). At the no stimulus level none of the channels reaches it, as expected, since at this level the response we get is only due to the background noise. In employing the averaging techniques, it was assumed that the EP is a deterministic signal independent of the BN and therefore is simply superimposed on it. The BN is assumed to be a Gaussian ergodic process. Since during the recordings there were instances of non stationary background noise, the assumptions are not strictly fulfilled. More complicated methods of analyses can be employed to fulfill the assumptions (Bendat). In employing the minimum Quality’ method it was assumed the variance ratio VAR(BN)/VAR(SP) follows the F-distribution, F(vi, ■ & ), with Vi=5 and v2 =256 the degrees of freedom of the numerator and the denominator. The value of v2 =256 corresponds to the number of ensembles, and therefore to 256 independent values used for the evaluation of the second order statistical terms. The value of vi=5 corresponds to the number of independent samples in every ensemble, and was estimated as such by Elberling and Don (1984). This assumptions is not strictly fulfilled as it is suggested by the spurious results when none stimulus is presented. When no stimulus is presented none of the channels or their combinations should reach the preset criterion that ensures the presence of ABR, since at this level the response we get is only due to the background noise. 45 1 4 0 1 35' 30- 1 25' ■ 2 0 is le s' 0‘ - 0 - critl 2: awake - Q - critl 3: awake - A - critl 4: awake - 0 ~ erit23: awake crit34: awake Line Chart, Split by: State 0.99 com. laval df-2 no s tim 35 stim level dB 43 45 40 35 30 8 " 1“ is 10 ' S ' 0 U n e C h a rt S p ilt B y : state 0.99 cortf. level (Jf-2 **0 - met 2: asleep critl 3: asleep critl 4: asleep ‘- V - oit23: asleep — erit24: asleep H I - oit34: asleep stim level dB 5 3 no stim 39 43 S3 Fig. 41 & 42. Number of Blocks Required to Reach the Criterion of 3.1 with 99% Confidence Level for Individual and Combinations of Channels 54 However, in Fig. 26 to 32, and Fig. 34 to 40 there are incidences at which the criterion is reached although no stimulus is present. This occurrences comprise an acceptable percentage of the total population, when the rate of false positive is set to 1% ( confidence level 99%). A new assumption for v1 could improve the rate of spurious results. Its new value will be used to evaluate again the corresponding probabilities to F(vi, v2 ). The results for Vi -2 are presented in Fig. 41 and Fig. 42. The spurious results decreased. For both states critl 4 seems the best combination among all. We notice that for the sleeping state the combination critl 4 does not present any unexpected results at no stimulus. The same observation is also made in Fig. 28, for Vi=5. In Fig. 30 and Fig. 42 it seems that a spurious result occurred in combinations of channels with channel 3. This indicates that the higher rate of false positive for v1 =5 is due to a mistake in channel 3 that propagates. Thus, the results for v1=5 are considered to be acceptable and within confidence limits. Since the combination of channel 1 with channel 3, i.e. critl 3 can not be used due to the high crosscorrelation between the channels, critl 4 (Fig. 32,40,41,42) seems to have the most overall satisfactory response. Fig. 25 and Fig. 33 show that channel 1 reaches criterion faster than channel 4. Thus, by comparing channel 1 to the combination critl 4 it is concluded that the time saving is over 30%, which is very significant. For the above conclusions to be valid, it had to be presumed that the channels are independent. This assumption is presently further investigated at the Electrophysiology Lab of House Ear Institute. 56 V. REFERENCES 1. Bandat, J. S. 1964. Mathematical analysis of average response values for non-stationary data. IEEE Trans on Bio-Medical Engineering, BME, July 1964:72-81. 2. Don, M., Elberling, C. (1994). Evaluating residual background noise in human auditory brain-stern responses. J. Acoust. Soc. Am., 96, No. 5, Pt. 1. 3. Don, M., Elberling, C., and Waring, M.W. (1984). Objective detection of averaged auditory brainstem responses. Scand. Audio. 13, 219-228. 4. Elberling, C., and Don, M.(1984). Quality estimation of averaged auditory brain stem responses. Scand. Audiol., 13, 187-197. 5. Elberling, C., and Don, M. (1987a). Threshold characteristics of the human auditory brainstem response. J. Acoust. Soc. Am., 81, 115-121. 6. Elberling, C., and Wahlgreen, O. (1985). Estimation of auditory brainstem responses, ABR, by means of Bayesian inference. Scand. Audiol. 14, 89- 96. 7. Hall III, J., W. HandBook of auditory evoked responses. Massachusetts: Allyn and Bacon, 1992. 8. Moore, E., J., ed. 1983. Bases of auditory brain-stern evoked responses. New York : Grune and Stratton. 9. Pickles, J., O. An introduction to the physiology of hearing. 2d ed. San Diego: Academic press, 1988. 57 10. Sayers, B., Ruggiero, C., and Feuerlight, J. (1981). 'Statistical variability of biomedical data: Parti. The influence of serial correlation on mean value estimates," Med. Inform., 6, no.1, 1-11. 11. Sayers, B., Ruggiero, C., and Feuerlight, J. (1981). ‘Statistical variability of biomedical data: Part2. The influence of serial correlation on power estimates and on comparative testing of samples,” Med. Inform, no. 3, 207-220. Table 1. List of Abbreviations T E R M D E F I N I T I O N AN V U I (eighth, auditory) cranial nerve AICA anterior inferior cerebellar artery PICA posterior inferior cerebellar artery MCA middle cerebral artery PCA posterior cerebral artery CN cochlear nucleus VCN ventral cochlear nucleus AVCN anterior ventral cochlear nucleus PVCN I posterior ventral cochlear nucleus DCN dorsal cochlear nucleus OCB olivo-cochlear bundle COCB crossed olivo-cochlear bundle UCOCB uncrossed olivo-cochlear bundle AS acoustic stria SM stria of Moiukow (dorsal acoustic stria) IAS intermediate acoustic stria SH stria of Meld (intermediate acoustic stria) DAS dorsal acoustic stria TB trapezoid body MTB medial trapezoid body SOC superior olivary complex ISO lateral superior olive MSO medial superior olive MCP middle cerebellar peduncle I L L . lateral lemniscus NLL nucleus of the lateral lemniscus LLV ventral nucleus of the lateral lemniscus LLD dorsal nucleus of the lateral lemniscus DCLL dorsal commissure of the lateral lemnsicus CP commissure of Probst IC inferior collicrulus CIC commissure of the inferior coiliculus PC paracentral nuclei cx cortex of inferior coiliculus TA tegmental areas BIC brachium of the inferior coiliculus MGB medial geniculate body (of thalamus) . VMCB ventral medial geniculate body DMGB dorsal medial geniculate body AR auditory radiation to cortex AC auditory cortex A I primary auditory cortex (koniocortex) in monkey and cat a n secondary auditory cortex in monkey and cat H P ectosylvian posterior area INS insula area TL temporal lobe STP superior temporal plane CC corpus callosum MTL medial temporal lobe •tile aUm block no. channel 1 Channel 2 charnel 2 channel 4 PI P2 P3 P4 dll PI dll P2 dll P3 dil P4 conl 12 I Awake no slim t .922 .605 .861 2.182 .533 .304 .362 .943 .467 .096 .638 ,057 .675 2 Awake no stim 2 .537 .683 1.340 2.770 .252 .303 .752 .082 .746 .637 .246 .016 .624 3 Awaka no trim 3 .571 1.101 1.230 2.217 .276 .640 .709 .947 .722 .360 .291 ,053 .740 4 Awaka no silm 4 .966 1.030 .655 1.749 .573 .600 .342 .676 .427 .400 .656 .124 .829 5 Awaka no ailm 5 1.045 1.116 .793 .840 .608 .646 .444 .476 .392 .352 .556 .522 .602 A Awaka no Kim 6 1.061 t493 1,010 .437 .617 .210 .566 .177 .363 .761 .412 .023 .701 7 Awaka no ailm 7 1.661 .619 2.287 .429 .656 .315 .953 .172 .144 .665 .047 .626 .901 A Awaka no slim 6 1 .306 .632 1.546 .761 .730 .325 .624 .421 .261 .676 .176 .579 .624 ft Awaka no aiim 0 1.585 !tSS 1.387 .803 .635 .419 .770 .302 .165 .561 .230 .696 .904 1 O Awaka no ailm io 1.630 .741 1.603 1.029 .648 .407 .640 .899 .162 .593 .160 .401 .910 11 Awaka no ailm 1 1 1.654 .625 1.585 .007 .854 .319 .835 .523 .146 .661 .168 .477 .900 1 2 Awaka no stim 12 1.370 .677 1.529 .664 .764 .359 .610 .404 .236 .841 .181 .500 .649 13 Awaka no slim 13 1,36ft .SI S 1 .233 .626 .771 .235 .706 .469 .229 .765 .204 .531 .625 14 Awaka no atkn 14 1.216 .523 1.068 ,706 .698 .241 .631 .361 .302 .759 .369 .619 .771 IS Awaka no altm 15 .076 .456 .876 .639 .567 .193 .504 .477 .433 .607 .406 .523 .651 16 Awaka no Blktl t 6 .686 .642 .634 .684 .366 .332 .250 .506 .634 .666 .750 ,402 .575 17 Awaka no e tJm 1 7 .792 .632 .624 .666 .444 .472 .316 .497 .556 .626 .662 .503 .706 16 Awaka no Bkn 1 6 .696 .774 .662 .759 .373 .431 .347 .420 .627 .560 ,663 .560 .643 1* Awaka no stim 1ft .747 .614 .645 .656 .411 .489 .334 .343 .569 .541 .666 .657 ,602 20 Awaka no ailm 20 .669 .033 .636 .566 .366 .540 .329 .274 .632 .460 .671 .726 .709 Awaka no stfm 21 .763 .736 .668 .636 .423 .404 .352 .328 .677 .596 .646 .672 .656 22 Awaka no sUm 22 .760 .909 .862 .533 .442 .522 .363 .249 .556 .476 .637 .751 .733 23 Awaka no skm 23 .631 1.116 .746 .044 .471 .649 .410 .334 .829 .361 .600 .666 .614 24 Awaka no slim 24 .925 1.133 .765 ,737 .835 .657 .430 .404 .465 .343 .501 .596 .640 25 Awaka no silm 25 .021 1.083 .887 .669 .532 .630 .510 .353 .468 .370 .490 .647 .627 26 Awaka no ailm 26 .054 .936 .674 .595 .853 .543 .601 .296 .447 .457 .409 ,704 .790 27 Awaka no slim 27 .041 .663 ,794 .612 .545 .507 .445 .309 .455 .493 .555 .691 .776 26 Awaka no silm 2 B .068 .676 .762 .669 .582 .604 .437 .353 .436 .496 .583 .847 .763 20 Awaka no aim 29 .960 .920 .739 .646 .570 .532 .404 .338 .430 .466 .596 .668 .706 30 Awaka no stim 30 1.041 .070 .695 .823 .606 .563 .372 .466 .304 .437 .626 .534 .620 31 Awaka no t in 31 ,95ft .069 .668 .669 .566 .556 .352 ,353 .444 .444 .646 .647 .803 32 Awaka no slim 32 .701 .666 .636 .664 .443 .500 .329 .364 .557 .491 .671 .630 .727 33 Awaka no silm 33 .072 .761 .851 .601 .565 .421 .339 .369 .435 .579 .681 .631 .748 34 Awaka no atlm 34 1.026 ,676 .626 .771 .597 .360 .320 .42ft .403 .640 .680 .571 .742 35 Awaka no slim 35 1.110 .645 .625 .644 .645 .334 .319 .460 .385 .666 .661 .520 .763 36 Awaka no atlm 36 1.149 .634 .749 .735 .665 .326 .413 .402 .335 .674 .567 .596 .774 37 Awaka no atlm 37 1.007 .721 .677 .762 .666 .392 .359 .437 .4 14 .606 .641 .563 ,746 36 Awaka no atlm 36 .676 .693 .644 .775 .602 .371 .334 .431 .4 98 .629 .666 .569 .667 3ft Awaka no aiim 3ft .620 ,680 ,615 .704 ,464 .361 .312 .379 .536 .639 .688 .621 .657 40 Awaka 36 db 1 .440 .325 .870 1.422 .166 .102 .360 .763 .814 .096 .640 .217 .269 4 1 Awaka 36 cfa 2 .490 1.360 .264 2.063 .216 .760 .066 .929 .764 .240 .932 .071 .612 42 Awaka 36 db 3 .300 .746 .486 2.351 .142 .412 .213 .950 .658 .566 .767 .041 .495 43 Awaka 36 c to 4 .613 1.132 .636 2.008 .310 .656 .329 .022 .600 .344 .671 .076 .703 44 Awaka 36 db 5 .474 1.206 .474 1.201 .204 .693 .204 .732 .796 .307 .706 .266 .756 45 Awaka 38 c ft> 6 .453 1.466 .696 1 .203 .16ft .79ft .373 .692 .611 .201 .627 .309 .837 46 Awaka 36 e fts 7 .SB 1 2.214 .770 1.162 .293 .947 .420 .682 .707 .053 .572 .316 .962 47 Awaka 36 c *a a .021 1.629 .865 1.147 .532 .610 .606 .664 .468 .161 .492 .336 .915 46 Awaka 36 db a 1.067 1.295 .765 1.201 .632 .734 .430 .691 .366 .266 .661 .309 .902 4ft Awaka 36 d> io .957 1.530 .840 .906 .555 .019 .478 .570 .445 .161 .522 .421 .920 S O Awaka 36 db 11 .009 1.451 .617 .641 .524 .794 .462 .476 .476 .206 .538 .522 .902 51 Awaka 36 C fl> 12 .007 1 .S O B .766 1.023 .523 .813 .425 .505 .477 .187 .573 .405 .911 Table 2, Page 1, Left. Database for Estimating the Q ualify C riterion Ul 'O slat* atlm block no. channel t channel 2 channel 3 channel 4 PI P2 P3 P4 dil PI dif P2 dif P3 dll P4 conl 12 S2 Awake 36 < t> 1 3 .954 1.360 .639 1.065 .553 .771 .330 .631 .447 .229 .670 .369 .898 S3 Awake 36 c ± > 1 4 .925 1.193 .977 1.224 .505 .687 .508 .702 .465 .313 .432 .298 .654 54 Awake 36 ct» 1 5 .647 1.426 .959 1.441 .462 .705 .558 .790 .518 .215 .444 .210 .009 SS Awake 36 cft) 1 6 .904 1.682 .993 1.329 521 .650 .578 .748 .479 .144 .422 .252 .931 S O Awake 36 db 1 7 1.010 1.972 1.16S 1 .682 .593 .917 .873 .856 .407 ,083 .327 .144 .986 57 Awak* 38 (6 te .093 1.984 1.063 1.791 .514 .916 .630 .005 .486 .004 .370 .115 .959 58 Await* 38 C t> 1 9 .659 1.661 .934 1 .841 .401 .690 .541 .850 .509 .102 .459 .150 .948 S O Await* 38 d> 20 .955 2.282 1.246 1 .833 .554 .953 .713 .054 .446 .047 .287 .146 .979 60 Awake 30 db 21 .792 2.015 .902 1.365 .444 .923 .571 .762 .555 .077 .429 .238 .957 6* Await* 38 <*) 22 .801 2.073 1.126 1.321 .450 .931 .653 .744 .950 .069 ,347 .256 .962 82 Awake 38 db 23 .611 1.046 1.064 1.120 .457 .913 .630 .850 .543 .067 .370 .350 .953 83 Await* 36 e ft) 24 .763 2.061 1.007 1 .355 .423 .932 .560 .756 .577 .060 .414 .242 .961 64 Awake 36 c ft> 25 .869 2.356 1.048 1.270 .490 .959 .810 .725 .502 .041 .390 .275 .979 65 Await* 36 (ft) 26 .995 2.355 1.110 1.363 .679 .959 .649 .701 .421 .041 .351 .239 .903 66 Await* 36 db 27 1.127 2.506 1.052 1.156 .654 .974 .612 .669 .346 .028 .385 .331 .991 67 Awaka 36 db 28 1.214 2.754 1.124 1 .176 .697 .961 .852 .679 .303 .019 .346 .321 .994 66 Awake 36 dt 29 1.160 2.400 1.020 1.053 .661 .963 .590 .613 .319 .037 .402 .367 .988 69 Awaka 36 c ft> 30 1.171 2.246 .600 1.136 .676 .950 .449 .650 .324 .050 .581 .342 .984 70 Awake 36 e ft) 31 1,100 2.251 .792 .980 .665 .950 .044 .670 .315 .050 .556 .430 .984 71 Awaka 36 e ft) 32 1.255 2.306 .895 .903 .716 .055 .372 .620 .204 .046 .626 .460 .987 72 Awaka 36 db 33 1.239 2.384 .732 1.041 .709 .960 ,400 .606 .291 .040 .600 .394 .968 73 Awakf 36 C ft) 34 1.322 1.620 .737 .959 ,745 .892 .404 :55S .255 .108 .590 .444 .973 74 Awaka 36 cft) 35 1.277 1.737 .599 1.020 .728 .073 .299 .594 .274 .127 .701 .405 .965 7S Awaka 36 db 38 1.400 1.519 .580 1.1 15 .775 .616 .285 .647 .225 .164 .715 .353 .959 78 Awaka 38 c ft» 37 1.706 1.570 .663 1.149 .666 .633 .346 .665 .134 .167 .652 .335 .978 77 Awaka 36 < ft> 36 1.746 1.547 .754 1.153 .675 .024 .416 .667 .125 .170 .584 .333 .978 78 Awaka 36 c fta 39 1.666 1.397 .654 1.239 .903 .774 .467 .709 .097 .226 .513 .291 .978 79 Awaka 43 c ft> 1 .269 .414 .613 1 .091 .061 .161 .459 .634 .919 .839 .641 .366 .229 80 Awaka 43 db 2 .655 _ . ,327 .693 .743 .486 ,244 .371 .400 .512 .750 .629 .592 .613 8 1 Awaka 43 db 3 ,722 .401 .737 .516 .393 .210 .404 .237 .607 .790 .696 .783 .520 82 Awaka 43 c ft) 4 .796 .221 1.086 .506 .447 .047 .621 .230 .553 .953 .379 .770 .472 63 Awaka 43 (ft) S .934 .363 1.015 .555 .641 .128 .591 .266 .459 .674 .409 ,734 .599 64 Awaka 43 cft) 6 1.224 .527 1.O07 .703 .702 .244 .621 .376 .296 .756 .379 .622 .775 65 Awaka 43 db 7 1.169 .451 .937 .736 .675 .168 .543 .403 .325 .612 .457 .597 .730 66 Awaka 43 c ft) 0 .680 .790 .624 .799 .511 .448 .486 .449 .489 .552 .534 .551 .730 67 Awaka 43 cft) 0 .793 .727 .755 .721 .444 .390 .417 .302 .556 .004 .503 .606 .665 68 Awaka 43 db 10 .760 .755 .956 .006 .439 .417 .555 .455 .561 .503 .445 .545 .873 89 Awaka 43 c ft> 11 .693 .066 .736 .933 .514 .569 .404 .640 .486 .491 .596 .460 .761 90 Awaka 43 cft) 12 1.141 1.054 .726 .710 .061 .612 .396 .390 .330 .398 .604 .610 .800 91 Awak* 43 c ft) 13 1.501 .035 .950 .006 .810 ,541 .556 .454 .190 .459 I4 4 4 .546 .913 *2 Awak* 43 c ft) 14 1.520 1.424 1.273 .608 .819 .784 .724 .455 .161 .216 .276 .545 .961 93 Awak* 43 cft) 1 5 1.692 1.470 1.649 .904 .863 .002 .096 .521 .137 .198 .104 .479 .973 04 Awak* 43 cft) 16 2.003 1.650 2.432 .937 .933 .655 .968 .543 .067 .145 .036 .4 57 .990 O S Awak* 43 db 17 2.452 1.805 2.913 1.047 .966 ,641 .966 .609 .034 .159 .014 ,391 .995 96 Awaka 43 c ft) 16 2.909 1.544 3.506 1.322 .960 .623 .998 .745 .012 .177 3.737E-3 ,255 .996 97 Awak* 43 c ft> 1 a 3.1 29 1 .407 3.704 1 .331 .991 .778 .997 .757 9.2S4E-3 .222 2.96SE-3 .243 .998 96 Awak* 43 c ft) 20 3.625 1.409 3.647 1.531 .997 .600 .997 .019 3.471E-3 .194 3.323E-3 .101 .999 99 Awaka 43 cft) 21 3.622 1.506 3.700 1.556 .997 .636 .097 .828 3.492E-3 .164 2.989E-3 .472 .999 too Awaka 43 db 22 3,663 1.422 3.017 1.351 .097 .703 .998 .757 3.216E-3 .217 1.937E-3 .243 .999 to t Awak* 43 da 23 3.573 1.389 3.031 1.374 .996 .771 .098 .766 3.05OE-3 .229 1.664E-3 .234 .999 102 Awak* 43 cft) 24 3.753 1.484 3.640 1.640 .997 .796 .998 .860 2.089E-3 .202 2.26QE-3 .1 50 .999 Table 2, Page 2, Left. Database for Estimating the Quality Criterion Hal* ailm block no. channel 1 channel 2 channel 0 channel 4 PI P2 P3 P4 dil P 1 dll P2 dir P3 dif P4 conl 12 103 Awaka 43 cto 25 3.990 1.535 3.912 2.029 .090 .021 .090 .925 1.6S4E-3 .179 1.957E-3 .075 1 .ooo 104 Awaka 43 (to 26 4.228 1.009 4.075 2.275 .999 .842 .999 .952 1 .037E-3 .156 1.411 E-3 .040 1 .ooo 105 Awaka 43 cto 27 4.279 1.544 4.046 2.525 .999 .823 .990 .970 9.361 E-4 .177 1.496E-3 .030 1.000 io e Awaka 43 cto 28 4.58ft 1.566 4.41 4 2.753 .909 .830 .000 .981 5.010E-4 .164 7.131 E-4 .019 1 .ooo 107 Awaka 43 db 29 4.950 1.665 4.525 3.087 1.000 .899 .990 .990 2.382E-4 .101 5.699E-4 .010 1.000 108 Awaka 43 (to 30 S.066 1.961 4.569 3.079 1.000 .91 5 .999 .090 1.01OE-4 .065 S.214E-4 .010 1.000 109 Awaka 43 cto 31 5.40A 1.049 4.769 3.326 1.000 .890 1 .O O O .904 9.563E-5 .104 3.341 E-4 6.256E-3 1.000 1 IO Awaka 43 cto 32 5.669 1.903 4.965 3.508 1.000 .900 1.000 .996 S.606E-5 .094 2.335E-4 3.66BE-3 1.000 111 Awaka 43 db 33 5.926 1.030 5.337 3.420 1.000 .01 1 1 .ooo .995 3.327E-5 .060 1.1QOE-4 5.215E-3 1.000 112 Awaka 43 cto 34 6.178 2.035 5.090 3.747 1.000 .926 1 .ooo .997 1.B95E-5 .074 1.708E-4 2.722E-3 1 .ooo 113 Awaka 43 cto 35 6.272 2.084 5.237 3.011 1.000 .932 1.000 .998 1 .840E-5 .008 1.347E-4 2.395E-3 1.000 114 Awaka 43 cto 30 6.339 2.071 1 5.340 3.759 1.000 .931 1 .ooo .997 1.439E-5 .069 1.O03E-4 2.657E-3 1 .ooo 115 Awaka 43 (to 37 6.353 1.077 5.452 4.266 1.000 .918 1.000 .999 1.3D9E-5 .082 0.710E-5 9.240E-4 1.000 110 Awaka 43 db 38 6.636 1.613 5.753 4.153 1.000 .869 1.000 .999 7.661E-6 .1 1 1 4,7276-6 1.2D6E-3 1 .ooo 117 Awaka 43 C to 39 7.037 1.506 0.210 4.208 1.000 .830 1.000 .099 3.500E-6 .104 1.039E-5 1.O60E-3 1 -00 0 118 Awaka 53 (to 1 .660 .802 1.965 1.410 .340 .451 .916 .782 .554 .549 .084 .218 .641 110 Awaka 53 cto 2 1.174 .716 2.641 t .565 .676 .387 .976 .835 .322 .613 .024 .165 .603 120 Awaka 53 db 3 1.264 .804 2.077 1 .878 .720 .452 .931 .901 .260 .548 .069 .009 .047 121 Awaka 53 (to 4 3.369 1.030 4.012 1.006 .994 .605 .996 .888 5.769E-3 .395 1.601 E-3 .112 .998 122 Awaka 53 db 5 3.062 1 .058 3.814 1.738 .909 .016 .090 .674 .01 1 .384 2.381 E-3 .126 .090 123 Awaka 53 (to 0 2.930 1 .430 3.272 1.685 .986 .786 .993 .861 .014 .212 0.060E-3 .139 .997 124 Awaka 53 da 7 2.950 1.277 3.347 2.372 .007 .720 .094 .960 .013 .274 0.O25E-3 .040 .900 125 Awaka 53 cto 6 3.081 .971 3.538 2.505 .990 .564 .996 .072 .010 .436 4.127E-3 .028 .006 125 Awaka 53 db 9 3.735 .096 3.931 2.517 .907 .579 .996 .970 2.7BBE-3 .421 1.6S4E-3 .030 .000 127 Awaka 53 (to IO 3.964 1.132 4.141 2.906 .990 .656 .999 .086 1.694E-3 .344 1.236E-3 .014 .099 128 Awaka 53 db 1 1 4.112 1.210 4.303 3.556 .990 .099 .999 .996 1.310E-3 .301 7.591 E-4 3.062E-3 1.000 120 Awaka 53 cto 12 4.541 1.154 5.008 3.546 .090 .660 1 .O O O .906 5.516E-4 .332 1.623E-4 4.062E-3 1.000 130 Awaka 53 (to 1 3 4.401 1.179 5.365 3.021 .900 .680 1 .O O O .997 7.320E-4 .320 1.039E-4 3.49BE-3 1.000 131 Awaka 53 db 1 4 4.392 1.125 5.191 3.707 .909 .653 1 .000 .997 7.454E-4 .347 1.470E-4 2.04BE-3 1 .oao 132 Awaka S3 db 1 5 4.764 1.114 6.51 1 3.627 1.000 .647 1 .000 .907 3.375E-4 .353 7.727E-S 3.457E-3 1 .ooo 133 Awaka 53 cto 10 6.322 1.191 6.703 3.337 1.000 .600 1.000 .994 1.1 34E-4 .314 6.734E-0 0.14SE-3 1 .ooo 134 Awaka S3 cto 17 6.096 1.120 7.533 3.500 1.000 .650 1 .O O O .096 2.356E-S .350 1.206E'ft 3.6506-3 1 .ooo 135 Awaka S3 (to 18 6.145 1.206 7.500 3.912 1.000 .093 1 .000 .008 2.133E-S .307 1.340E-0 1,9576-3 1 .ooo 130 Awaka S3 cto 19 6.292 1.348 7.029 3.964 1.000 .755 1.000 .096 1.583E-S .245 1.D59E-6 1.7B3E-3 1.000 137 Awaka 53 cto 20 6.170 1.443 7.517 3.973 1.000 .791 1.000 .998 2.028E-S .209 1.328E-0 1.732E-3 1 .ooo 30 Awaka 53 cto 21 6.000 1.537 7.336 3.054 1.000 .821 1.000 .906 2.B63E-S .179 1.913E-0 2.198E-3 1 .ooo 39 Awaka A3 db 22 6.435 1.488 7.983 4.060 1.000 .006 1.000 .909 1.104E-8 .194 5.197E-7 1.4546-3 1 .ooo 1 40 Awaka S3 cto 23 7.100 1.013 0.641 5.036 1.000 .889 1 .ooo 1 .O O O 2.521 E-6 .111 1.39DE-7 2 0766-4 1 .ooo 141 Awaka 53 db 24 7.507 1.578 9.068 4.069 1.000 .833 1 .ooo 1 .ooo 1.35SE-0 .167 S.033E-0 2.320E-4 1.000 142 Awaka 53 cto 25 7.540 1.535 9.405 4.710 1.000 .021 1 .000 1.000 1.26BE-6 .170 2.696E-6 3.6566-4 1.000 143 Awaka 53 cto 26 7.914 1.547 0.942 5.013 1 .ooo .024 1 .ooo 1 .ooo 5.970E-7 .170 1.050E-8 2.122E-4 1 .ooo 144 Awaka 53 cto 27 7.749 1.687 9.387 5.124 1 .ooo .057 1 .ooo 1.000 Q.320E-7 .143 3.147E-B 1.60SE-4 1 .ooo 145 Awaka 53 cto 2ft 7.756 1.565 9.513 5.035 1 .ooo .030 1 .ooo 1.000 B.204E-7 .170 2.452E-8 2.030E-4 1 .000 14ft Awaka S3 ito 29 7.640 1.907 9.863 6.171 1.000 .006 1 .ooo 1.000 6.920E-7 .004 1.227E-8 2.0246-5 1 .ooo 147 Awaka 53 db 30 7.404 1.905 9.729 6.909 1,000 .919 1 .ooo 1.000 1.066E-6 .OBI 1.509E-8 4.534E-6 1 . o o o 148 Awaka 53 cto 31 7.930 2.395 10.400 0.997 1.000 .962 1 .000 1.000 5.669E-7 .038 4.261E-9 3.795E-6 1.000 140 Awaka 53 cto 32 ft.lftft 2.562 10.335 7.084 1.000 .972 1.000 1.000 3.444E-7 .028 4.841E-9 9.4846-7 1.000 ISO Awaka 53 db 33 6.498 2.548 11.090 7.990 T .O O O .972 1 .ooo 1.008 1.051E-7 .028 1.1O5E-0 5.12SE-7 1 .ooo 151 Awaka S3 (to 34 0.686 2.042 12.515 7.900 1 .ooo .976 1 .ooo 1.000 1.741E-6 ,084 7.0376-11 0.0436-7 1 .ooo 152 Awaka 53 db 35 9.972 2.91 9 13.055 0.072 1 . o o o .986 1 .000 1.000 9.093E-9 .014 2.S09E-11 4.340 E-7 1 .ooo 153 Awaka 53 cto 3ft 10.267 7.020 oao 5.32OE-0 .020 1.252E-11 9.0996-7 1.000 Table 2, Page 3, Left. Database for Estimating the Quality Criterion •111* ailm block no. channel 1 channel 2 channel a channel 4 PI P2 P3 P4 6 1 1 PI on P2 dll P3 dll P 4 eonl f 2 154 Awak* S3 c ft> 37 10.661 2.965 14.153 8.107 1.000 ,9B7 1.000 1.000 1.727E-9 .013 2.9725-12 4.051 E-7 1.000 155 Awak* 53 cfti 38 10,968 3.195 14.291 8.299 1.000 .992 1.000 1.000 1.402E-9 8.129E-3 2.427E-12 2.7576*7 1.000 150 Awaka 53 < * > 39 12,493 2.636 15.725 0.771 1.000 .976 1.000 1.000 7.34E* 11 .024 1.689E-13 1,072£>7 1.000 157 A*I*#P no a ton 1 1.220 1.010 .491 1.207 ,700 .586 .217 .694 .300 .412 .783 .306 .876 150 Aalaap no aton 2 1,234 1.402 .863 1.121 .700 .776 .493 .650 .294 .224 .507 .350 .934 159 AsIm o no ailm a .961 1.152 .748 .989 .570 .687 .412 .575 .430 .333 .568 .425 .657 160 Aalaap no aH m 4 .920 .745 .754 .996 .532 .410 .438 .579 .466 .590 .562 .421 ,723 161 Aalaap no *km 5 1.156 .668 1.051 .070 .669 .367 .612 .504 .331 .633 .360 .496 ,790 162 Aalaao no akm 6 1.124 .658 .878 .960 .652 .344 .504 .557 .348 .656 .496 .443 .772 163 Aataao no sUm 7 .976 .606 .842 .920 .567 .454 .479 .532 .433 .540 .521 .460 .764 164 Aslavp no slim 8 1.040 .979 .774 .709 .60S .569 .431 .383 .395 .431 .569 .617 .630 165 Aalaao no asm 9 .915 .931 .662 .726 .528 .539 .506 .397 .472 .461 .494 .603 .782 166 Ast*ao no a iE m IO .924 1.046 .745 .795 .534 .609 .410 .440 .466 .391 .590 .554 .616 167 Aalaap no wlm 1 1 .630 1.218 .647 .755 .471 .699 .336 .417 .520 .301 .604 .583 .041 166 Aalaao no sU m 1 2 .689 .059 .729 .672 .366 .656 .398 .355 .632 .444 .602 .645 .720 169 Astaao no ailm 13 .666 .895 .750 .710 .496 .515 .413 .304 .504 .465 .567 .616 .755 170 Afttaao no atlm 14 .676 1.009 .767 .768 .350 .587 .418 .426 .642 .413 .582 .574 .735 171 Aslaao no alim 15 .724 .975 .680 .755 .394 .566 .491 .417 .606 .434 .509 .593 .737 172 Aalaao no lUm 16 ,806 .960 .929 .742 ,454 .557 .537 .407 .546 .443 .463 .593 .758 173 Ai Im d no atlm 17 .654 .944 .696 .779 .487 .647 .518 .434 .513 .453 .484 .566 .768 174 AsIm d no atlm 16 .799 .061 1.003 .949 .449 .570 .584 .550 .551 .430 .416 .450 .763 176 Aataao no atlm 10 .693 .944 .960 .879 .514 .547 .557 .504 .466 .453 .443 .496 .760 176 Aalaao no ailm 20 .665 .866 .696 .955 .506 .496 .516 .554 .492 .604 .484 ,446 .752 177 AsIh o no stim 21 .909 .841 .927 .059 .575 .476 .536 .511 .4 25 .522 .464 .469 .778 176 Aalaao no ailm 22 .930 .768 .912 .664 .536 .426 .526 .500 .462 .574 .474 .492 .735 179 A m i mo no aton 23 .661 .692 .993 .797 .506 .370 .576 .447 .494 .630 .422 .553 .609 160 Aalaao no aton 24 .624 .850 .905 .750 .466 .485 .522 .413 .534 .515 .470 .587 .725 181 Aalaap no alim 25 .826 1.107 .614 .634 ,468 .643 .459 .473 .532 .357 .541 .527 .010 162 Ailaao no aton 26 .760 .915 .997 1 .132 .421 .526 .560 .650 .579 .472 .420 .344 .727 183 Aalaao no attm 27 .757 .751 1.243 1.518 .416 .414 .711 .815 .582 .586 .269 .165 .659 184 A^W aP no aton 28 .934 .696 1.241 1.155 ,541 .516 .710 .668 .459 .464 .290 .332 .778 165 Aalaao no ailm 29 .665 .810 1.099 1.312 .495 .457 .639 .741 .505 .543 .361 .259 ,728 166 Aataao no ailm 30 .862 .603 .906 1.360 .493 .452 .524 .760 .507 .546 .476 .240 .722 187 A«m p no ailm 31 .663 .771 .679 1.161 .507 .429 .504 ,071 .493 .571 .496 .329 .718 188 Aalaap no atlm 32 .904 .764 .650 1.107 .521 .423 .491 .643 .479 .577 .609 .357 .724 169 Aalaao no atlm 33 .908 .670 .685 1.075 .524 .496 .365 .825 .476 .502 .035 .375 .701 190 Aalaao no atim 34 .917 .766 .910 1.033 .630 .425 .525 .601 .470 .675 .470 .399 .729 191 Aalaao no alim 35 .968 .760 .018 1.052 .562 .435 .462 .612 .436 .665 .536 .386 .753 192 Aalaao no alim 36 .909 .760 .621 1.158 .524 .421 .464 .670 .476 .679 .530 .330 .724 1 93 Aalaao no slim 37 .917 .785 .627 1.227 .530 .439 .469 .703 .470 .561 .531 .297 .730 1 94 Aalaao no alim 38 .912 .662 .640 1.111 .526 .493 .478 .645 .474 .507 .522 .356 .700 195 Aalaao no atim 39 .683 .949 .610 1.068 .507 .550 .457 .621 .493 .450 .543 .379 .770 196 Aalaao 38 < ± > 1 .402 .635 .714 .902 ,153 .327 .387 .520 .847 .673 ,013 .480 .430 197 Aalaao 38 < f> 2 .569 .632 .570 1.991 .276 .325 .262 .920 .724 .675 .718 .080 .511 198 Aalaao 38 (t) 3 .760 .629 .630 2,534 .421 .470 .323 .971 .579 .530 .677 .029 .693 199 Aataao 38 < * > 4 .777 .490 .767 1.609 .433 .216 .420 .669 .507 .704 .574 .111 .356 200 Aataao 38 < * > 5 1.152 .902 .690 1.4 B6 .667 .520 .512 .605 .333 .460 .488 .195 .840 201 Aalaao 36 db 6 1.082 .625 .956 1.620 .029 .487 .555 .691 .371 .533 .445 .109 .602 202 Aalaao 36 d> 7 .650 .556 .777 1 .520 .485 .266 .433 .618 .515 .732 .587 .182 .623 203 Aataao 36 < t> 8 .797 .616 .790 1.666 .447 .237 .442 .627 .553 .763 .556 .173 .379 _ £ 0 4 Aalaap 36 d> 9 .644 .469 .737 1.378 .334 .201 .404 .767 .660 ,799 .596 .233 .4 07 Table 2, Page 4, Left. Database for Estimating the Quality Criterion ■1*1* sflm block no. CllMlWl 1 channel 2 chsnnat S channel 4 PI P2 P3 P4 dif PI dtf P2 dif P3 dll P4 cool 12 205 As Im d 3ft A IO .043 .480 .754 1.390 .333 .194 .410 .774 .667 .800 .504 .226 ,462 20ft ASiMH 3ft A 11 .650 .520 .073 1.550 .330 .239 .356 .025 .682 .761 .644 .175 .496 207 Arf**o 3ft db 12 .625 .635 .722 1.355 .319 .327 .393 .758 .001 .073 .607 .242 .542 20ft .. .AsIm p 3ft A 13 .703 .647 .770 1.538 .37ft .330 .428 .B21 .622 .864 .572 .179 .567 200 M te o 3ft cfti 14 .704 .544 .050 1.093 .379 .257 .550 .603 .621 .743 .444 .137 .539 210 A H m 3ft A 15 .724 .874 .992 1.054 .394 .356 ,577 .054 .60ft .644 ,423 ,14ft .010 2*1 AsIm d 3ft A 1ft .7B5 .002 1.002 1.817 .439 .493 .563 .090 .581 .507 .417 .1 IO .715 212 As J sod 3ft A 17 .072 .991 1.051 2.001 .500 ,57ft .-91* ,921 .900 .424 .308 .079 .760 213 As Im d 3ft A 18 .765 1.032 .990 1.992 .424 .601 .576 .920 .570 .399 ,424 .080 ,770 214 A«Im p 3ft A 1ft .74ft 1.200 1.187 1.054 .410 .694 .674 .654 .590 .30ft .320 ... .140 .620 215 A*)«p 3ft A 20 .Baa t.139 1.163 1.751 .510 .060 .082 .077 .490 .340 .318 .123 ,833 21ft AHwo 38 A 21 .900 1.278 1.141 1.000 .516 .726 .001 .000 .482 .274 .339 ,140 .860 217 As Im o 3ft A 22 .917 1.004 1.184 1.531 .530 .584 .063 .61 9 .470 .4(0 .317 .181 .604 21ft Atieeo 3ft A 23 1.029 1.104 1.227 1.590 .599 .641 .703 .637 .401 ,359 .297 .103 .650 210 Ai Im d 3ft A 24 1.030 1.255 1.227 1.325 .603 ,71ft .703 .746 .397 .204 .297 .204 .087 220 AltMP 3ft A 25 1.145 1.226 1.22ft 1.293 .663 .704 .703 .733 .337 .296 .297 .287 .900 221 AtiUP 3ft A 26 1.237 1.201 1.270 1.200 .700 .691 .723 .730 .292 .309 .277 ,270 .910 222 Aitoap 3ft A 27 1.48ft 1.252 1.226 1.241 .806 .715 .704 .710 .194 .265 .290 .290 .945 223 Ai Im d 30 A 2ft 1.304 .911 1.029 1.055 .737 .520 .599 .614 .203 .474 .401 .366 .075 224 AttMD 3ft A 29 1.324 .652 .964 .094 .746 .466 .560 .514 .254 .514 .440 .480 .069 225 AtiMD 3ft A 30 1.507 1.036 1.012 .799 .612 .603 .569 .449 .160 .397 .411 .551 .925 220 Aalppp 3ft A 31 1.010 .902 1.103 .877 .842 .520 .641 .503 .tS6 .480 .359 .497 .924 227 A*Ih d 3ft A 32 1.960 .667 1.301 .950 .915 .351 .730 .066 .085 .649 .264 .445 .945 22ft Atl««D 38 A 33 2.133 .706 1.425 . BBS .936 .301 .784 .575 .002 .019 .216 .425 .962 22ft ASlMD 3ft A 34 2.264 .684 1.521 .990 .951 .200 .816 .581 .049 .712 .104 ,419 .905 230 AsIm o 3ft A 35 2.398 .554 1.407 1.011 .902 .205 .606 .588 .030 .735 .104 .412 .972 231 AsIm d 3ft A 36 2.315 .634 1.503 .005 .956 .326 .635 .500 .044 .074 .105 .492 .970 232 A tlm 3ft A 37 2.464 .571 1.027 .934 .27ft .892 .541 .032 .722 .108 .459 " 977 23? W m p 3 1 1 A 38 2.476 .804 1.906 1.059 .967 _.227 .900 .616 .033 .773 .094 .364 .976 234 As Im p 3ft A 3ft 2.492 .926 t .976 1.106 .906 .245 .910 .642 .032 .755 .082 .350 .976 235 AtlND 43 A 1 1.400 1.204 1.247 . .1,474 .775 .729 .712 .601 .225 .271 .280 .199 .939 236 As Im d 43 A 2 1.840 .960 1.810 1.493 .695 .562 .869 .000 .105 .430 .111 .192 .054 237 Ai Im d 43 A 3 2.262 1.11ft 2.266 1.442 ,951 .640 .953 .790 .049 .352 .047 .210 .963 238 AsIm d 43 A 4 1.749 1.261 2.199 1.736 .07ft .719 .945 .074 .124 .281 .055 .12ft .905 23ft AllMD 43 A 5 1.669 1.202 2.175 1.599 .626 .691 .943 .639 .172 .309 .057 .161 .947 240 AsIm d 43 A 6 1.051 1.380 2.402 2.194 .097 .708 .902 ,945 .103 .232 .030 .059 ,97ft 241 AsIm d 43 A 7 1.906 1.413 2.314 2.417 .907 .760 .950 .963 .093 .220 .044 .037 .979 242 AsIm d 43 A a 1.723 1.733 1.970 2.120 .670 .073 .917 ,937 .130 .127 .083 .0A3 .903 243 AsIssD 43 A 9 2.300 2.29ft 2.340 2.415 ,955 .954 .958 ,963 .045 .040 .042 .037 .998 244 Ai Im p 43 A IO 2.371 2.44*1 2.190 2.47ft .900 .965 .944 ,907 .040 .035 .050 .033 .999 245 A tlu p 43 A 11 2.787 2.472 2.54ft 2.574 .962 .967 .971 .973 .018 .033 .029 .027 .999 248 AsIm p 43 A 12 2.682 2.520 2.631 2.401 .977 .970 .976 ,902 .023 .030 .024 .036 .999 247 AsS m o 43 A 1 3 2.830 2.329 2.671 2.45ft .903 .907 .977 .966 .017 .043 .023 .034 .999 24ft AsIm d 43 A 14 2.000 2.343 2.80ft 2.605 .982 .958 .983 .974 .018 .042 .017 .020 .999 24ft AsIm d 43 A 15 2.92ft 2.37ft 2.884 2.61ft .966 .961 .905 .975 .014 .039 .015 .025 .999 250 AsIm p 43 A 1ft 2.027 2.254 3.030 2.575 ,98ft .959 .989 .973 .014 .041 .011 .027 .999 251 AsIm d 43 A 17 2.59ft 2.30ft 2.929 2.656 .974 .960 .98ft .977 .020 .040 .014 .023 .999 252 AsIm d 43 A 1ft 2.474 2.142 3.159 2.306 .967 .939 .991 .955 .033 .001 6.725E-3 .045 .990 253 AslMp 43 A 1fi 2.542 2.169 3.224 2.097 .971 .942 .692 .934 .029 .058 7.079E-3 .006 .990 254 AsIm d 43 A 20 2.6S0 2.169 3.305 2.104 .977 .942 .994 .935 .023 .058 S.ftlSE-3 .005 .999 255 As Im d 43 A 21 2.772 2.163 3.532 2.168 .961 .943 .990 .942 .019 .057 4.176E-3 .056 .999 Table 2, Page 5, Left. Database for Estimating the Quality Criterion S lltt •tiro block no. channel 1 channel 2 channel 3channel 4 PI P2 P3 P4 dll pi dll P2 dll P3 dif P 4 ZS6 Asltto 43 c t o 22 2.613 1.774 3.417 1.942 .983 .681 .995 .912 .017 .119 5.246E-3 .066 .998 257 Asltto 43 c t o 23 2.666 1.667 3.706 2.055 .985 .857 .997 .926 .015 .143 2.9S4E-3 .072 .096 25# Asltto 43 db 24 3.275 1.677 4.159 2.IBS .993 .659 .999 .944 6.945E-3 .141 1.1D2E-3 .056 .099 259 Asltto 43 c t o 25 3.243 1.657 4.309 2.501 .993 .695 .999 .969 7.3976-3 .145 B.B12E-4 .031 .999 200 Asltto 43 db 26 3.165 1.716 4.061 2.496 .992 .669 .999 .960 6.2916*3 .131 1.304E-3 .031 ,999 201 * * •• £ 43 c ft> 27 3.304 1.690 4.398 2.427 .993 .663 .999 .964 6.559E*3 .137 7.3646-4 .036 .999 262 Asltto 43 c t o 26 3.425 1.761 4.799 2.769 .995 .679 1.000 .962 S.164E-3 .121 3.274 E-4 ,016 263 Asltto 43 db 29 3.623 1.636 5.279 2.952 .997 .694 1.000 ,987 3.465E-3 ,106 1.237 E-4 .013 1.000 264 Asltto 43 db 30 3.611 1.776 5.341 3.025 .996 .862 1.000 .909 3.S69E-3 .ite 1.001 E-4 .011 1,000 265 Asltto 43 c t o 31 3.597 1.649 5.223 2.965 .996 .652 1.000 .987 3.670E-3 .148 1.386 E-4 .013 .999 266 Asttto 43 db 32 3.699 1.736 5.374 3.217 .996 .874 t.000 .992 2.006E-3 .126 1.020E-4 7.785E-3 1.000 267 Asltto 43 db 33 3.972 1.660 5.315 3.265 .996 .902 1.000 .993 1.735E-3 .096 1.150E-4 6.6D9E-3 1.000 266 Asltto 43 c t o 34 4.532 1.972 5.907 3.580 .999 .917 1.000 .096 3.619E-4 .063 3.456E-5 3.7966-3 1.000 26S Asltto 43 db 35 5.129 2.266 6.170 3.775 1.000 .053 1.000 .907 1,6786*4 .047 2.026E-5 2.5746-3 270 AtlttO 43 db 36 5.400 2.343 6.565 3.799 1.000 .956 1.000 .998 9.505E-5 .042 8.739E-6 2.453E-3 1.000 271 Asltto 43 c t o 37 5.636 2.311 7.507 3.643 1.000 .955 1.000 .907 3.904E-S .045 1.353E-6 3.349E-3 272 Aalaap 43 db 36 6.351 2.266 0.042 3.643 1.000 .952 t . o a o .907 1.404E-5 .048 4616E-7 3.3496-3 1.000 273 Asltto 43 c ft) 3ft 6.939 2.460 0,784 3.954 1.000 .968 1.000 .908 4.267E-6 .032 1.04SE-7 1.799E-3 274 Asltto S3 db 1 2.972 1.024 2.016 1.695 .987 .596 ,923 .664 .013 .404 .077 .136 27S Asltto S3 c t o 2 5.069 1.864 4.231 2.412 1.000 .902 .999 .963 1.8956-4 .098 1.031E-S .037 1.000 276 Asltto 53 db 3 6.355 1.537 5.9i a 2,774 1.000 .821 1.000 .082 T.393E-5 .179 3.3B2E-8 ,018 1.000 277 Aaittfi 53 C ft) 4 6.627 1.559 6.967 2.498 1.000 .628 1.000 .969 5.353E-6 .172 4.0326-6 .031 276 Asltto 53 c t o 5 7.497 2.030 7.927 2.765 1.000 .925 1.000 .992 1.382E-6 .075 6.B18E-7 .018 279 Asltto 53 db 6 6.514 1.753 9.316 2.919 1.000 " !B77 1.000 .966 2.447E-8 .123 3.624E-8 .014 1.000 260 Asltto 53 < ft> 7 10.463 2.056 9.570 2.990 1.000 .929 1.000 .968 3.7656-0 .071 2.190E-8 .012 1.000 241 . Aalaap 53 db 6 11.074 2.194 10.727 3.843 1.000 .945 1.000 .906 1.140E-9 .055 2.24SE-9 2.247E-3 1.000 242 Asltto 53 d) 9 12.183 2.926 12.104 4.725 l . o o o .966 f .000 1.000 1.33E-ID .014 t.SSOE-10 3.803E-4 1.000 263 Asltto S3 db 10 12.673 2.814 13.062 4.940 1.000 .963 1.000 1.000 5 2E-I1 .017 2.3636-11 2.461 E-4 1.000 264 AtlttD S3 c t o 1 1 14.019 2.480 14.193 5.319 1.000 .988 1.000 1.000 4.0SE-12 .032 2.917E-12 1.14IE-4 1.000 266 Asltto 63 c ft> 12 14.519 2.361 15.667 5.790 1.000 .939 1.000 1.000 1.50E-I2 .041 1.079E-13 4.365E-5 266 Asltto 63 (ft) 13 17.131 2.523 17.996 8.329 1.000 .970 1.000 1.000 1.3E-14 .030 2.723E-1S 1.4696*5 267 Asltto 53 db 14 16.146 2.811 19.341 8.716 1.000 .963 1.000 1.000 3.S4E-16 ,017 2.S14E-I6 6.67SE-6 1.000 266 S3 c t o 1 5 20.941 3.121 21.444 7.622 1.000 .991 1.000 1.000 1.57E-17 9.400E-3 6.7766-16 7.164E-7 1.000 269 Asltto S3 e ft) 1 6 21.491 3.062 22,649 7.316 1.000 .990 1.000 1.000 6,186*10 .010 7.047E-19 1.902E-6 1.000 290 Aalaap 53 c t o 17 21.690 2.917 24.552 7.204 1.000 .986 1.000 1.000 2.93E-1B .014 1.084E-1 0 2.4976-6 291 AtlttD S 3 db 1 6 22.676 2.683 25.764 7.603 1.000 .985 1.000 1,000 5.42E-19 .015 0.000 7.464E-7 292 Asltto 53 c t o 19 22.997 3.595 26.664 8.242 1.000 .996 1.000 1.000 4.346-19 3.68SE-3 0.000 3.090E-7 293 294 Asltto S 3 c t o 20 23.7S0 3.232 28.169 8.725 1.000 .992 1.000 1.000 3.2SE-19 7.558E-3 2.7T1E-19 1.1766-7 1 — 53 db 21 24.230 3.061 26.596 9.226 1.000 .990 1.000 1.000 2.17E-10 .010 1.62BE-1 9 4.316E-6 295 AtlttD S3 c t o 22 25.622 3.036 30.702 9.944 1.000 .989 1.000 i . o a o 0.000 .011 0.000 1.04SE-6 1.000 296 —^ ‘ J L 53 c t o 23 26.991 3.246 31.668 10.446 1.000 .993 1.000 1.000 1.08E-10 7.324E-3 3.795E-19 3.8036*9 297 Ailtto 53 c ft) 24 28.137 3.677 33.313 10.901 1.000 .997 1.000 1.000 0.000 3.130E-3 1.064E-19 1.S96E-9 296 Aalaap 53 c t o 25 29.715 3.936 35.136 12.045 1,000 .996 1.000 1.000 0.000 1.86SE-3 1.O84E-10 1.74E-10 299 Asltto 53 ( f t ) 26 30.176 3.979 36.223 12.063 1.000 .996 1.000 1.000 5.42E-20 1.7116-3 0.000 1.66E*10 300 Asltto S3 c t o 27 31.294 3.912 37.664 12.306 1.000 .998 1.000 1.000 0.000 1.9S7E-3 0.000 I.05E-10 301 Asltto 63 c t o 2Q 32.465 4.075 38.647 12.603 1.000 .999 1.000 1.000 S.42E-20 1.411 E-3 I.626E-19 5.9SE-11 302 Asltto 53 c t o 29 33.711 4.224 39.965 13.265 1,000 ,999 1.000 T.000 2.17E-t9 1.046E-3 0.000 1.626*11 303 Asltto S3 cto 30 34.907 4.383 40.923 14.390 1.000 .999 1.000 1.000 0.000 7.S91E-4 0.000 2.02E-12 304 A»l*n 53 db 31 35.764 4.536 41.161 14.966 1.000 .999 1.000 1.000 5.42E-20 5.5746*4 0.000 B.61E-13 305 AslttD 53 c t o 32 36.609 4.543 42.614 15.713 1.000 .999 1.000 1.000 3.25E-19 5.4856*4 1.626E-19 1.73E-13 306 Asltto 53 c t o 33 37.603 4.756 44,037 16.214 1.0Q0 t.000 1.000 1.000 3.5726-4 1.626E-19 8.666-14 1 . o o o Table 2, Page 6, Left. Database for Estimating the Quality Criterion ■111* •Um block no. channel 1 channel 2 channel 3 channel 4 PI P2 P3 P4 dif Pi dif P2 d! P3 d! P4 conl 12 307 Ai Im d S3 * 34 36.423 4.835 44.925 16.801 1.000 1.000 1.000 1.000 0.000 3.044E-4 0.000 2.36E-14 1.000 306 AsIm d 53 db 35 40.411 4.990 46.923 17.495 1.000 1.000 1.000 1.000 0.000 2.224E-4 0.000 6.72E-15 1.000 309 AsIm d 53 d> 36 43.107 5.462 49.453 16.790 1.000 1.000 1.000 1.000 o.ooo 8.S35E-S 5,421 E-20 6.65E-16 1.000 310 AsIm d 5 3 * 37 45.007 5.596 51.070 19.559 1.000 1.000 1.000 1.000 0.000 6,5025-5 2.711E-19 1.716-16 1.000 311 AsIm d 53 * 36 47.543 5.6B6 54.023 20.404 1.000 1.000 1.000 1.000 0.000 3.6C9E-5 0.000 3.89E-17 1.000 312 AsIm d 53 * 39 49.101 5.967 55.167 20.789 1.000 1.000 1.000 1.000 0.000 3.061 E-S 0.000 2.01 E-17 1.000 Table 2, Page 7, Left. Database for Estimating the Quality Criterion cant 13 con! 1 4 .074 cant 24 con I 34 c r it ia O fll23 erl!34 .702 ,981 .084 .080 .002 .342 .005 .700 .710 .805 .737 .047 .70S .702 .804 .085 .080 .081 .003 _.OS4 .840 ■ 081 .898 .030 .030 ■ Q O S .888 O O 1 1 1 2 .078 J57 .070 ♦77S .504 .844 14 .705 .772 .599 18 17 _ _ .S 2 4 .821 .740 .888 .720 .400 .331 O O .640 .620 .687 ■ S O I - 12. .640 .601 o o 20 21 .570 JB26 .541 .812 414 .605 ■ 564 O O 2 2 ^644 .64 1 .521 7 9 3 .607 _ i± 2 5 .730 .771 .723 .665 .419 .772 O O 26 27 .746 .686 ,686 .726 -744 ■ 721 .617 .635 O O - 22. 31 ,753 .713 32 3 3 .626 ,646 .725 .561 .573 O O .617 .864 .583 JUS .547 .61 1 ,646 O O .640 38 30 _^668 7fl3l .717 .667 ^581 .642 .630 .621 40 41 .624 .045 ,661 O O .260 ■ 776 .083 .034 .760 ,756 _ q o 44 45 ■ 387 .402 .767 .760 46 47 .606 .770 .775 .643 .060 .607 ,818 .91 1 .650 .835 O O .766 .813 ,006 ■ 752 .024 .826 ,780 ,710 .767 Table 2, Page 1, Right. Database for Estimating the Quality Criterion O s C T \ conl 13 conl 14 conf 23 conl 24 Conl 34 crltl2 crlll 3 crlll 4 crlt23 crl!24 cri134 52 .701 .835 .047 .91 0 .753 O o d d d o S3 .700 .661 .005 .007 .871 O o o o o o 54 .770 .801 .005 .055 .007 O o o o o o SS .700 .070 .039 .904 .093 o o d o o o 55 .007 .041 .0 73 .088 .953 o o o o o o 57 .020 .044 .959 .090 .057 o o a o 1 o sa .700 .024 .053 .085 .93 1 o o o o d o so .072 .035 .080 .003 .050 o o o o 1 o so .701 .060 .007 .982 .808 o o o o o o A 1 .550 ,B B O .070 .082 .91 1 a o o o o o 52 .700 .810 .088 .009 .87 1 o o o o o o A3 ,701 ,860 .072 .083 .900 o o o o o o A 4 .004 .002 .084 .089 .893 o o o o o o 05 .052 .009 .080 .090 .9*0 o o o o 1 o A A .556 .085 .090 .091 .872 1 o o o 1 o 07 .805 .003 .003 .994 .008 1 o d 1 1 o A 6 .872 .875 .085 .080 .845 o o o o o o B O .022 .689 ,072 .083 .012 o d o o o o 70 .025 .864 .072 .070 .701 o o o o o o 71 .022 ,664 .072 .070 .890 o o o o d o 72 .020 .085 .070 .084 .704 9 o o o o o 73 .040 .087 .030 .052 .730 o o o o o o 74 .605 .080 .91 1 .940 .71 S o o o o o o 75 .030 .021 .864 .935 .748 o o o o o o 7 0 .013 .055 .801 .044 .782 o o o d o o 77 .027 .050 .007 .942 .800 o o d d o o 70 .950 .072 .004 .934 .051 o o o o o o 70 .003 .004 .540 .603 .002 o o o o o o eo .070 .007 .525 .553 .028 o o o o o o B 1 .030 .537 .529 .307 .545 o o o o o o 02 .701 .574 .039 .200 .708 o o o o o o 03 .6 1 2 .503 .043 .358 .690 o o o o o o B4 .007 .015 .713 .530 .70 4 o d o o o o 05 ,051 .O O O .820 .51 5 .727 d o a o o o 55 .730 .730 .700 .096 .708 o o o o o o 07 .070 .802 .640 .033 .640 d o o o o o o a .750 .005 .740 .682 .757 o o o o o o B fl .710 .770 .700 .774 .720 o o o o o d ao .705 .703 .70S .703 .031 o o o o o o 01 .61 a .090 .708 .749 .757 o o o o o o 02 .650 .901 .040 .882 .850 o o o o o o 03 .080 .034 .070 .005 .950 o o o d o o 04 .008 .969 .005 .034 .884 1 1 o 1 o o 05 1 .ooo .007 .900 .938 .904 1 1 o 1 o 1 o o 1 .O O O .007 .900 .055 .900 1 1 1 1 o 1 07 1 .O O O .O O O .900 .040 .090 1 1 1 1 o 1 oa 1.000 .O O O .009 .065 .090 1 1 1 1 o 1 O O 1 .O O O .090 1 .O O O .072 .999 1 1 1 1 o 1 O O 1 .ooo .O O O 1 .O O O .047 1 .O O O 1 1 1 1 o 1 101 1.000 .O O O 1 .ooo .946 1.800 1 1 1 1 d 1 1 02 1 .ooo 0 0 0 1 .O O O .970 1 .O O O 1 1 1 1 o 1 Table 2, Page 2, Right. Database for Estimating the Quality Criterion oonf 13 oonf 14 oonf 23 conl 24 conl 34 erll 1 2 orltl 3 orll 1 4 erl123 orlt?4 erll34 103 1 .ooo 1 .ooo 1 .ooo ,9S7 1 .900 1 1 1 1 O 1 104 1 .ooo 1 .ooo 1 .ooo .992 1 .ooo 1 1 1 1 1 1 109 1 .ooo 1 .ooo 1 .ooo .995 4 b 0 0 1 1 1 1 1 1 1 O S 1 .ooo 1 .ooo 1 .ooo .997 1 .ooo 1 1 1 1 1 1 1 07 1 .ooo 1 .ooo 1 .ooo .999 1 .ooo 1 1 1 1 1 1 O B 1 .ooo 1 .ooo 1.000 .999 1 .ooo 1 1 1 1 1 1 1 O O 1 .ooo 1 .ooo 1 .ooo .999 1 .ooo 1 1 1 1 1 1 1 1 O 0 0 0 1 .ooo 1 .ooo 1 .ooo a 0 0 1 1 1 1 1 1 111 1 .ooo 1 .ooo 1 .ooo 1 .O O O 1 .ooo 1 1 1 1 1 1 112 1 .ooo 1 ,Q O Q 1 .ooo 1 .O O O 1 .000 1 1 1 1 1 1 113 1 .ooo 1 .ooo 1 .ooo 1 .O O O 1 .ooo 1 1 1 1 1 1 114 1 .ooo 1 .ooo 1 .ooo 1 .O O O 1 .ooo 1 1 1 1 1 1 I I S 1 .ooo 1 .ooo 1 .ooo 1 .O O O 1 .ooo 1 1 1 1 1 1 l i f t 1 .ooo 1 .ooo 1 .ooo 1 .O O O 1 .ooo 1 1 1 1 1 1 117 1 .ooo 1 .ooo 1 .ooo 1 .O O O 1 .ooo 1 1 1 1 1 1 1 1 ft .9 4 5 SS7 .954 . B ftO .902 o o O 110 .002 .947 ,o as .009 .990 o 1 o o O 1 120 .OBI .972 .002 .94ft .993 o o o o 121 1 .ooo .999 .099 .95ft 1 .ooo 1 1 1 1 o 1 1 22 1 .ooo .999 .999 .951 1 .ooo 1 1 1 1 o 1 1 23 1 .ooo .99ft .999 .97 1 .999 1 1 1 1 o 1 1 24 1 .ooo .999 .99ft .909 1 .O O O 1 1 1 1 o 1 1 2S 1 .ooo 1 .O O O .996 .998 1 .O O O 1 1 1 o 1 1 2ft 1 .ooo 1 .O O O .909 .907 1 .O O O t 1 1 1 o 1 1 27 1 .ooo 1 .O O O 1 .O O O 695 1 .O O O 1 1 1 1 1 1 2 8 0 0 c 1 .ooo i .ooo .999 1 .O O O 1 1 1 1 1 1 1 20 0 0 0 1 .O O O 1 .ooo .999 1 .ooo 1 1 1 1 1 1 1 30 1 .ooo 1 .O O O 1 .ooo .999 1 .ooo 1 1 1 1 1 131 1 .ooo 1 .O O O 1 .ooo .999 1 .000 1 1 1 1 1 1 32 4 b 0 0 1 .ooo 1 .ooo .909 1 .ooo 1 1 1 1 1 f 1 33 1 .ooo 1 .ooo 1 .ooo .99ft 1 .O O O 1 1 1 1 1 1 1 34 0 0 0 1.000 1 .ooo .999 1 .ooo 1 1 1 1 1 1 1 35 1 .ooo 1 .ooo 1 .ooo .999 1 .ooo 1 1 1 1 1 1 1 3ft 1 .ooo 1 .ooo 1 .ooo 1 .O O O 1 .ooo 1 1 1 1 1 1 1 37 1.000 1 .ooo 1 .ooo 1 .ooo 1 .ooo 1 1 1 1 1 1 3 B 1 .ooo 1 .ooo 1 .ooo 1 .ooo 1 .ooo 1 1 1 1 1 1 1 30 1 .ooo 0 0 0 * • 1 .ooo 0 0 0 w m m 1 .o o o 1 1 1 1 1 1 1 40 1 .ooo 4 b 0 0 1 .ooo 1 .ooo .ooo 1 1 1 1 1 1 141 1 .ooo 1 .ooo 1 .ooo 1.000 1 .ooo 1 1 1 1 1 1 1 42 0 0 0 ▼ » 1,000 1 .ooo 1 .ooo 1 .ooo 1 1 1 1 1 1 143 lo |C i° 1 .ooo 1 .ooo 1 .ooo 1 .ooo 1 1 1 1 1 1 144 1 O O O 1.000 1 .000 1 .ooo 1 .ooo 1 1 1 1 1 14S c 0 0 1.000 1 .ooo 1 .o o o 0 a q: 1 1 1 1 1 1 40 1 .ooo 1 .ooo 1 .ooo 1 .ooo 1 .ooo 1 1 1 1 1 1 47 o 0 0 1 .ooo 1 .ooo 1 .ooo 1 .ooo 1 1 1 t 1 1 14ft 1 .ooo 1 .ooo 1 .ooo 1 .O O O 1 .ooo 1 1 1 1 1 1 .14® 1 .ooo 1 .ooo 1 .ooo 1 .ooo 1 .ooo 1 1 1 1 1 1 1 S O 1 .ooo 1 .ooo 1 .ooo 1 .ooo 1 .ooo 1 1 1 1 1 1 i a i 1 .ooo 1 .ooo 1 .ooo 1 .ooo 1 .ooo 1 1 1 1 1 1 1 52 1 .ooo 1 .ooo 1.000 1 .ooo 1 .ooo 1 1 1 1 1 1 1 S3 1 .ooo 1 .ooo 1 .ooo 1 .o o o 1 .ooo 1 1 1 1 1 1 Table 2, Page 3, Right. Database for Estimating the Quality Criterion C \ 0 0 conl 23 mf 24 conf 34 0 0 6 o o o o o o OOP o o o o o o I s s o o o O O P O O P 57 1 58 05 007 023 B Q 4 .750 3 2 1 O f 071 I 02 2 2. 734 71 702 073 O B B O O B 1 66 733 71 S 760 038 O I 70 703 666 BO 703 703 730 173 174 173 1 70 O O .738 71 .75 600 1B1 720 .007 801 6 0 5 .050 a a 1 04 004 IBS £17 7 3 0 .730 s e e B37 1 B O 020 697 022 81 222. 76 1 02 B O B 823 702 753 735 50 30 B O B3 87 B O 842 oa 884 B B O 550 200 B O O 42 95 02 58 69 203 57 o a 204 003 45 Table 2, Page 4, Right. Database for Estimating the Q uality C riterion 1 — 1 eont 13 eonf 14 conf 23 conf 24 Conf 34 crll12 o rfin erll 1 4 cr|«23 crlt24 crt!34 Izosl .01 1 .549 .630 .515 .566 O o o o o o 1200 1 , 574 .884 .510 .567 .687 O o o o o o 1207 1 .500 .035 .501 .537 .553 O o o o o e > .044 .880 .620 .681 .896 O o o o o o .724 .015 .670 .696 .930 O o o o o o 210 .744 .Oil .725 .900 .935 O o o o o o 21 1 .700 .035 .706 .944 .954 O o o o o o 212 .aoe .050 .536 .967 .680 O o o o o o 213 .750 .054 .531 .068 .960 O o o o o o 214 .ooo .014 .O O O .055 .052 O o o o o o 215 .044 .040 .002 ,056 .061 O o o o o o 210 .037 .033 .007 .962 .053 o o o o o 0 217 ,*51 .015 .668 .025 .043 o o o o o o 210 .Bfll .035 .594 .04 1 .052 o o o o o o 210 .002 .600 -016 .025 .925 o o o o o o 220 .200 ,010 .012 .021 .021 o o o o o o 221 .019 .021 .014 .016 .025 o o o o o o 222 .042 .044 .015 .017 .91 4 o o o o o o 223 .005 .SO B .810 .617 .045 o o o o o o 224 .O O O .*77 .774 .760 .766 o o o o o o 220 .023 .806 .637 .761 .773 o o o o o o 220 .043 .022 .627 .761 .621 o o o o o o 227 .070 .052 .620 .71 1 .682 O o o o o o 220 .007 .074 .555 .737 .905 o o o o o o 220 .O O 1 ,080 .560 .701 .923 o 1 o o o o 230 .003 .054 -557 .697 .920 o 1 o o o o 231 .003 .075 .660 .569 .910 o 1 o o o o 232 .007 .OSS .022 .566 .050 o 1 o o o o 233 .007 .087 .020 .703 .964 o 1 o o o o 234 .007 .009 .038 .730 .971 o 1 o o o o 235 .035 .055 .022 .046 .943 o o o o 1 o 230 .ooo .050 ,051 .916 .070 o o o o o 237 .O O O ,090 .054 .026 .900 o 1 o o o 1 230 .003 .004 .085 .064 .903 o 1 o o o 1 230 .ooo .072 .052 .950 .901 o 1 o o o 1 240 .O O O .004 ,001 .087 .096 o 1 1 1 o 1 24 1 .O O O .007 .600 .092 .906 o 1 1 1 1 1 242 .O O O .002 .089 .092 .099 O 1 o 1 1 243 -O O O .005 .006 .098 .096 1 1 1 1 1 244 .ooo .090 ,600 .099 .996 ' 1 1 1 1 1 24 S .ooo 1 .O O O .990 .090 .999 1 1 1 1 1 240 .ooo .009 .099 .090 .099 1 1 1 1 1 247 1 .ooo .090 .990 .090 .990 1 1 1 1 1 1 240 1 .ooo 1 .005 .900 .909 1 .O O O 1 1 1 1 1 1 240 1 .ooo 1 .O O O .090 .900 1 .O O O 1 1 1 1 1 1 250 1 .ooo 1 .ooo 1 .O O O .090 1 .O O O 1 1 1 1 1 25 1 1 .ooo .O O O .009 .090 1.000 1 1 1 1 1 1 252 1 .ooo .B O O .090 .997 1 .ooo 1 1 1 1 1 253 1 .ooo .B O S 1 .O O O .006 .990 1 1 1 1 1 1 254 1 .ooo .005 1 .O O O .095 1.060 1 1 1 1 1 1 255 1 .ooo .009 1 .O O O .007 1 .O O O 1 1 1 1 1 1 Table 2, Page 5, Right. Database for Estimating the Quality Criterion -J O conl 13 oonf H I conl H Z conf 34 crlt12 crliia crlll 4 or|t?3 orl!24 erll34 2 S ft 1OOO .198 .990 .990 1 .ooo 1 1 1 1 o 1 2S7 1ooo .009 1OO O .990 1 .0 0 0 1 1 1 1 o 1 258 1 ooo 1ooo 1ooo .902 1 .ooo 1 1 1 1 1 1 250 - 1 l O O O .. 1 ooo ooo .990 1 .ooo 1 1 1 1 1 1 2 0 0 1ooo 1ooo 1OO O .990 1 .ooo 1 1 1 1 1 1 2 « 1 ooo _ 1 ooo 1ooo .990 1 .ooo 1 1 1 1 1 1 2 02 1 ooo 1ooo 1ooo .900 1 .0 0 0 1 1 1 1 1 263 _ _ _ _ _ _ _1 ooo 1ooo 1ooo .999 1 .0 0 0 1 1 1 1 1 2A 4 1ooo 1ooo 1ooo .999 1 .ooo 1 1 1 1 1 285 1ooo 1ooo 1ooo .990 1 .ooo 1 1 1 1 1 1 2 ft0 1ooo 1ooo 1ooo .999 1 .0 0 0 1 1 1 1 1 1 207 1 ooo 1ooo 1ooo .999 1 .ooo 1 1 1 1 1 I 2 00 1ooo 1.ooo 1ooo 1ooo 1 .ooo 1 1 1 1 1 1 2 00 1ooo 1 .ooo 1ooo 1ooo 1 .ooo 1 1 1 1 1 1 270 1ooo 1ooo 1ooo 1ooo 1 .ooo 1 1 1 1 1 271 1ooo 1ooo 1ooo 1ooo 1 .ooo 1 1 1 1 1 1 272 1ooo 1ooo 1ooo 1ooo 1 .0 0 0 1 1 1 1 1 1 273 1ooo 1ooo 1ooo 1ooo 1 .ooo 1 1 1 1 1 1 274 .OOO .990 ■®MU .940 .990 1 1 1 _ 1 .OOP 1ooo 1ooo .990 1 .ooo 1 1 1 1 1 270 1ooo 1 .ooo 1ooo .997 1 .ooo 1 1 1 1 1 277 1ooo 1 .ooo 1ooo .990 1 .ooo 1 1 1 1 1 270 1ooo 1ooo 1ooo .999 1 .000 1 1 1 1 270 1 .ooo 1ooo 1ooo .990 1 .ooo 1 1 1 1 t 2 0 0 t .ooo 1 .ooo 1ooo .999 1 .ooo 1 1 1 1 1 2 0 1 1 .ooo 1ooo 1ooo 1ooo 1 .ooo 1 1 1 1 1 1 2 0 2 1 .ooo 1 ooo 1 ooo 1 ooo 1 .ooo 1 1 1 1 203 1 .ooo 1 ooo 1 ooo 1 ooo 1 .ooo 1 1 1 1 1 1 204 ooo 1 ooo 1 ooo 1 ooo 1 .ooo 1 1 1 1 1 1 200 1 ooo .ooo 1 .ooo 1 ooo 1 .ooo 1 1 1 1 1 1 2 0 0 t ooo 1 .ooo ooo < ooo 1 .ooo 1 1 1 1 1 207 1 .ooo 1 .ooo ooo 1ooo 1 .ooo 1 1 1 1 1 1 2 0 0 .ooo 1.ooo ooo 1ooo 1 .ooo 1 1 1 1 1 2 0 0 1 .ooo 1.ooo ooo 1ooo 1 .0 0 0 1 1 1 1 1 2 0 0 1 .ooo 1.ooo .ooo 1.ooo 1 .0 0 0 1 1 1 1 1 1 2 0 1 1 .ooo 1.ooo .ooo 1 .ooo 1 .0 0 0 1 1 1 1 1 2 0 2 1 .ooo 1.ooo .ooo 1ooo 1 .0 0 0 1 1 1 1 1 1 203 1 .ooo 1 .ooo .ooo 1 .ooo 1 .0 0 0 1 1 1 1 1 1 204 1 .0 0 0 1 .ooo .ooo 1 .ooo 1 .0 0 0 1 1 1 1 1 1 2 0 0 1 .ooo 1 .ooo ooo 1 .ooo 1 .0 0 0 1 1 1 1 1 1 2 0 0 1 .ooo 1.ooo .ooo 1 .ooo 1 .0 0 0 1 1 1 1 1 1 207 1 .ooo 1.ooo .000 1 .ooo 1 .0 0 0 « 1 1 1 1 1 2 0 0 1.OOP ■ O O O .ooo 1 .ooo 1 .0 0 0 1 1 1 1 1 200 1 .OOP 1 .ooo .ooo 1 .ooo 1 .0 0 0 t 1 1 1 1 1 300 1 .ooo 1 .ooo .ooo 1 .OOP 1 .0 0 0 1 1 1 1 1 301 1 .ooo 1.OOP .ooo 1 .OOO 1 .ooo 1 1 1 1 1 1 T ooo 1 .OOP ■ OO O 1 .OOO 1 .0 0 0 1 1 1 1 1 1 303 1 .ooo 1 .ooo .OOO 1 .ooo 1 .ooo 1 1 1 1 1 304 i .ooo 1 1.0 00 1 .ooo 1 .ooo 1 .ooo 1 1 1 1 1 1 300 i .ooo ____ I .OOP 1 ■ Q O O 1 ■ OO O 1 .ooo 1 1 1 1 1 1 300 i .OOO C Z l jOOO^ 1 .ooo .OOP 1.000 1 1 1 1 1 1 Table 2, Page 6, Right. Database for Estimating the Quality Criterion conl 13 conl 14 conl 23 conl 24 conl 34 crlt12 crina crlt14 crl!23 crll24 crll34 307 1.000 1.000 t.000 1.000 1.000 1 i 1 1 1 1 306 1.000 1.000 1.000 1.000 1.000 1 i 1 1 1 1 309 1.000 1.000 1.000 1.000 1.000 1 1 1 1 1 1 310 1.000 1.000 1.000 1.000 1.000 1 1 1 1 1 1 311 1.000 1.000 1.000 1.000 1.000 1 1 1 1 1 1 312 1.000 1.000 1.000 1.000 1.000 1 i 1 1 1 1 Table 2 Page 7, Right. Database for Estimating the Q uality C riterion N ) 73 S ta te A v e ra g in g C h a n C om parison M e a n t-v a lu e s N = 6 , D F = 8 m e a n r (tran sfo rm ed ) 1 A W AK E S ta n d a rd 1 * 2 .8 9 2 .3 0 1 2 A W A K E S ta n d a rd 1 - 3 2 .4 1 4 .6 4 9 3 A W A K E S ta n d a rd 1 * 4 .6 7 2 .2 3 1 4 A W A K E S ta n d a rd 2 * 3 - . 1 5 1 - . 0 5 3 5 A W A K E S ta n d a rd 2 - 4 .9 9 4 .3 3 2 6 A W AK E S ta n d a rd 3 * 4 1 .0 5 0 .3 4 8 7 A W AK E B aye s 1 - 2 1 .0 0 6 . 3 3 5 ! 8 A W A K E B a y e s 1 - 3 2 .0 0 7 .5 7 9 0 A W A K E B aye s 1 - 4 .3 1 8 .1 1 2 1 0 A W A K E B a y e s 2 - 3 - . 4 2 0 - . 1 4 7 11 A W A K E B a y e s 2 - 4 1 .4 3 6 .4 5 3 1 2 AW AK E B a y e s 3 - 4 .2 2 0 . 0 7 8 1 3 SLEEP S ta n d a rd 1 - 2 .5 4 9 .1 9 1 1 4 SLEEP S ta n d a rd 1 - 3 3 .4 8 3 . 7 7 6 1 5 SLEEP S ta n d a rd 1 - 4 1 .6 2 7 . 4 9 9 1 6 SLEEP S ta n d a rd 2 - 3 - . 5 0 5 - . 1 7 8 1 7 SLEEP S ta n d a rd 2 - 4 .4 4 8 .1 5 6 1 6 SLEEP S ta n d a rd 3 - 4 1 .5 5 4 .4 8 2 1 9 SLEEP B aye s 1 - 2 - . 4 3 0 - . 1 5 0 2 0 SLEEP B aye s 1 - 3 2 .2 4 2 .6 2 1 2 1 SLEEP B aye s 1 * 4 1 .5 0 4 .4 6 9 2 2 SLEEP B aye s 2 - 3 - 1 .3 1 1 - .4 2 1 2 3 SLEEP B aye s 2 - 4 - . 1 0 4 - . 0 3 7 2 4 SLEEP 'M S M A N 'V A 'i U f f l V & V X l l V . t i 'J B aye s 3 - 4 VAVAV.W.UA\l*AWAVWVAvJ/AlAvi,' 1 .4 2 1 .4 4 9 Table 3. Crosscorelation of Between Channels INFORMATION TO USERS This manuscript has been reproduced from the microfilm master. UMI films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter face, while others may be from any type of computer printer. The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleed through, substandard margins, and improper alignment can adversely affect reproduction. In the unlikely event that the author did not send UMI a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note wifl indicate the deletion. Oversize materials (e.g., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand comer and continuing from left to right in equal sections with small overlaps. Each original is also photographed in one exposure and is included in reduced form at the back of the book. Photographs included in the original manuscript have been reproduced xerographically in this copy. Higher quality 6” x 9" black and white photographic prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact UMI directly to order. A Bell & Howell information Company 300 North Zeeb Road. Ann Arbor. Ml 48106-1346 USA 313/761-4700 800/521-0600 UMI Number: 1376461 UMI Microform 1376461 Copyright 1995, by UMI Company. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code. UMI 300 North Zeeb Road Ann Arbor, MI 48103

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Auditory brainstem responses (ABR): variable effects of click polarity on auditory brainstem response, analyses of narrow-band ABR's, explanations

PDF

Estimation of upper airway dynamics using neck inductive plethysmography

PDF

Respiratory system impedance at the resting breathing frequency range

PDF

Cross-correlation methods for quantification of nonlinear input-output transformations of enural systems using a Poisson random test input

PDF

A physiologic model of granulopoiesis

PDF

Design of a portable infrared spectrometer: application to the noninvasive measurement of glucose

PDF

Three-dimensional functional mapping of the human visual cortex using magnetic resonance imaging

PDF

Comparison of evacuation and compression for cough assist

PDF

Fault tolerant characteristics of artificial neural network electronic hardware

PDF

Pulse oximetry failure rates

PDF

Flourine-19 NMR probe design for noninvasive tumoral pharmacokinetics

PDF

Human Skeletal Muscle Oxygenation And Perfusion: Non-Invasive Measurement By Near-Infrared Spectroscopy

PDF

Functional water MR spectroscopy of stimulated visual cortex using single voxel

PDF

A computational model of NMDA receptor dependent and independent long-term potentiation in hippocampal pyramidal neurons

PDF

Propofol Effects On Eeg And Levels Of Sedation

PDF

A kinetic model of AMPA and NMDA receptors

PDF

Comparisons of deconvolution algorithms in pharmacokinetic analysis

PDF

Decomposition Of Neuronal Function Using Nonlinear Systems Analysis

PDF

A proposed role for the superior colliculus within the feedback loop of the brainstem saccade generator

PDF

Seizure activity does not impair LTP induction in young rat hippocampus

Asset Metadata

Creator Iatrou, Maria (author)

Core Title Auditory brainstem responses (ABR): quality estimation of auditory brainstem responsses by means of various techniques

School School of Engineering

Degree Master of Science

Degree Program Biomedical Engineering

Degree Conferral Date 1995-08

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag engineering, biomedical,OAI-PMH Harvest

Language English

Contributor Digitized by ProQuest (provenance)

Advisor Marmarelis, Vasilis A. (committee chair), Khoo, Michael Chee-Kuan. (committee member), Maarek, Jean-Michel (committee member)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c18-1802

Unique identifier UC11357914

Identifier 1376461.pdf (filename),usctheses-c18-1802 (legacy record id)

Legacy Identifier 1376461-0.pdf

Dmrecord 1802

Document Type Thesis

Rights Iatrou, Maria

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA