AU599272B2 - Improved evoked response audiometer for sleeping subjects - Google Patents

Description

V

599272 COMPLETE SPECIFICATION FOR OFFICE USE Application NRirmber: Lodged; Complete Specification Lodged: Accepted: Published: Priority: Related Art: Class Int. Class .rs 7 W o o 900 0 0 o a 0000 o 00 069 S 0 1 TO BE COMPLETED BY APPLICANT Name of Applicant: THE UNIVERSITY OF MELBOURNE Address of Applicant: Grattan Street, Parkville, Victoria 3052, Australia Actual Inventors: Lawrence Thomas COHEN and Field Winstfn RICKARDS Address for Service: SMITH SHELSTON BEADLE 207 Riversdale Road Box 410) Hawthorn, Victoria, Australia Complete Specification for the invent ,on entitled: IMPROVED EVOKED RESPONSE AUDIOMETR FOR SLEEPING SUBJEFS The following statement is a full description of this invention, including the best method of performing it known to us: Page Our Ref: TNB:WB:6-jelbuni.pl 2 1 TITLE: IMPROVED EVOKED RESPONSE AUDIOMETER FOR SLEEPING 2 SUBJECTS 3 Field of the Invention: 4 This invention relates to an improved evoked response audiometer to be used with sleeping subjects, with special 6 reference to neonates, young children and mentally 7 handicapped persons.

8 Background of the Invention: 9 The diagnosis of deafness at an early stage is most 1C important to enable the early fitting of hearing aids and in S 11 the application of educational programs to assist language Q 0 aoo 12 development in the hearing impaired child, Current ooo" 13 procedures in the early diagnosis of deafness include the So" 14 "Cribogram" and brainstem evoked response.

Q a S 15 Auditory evoked potentials recorded from the scalp in 16 humans have now been described in many studies. These 17 potentials have been classified into three main groups.

0 00 00 18 These groups are: S 19 brainstem evoked potentials which are 20 approximately 0.5 microvolts in amplitude and occur during S 21 the first 10 milliseconds following the presentation of an 22 abrcpt sound stimulus, usually a click.

23 (ii) the middle latency responses which are 24 approximately two microvolts in amplitude and occur between 7 and 50 milliseconds following the presentation of a click 26 or tone pip and, 27 (iii) slow responses, about 10 microvolts in 28 amplitude, following the onset of a tone burst and have 29 latencies between 50 and 500 milliseconds.

881110,ltbspe.030,unimell.spe, i 3 1 Currently, the brainstem potential is receiving most 2 attention both as a neurological and an audiological tool.

3 It does, however, have the disadvantage of using abrupt 4 stimuli. This is necessary since this response reflects synchronous firing patterns in the auditory pathway in the 6 brainstem. Stimuli of slower onset fail to achieve the 7 synchrony necessary for the recording of the various peaks.

8 As a result of this limitation only high frequency hearing 9 information is measured.

The middle latency responses are also currently o. oo 11 receiving attention as a measure of low frequency hearing a°c A 12 with low frequency tone bursts being repeated forty times 0oo 13 per second to evoke a periodic response. This response is 0 0Q 14 affected by sleep and therefore has limited application in eo 0 S 15 the testing of babies.

16 The periodic 40 Hz middle latency responses are a 17 subgroup of the auditory steady-state evoked potentials.

18 These are periodic responses, recorded from the scalp to a 19 continuous periodically varying stimulus, for example an 20 amplitude modulated tone. The periodicity of the response 21 is the same as the period of the madulation waveform. These

I

22 can be recorded over a wide range of modulation and carrier 23 frequencies.

24 Several classes of subject can be tested by evoked potentials only (or most conveniently) during sleep, for 26 example infants and young children and people with mental 27 retardation.

28 Existing steady-state literature shows that for low 29 modulation rates (less than or equal to 60 Hz) responses are 881110,ttbspe.030,uAimell.spe,

S.

-4 1 variable in sleeping subjects and considerably smaller in 2 amplitude than for the waking state.

3 Responses at all modulation frequencies do decrease 4 with sleep or sedation but we have found that the background noise level of the EEG decreases dramatically at high 6 modulation frequencies and consistent responses remain 7 present. (see Figure 11 and Figure 12). This means that 8 detection of responses in sleep can easily be perFGrmed at 9 high modulation frequencies (greater than 60 Hz).

Summary of Invention and Objects: oo 12 It is therefore an object of the present invention to o00 0 12 provide an improved evoked response audiometer which will 600 ooo 0 13 make use of the optimum modulation frequencies for steadyo 00 o 14 state evoked potential testing during sleep, to allrw the 0 0 15 efficient assessment of hearing of a variety of difficult- 16 to-test patients. These include neonates, infants, young 17 children and mentally retarded patients.

o4 18 The invention provides an evoked response audiometer 19 comprising means of supplying to the patient an auditory s 20 signal consistiag of a carrier frequency which is S 21 periodically modulated (for example amplitude modulation, oQ, o 22 frequently repeating tone bursts or tone pips, frequency 23 modulation or beats) such that the stimulus is at least 24 substantially frequency specific, said auditory signal being presented for a sufficiently extended period of time to 26 enable phase-locked steady-state potentials to be evoked in 27 the brain, means for sampling the brain potential signals 28 evoked by said signal, and means for analysing said brain 29 potentials to determine whether phase-locking of said brain 881110,ltbspe.030,unimell.spe, 5 5 1 potentials to the modulated auditory signal has occurred, 2 said auditory signal means being controlled so that said 3 auditory signals are periodically modulated at frequencies 4 in excess of 60 Hz, the frequency of modulation being varied in a generally increasing manner for auditory signals of 6 higher frequencies.

7 In a preferred form, said frequency of modulation is 8 about 60-115 Hz for auditory signals having frequencies less 9 than or equal to 1.5 kHz, and said frequency of modulation is about 65-200 Hz (or more) for auditory signals having s...IO 11 frequencies in excess of 1.5 kHz.

o 0 oo° 12 It should be appreciated that the modulation frequency 13 used will depend on the frequency of the auditory signal as o0 o 1 4 well as on the subject being tested. As an indication of 00 the range of modulation frequencies which may be used, the 16 following table is provided: 17 Normal sleeping neonates 18 Auditory Signal Modulation Frequency 19 500 Hz: about 60-140 Hz, preferably 65-95 20 Hz, and most preferably about 72 Hz 21 1.5 kHz: about 60-165 Hz, preferably 75-110 22 Hz, and most preferably about 85 Hz 23 4 kHz: about 65-200 Hz, preferably 85-110 24 Hz, and most preferably about 97 Hz Hence, 60-165 Hz for CF 1.5 kHz 26 65-200 Hz for CF 1.5 kHz 27 Normal sleeping adults 28 250 Hz: about 70-130 Hz, preferably about 29 80-115 Hz, and most preferably about 881110, Itbspe.030,unimell.spe, -6 1 85-95 Hz 2 500 Hz: about 70-180 Hz, preferably about 3 80-115 Hz, and most preferably about 4 85-95 Hz 1 kHz: about 70-200 Hz, preferably about 6 80-115 Hz, and most preferably about 7 95 Hz 8 2 kHz: about 75-200 Hz, preferably about 9 85-195 Hz, and most preferably about 105-160 Hz n.o. 11 4 kHz: about 75-200 Hz, preferably about 12 85-200 Hz, and most preferably 13 about 120-190 Hz 4 44 14 Hence, 70-180 Hz for CF 1 kHz e4 75-200 Hz for CF 1 kHz 16 It is expected that as infants mature, their responses 17 will become more like those of adults. Accordingly, there 18 will be a shift of optimum MF ranges 19 Thus other modulating frequencies will be determined 20 experimentally for other types of putients.

o 21 The use of modulation frequencies in excess of 60 Hz to 22 evoke the responses allows the most efficient detection of a 23 response in the type of patient being tested and at the 24 carrier frequency being used. The system may be designed to choose the optimum modulation frequency automatically, based 26 on the type of subject and the carrier frequency used.

27 The audiometer embodying the present invention has the 28 advantage over prior art audiomears in that it may make use 29 of the widest possible range of modulation types (limited 881110,1tbspe.030,unimell.spe, 7 1 only by the requirement of reasonable frequency 2 specificity), in that it makes use of the modulation 3 frequencies that allow most efficient detection of a 4 response during sleep (namely those in excess of 60 Hz) and that it employs a frequency specific stimulus. It also 6 detects a response in real-time enabling the transfer to a 7 new stimulus automatically.

8 The brain potentials are preferably recorded by means 9 of electrodes on the vertex or forehand and on the mastoids of the patient. In the preferred embodiment, the patient is 11 presented with a band limited tone buzst or a tone that is 4 4 12 simultaneously amplitude and frequency modulated or an 13 amplitude modulated tone. The EEG signal is Fourier 0 4 14 analysed to extract the components at the modulation 15 frequency and its second harmonic, as these have been found 16 to be the predominant components of the response. The use 17 of low-pass filters following the multiplication of the EEG 1 4 18 signal by the modulation frequency waveform and its second Oo 19 harmonic provides a time "window" which samples the EEG 6° 20 waveform for an interval of, typiclly, 64 periods of the 21 modulation waveform. The filters are sampled twice every 22 such interval (that is, typically, once every 32 modulation 23 periods) resulting in a set of samples, each of which 24 contains measurements of amplitude and phase of the EEG components present in very narrow frequency bands centred on 26 the modulation frequency and its second harmonic. The phase 27 measurements are made relative to the modulation frequency 28 envelope.

29 The sets of samples are analysed to provide mean 881110,ltbspe.030,unimell.spe 8 1 2 3 4 6 7 8 9 0 i o a 12 13 0 '0 o o 13 14 0 0 o 4 0" 15 16 t 17 18 19 S 21 I I IItII 22

I

23 24 26 27 28 29 amplitudes, mean phases and probabilities that the distributions of the angles of the samples could have occurred by chance in the absence of a phase-locked response). The said probabilities enable the system to decide in real-time whether a response is present. As the system is able to vary both the loudness and the carrier frequency of the auditory signal presented to the patient, it allows objective testing of hearing, which may be performed automatically.

Brief Description of the Drawings: One preferred embodiment of the invention will now be described with reference to the accompanying drawings in which: Figure 1 is a schematic diagram of a preferred audiometer embodying the invention; Figure 2 is a more detailed block diagram of part of the audiometer of Fig. 1 showing the manner in which most stimulus data transmission from computer to audiometer is effected ('strobed' data transmission); Figure 3 is a block diagram of the stimulus generation portion of the audiometer; Figure 4 is a block diagram of the signal detection portion of the audiometer; Figure 5 is a block diagram of a part of the audiometer of Fig. 1 showing the manner in which further data transmission from computer to audiometer is effected ('static' data transmission); Figure 6 is a timing diagram showing the ADC timing for sampling of the Hanning filters; 881110,ttbspe.030,unimell.spe, l-il II- il i i -ICU- I lla( IE~ 1 2 o 'u 0 o 4 i V O o 0 o 0 o VIo 4 4a 0 0 0 o 9 Figure 7 is a more detailed block diagram showing the circuit for generating the carrier, modulation sine waves and quadrature components of the stimulus signal; Figure 8 is a more detailed block diagram showing the ciruit for controlling the modulation and level of the stimulus signals; Figure 9 is a more detailed block diagram showing the signal detection circuitry; Figure 10 shows average response amplitudes versus modulation frequency for awake adult subjects with amplitude modulated stimuli at various carrier frequencies (55 dBHL binaural); Figure 11 shows average response amplitudes versus modulation frequency for awake and asleep adult subjects with amplitude modulated stimuli at 4 kHz carrier frequency dBSL binaural); Figure 12 shows narrow-band background EEG noise as a function of modulation frequency for awake and sleeping adult subjects; Figure 13 shows a detection efficiency function for dBSL binaural stimulation at 4kHz in awake and sleeping subjects, as a function of modulation frequency; Figure 14 shows the detection efficiency function of the method at various carrier frequencies for awake and sleeping subjects for 30 dBSL binaural stimulation as a function of modulation frequency, and Figure 15 shows the detection efficiency function of the method at three suitable carrier frequencies for sleeping neonates for 55 dBHL monaural stimulation as a 881110,1tbspe.030,unimell.spe, 1 function of modulation frequency.

2 Referring to Figure 1 uf the drawings, the presently 3 preferred prototype audiometer configuration is shown 4 schematically. The audiometer is controlled by any suitable microcomputer, such as an IBM IT-type, while the EEG 6 signals taken from the patient are amplified by a suitable 7 amplifier, such as a Madsen BPA 77. Selection of carrier 8 frequency and modulation frequency (from, the computer 9 keyboard or automatically under program control) sets the frequency of clock oscillators in the "Function Generator" S 11 section. These clock oscillators control the frequencies of S 12 numerous tracking filters (switched capacitor and other) in 13 both the "Function Generator" and "Detector" sections. The o oO S 14 non-switched capacitor filters are all similar 12dB per S 15 octave low-pass filters which perform such tasks as anti- 16 aliasing as described further below. The computer catt also 17 vary the loudness of the stimulus to each ear separately 18 over a range of approximately -10 to 120 dBHL. This is made S 19 possible by the "Range Setter" section which attenuates the S 20 output to the headphones in increments of 40 dB. Such high 21 sound levels allow the use of the system in measuring the 22 hearing of profoundly deaf infants and children. Protection 23 against accidental overly loud stimuli is threefold: 24 software level limitation, preset hardware cutout and disablement of the highest 40 dB of the system. The oottput 26 of the function generator drives two buffer amplifiers which 27 drive the headphones (or other transducers such as 1Etymotic 28 Research Tubephones).

29 The EEG resulting from the stimulus is picked up using 881110, Itbspe.030,unimell.spe, i i"

'I

I'

o oc os0 o t o n. 0 00r 00le 00 1r 00 0o 11 silver-silver chloride (or gold) electrodes on the scalp, is amplified by the EEG amplifier, which is a high impedance, high gain, low noise preamplifier. As described further below, the output of the preamplifier passes through an anti-alias filter and preliminary band-pass filters (one each for the modulation frequency and its second harmonic).

The signal (in two channels for the two harmonics) is further amplified and multiplied by the sine and cosine components of the two harmonics. The resulting four channels pass through four anti-alias filters, followed by four low-pass filters. A wide variety of low-pass filters would suffice for this role but here four-pole filters are employed, giving rise to an effective time "window" which approximates the shape of a Hanning window. The duration of the "window" is, typically, 64 modulation periods.

Typically, a sampling pulse is generated every 32 cycles of the modulation waveform, resulting in optimum sample overlap so as to have the maximum inter-sample time while maintaining 100% sampling efficiency. The outputs of the low-pass filters provide Fourier analysis of the EEG at the modulation frequency and its second harmonic, both in-phase (cosine) and 90 0 -shifted (sine) terms. The detection of a response would usually require the collection of from 20 to 256 samples.

The computer analyses the samples (which contain amplitude and phase information at the two frequencies) to obtain the mean amplitudes and mean phases as well as probabilities that the sample distributions might have arisen by chance in the absence of a phase-locked 8811t0,1tbspe.O3Ounimellspe, 12 1 response). This processing is done in real time and the 2 computer stops sampling as soon as the required i.tistical q criteria for the presence of a response have been satisfied.

4 Referring now to Figures 2 to 9 of the drawings, which show the essential elements of the audiometer of Fig. 1 in 6 greater detail, data determining carrier frequency (CF), 7 modulation frequency frequency modulation depth (DF), 8 amplitude modulation depth stimulus intensity to left 9 ear (LA) and intersity to right ear represented in Fig. 2 as "output from computer" 1, are strobed into pairs o00ooo 11 of Hex D flip-flops, for example CF1 and CF2, there being 12 twelve bits of information for each of the above parameters.

0100 oo' 13 The number appearing in each circuit block refers to a o 00 0 14 commercially available circuit component, Prior to the 0 0 0 0 0 00 15 transmission of this data, the audiometer circuit must be 16 informed as to which pair of data flip-flops should be 0 00 0 0 o 17 strobed, for example CF1 and CF2 (similar pairs of flip- 0 0 0 o 3 0 Uo 18 flops (not shown) are provided for the remaining parameters 0 a o a 19 MF, DF, DA, LA and RA), and which channel of the system it O01 D 20 to be addressed (in the present prototype system there is o 1 i u21 ly one channel but there is provision for more if 0 0 0o o 22 desired). This initialising information is strobed into the 23 Hex D flip-flops and 3. For any strobing to have effect, 24 bit 15 must be set. For the initializing strobing into flipflops 2 and 3, bit 13 must be low. For initialization, the 26 bits selecting CF (bit 1) and No. 1 channel (bit 7) must be 27 set (and no others among bits 1 to 12). Then the computer 28 must strobe bit 14 with a positive-going pulse (duration 29 approximately 4 us). The setting of bit 1 of both flip-flop 881110, tbspe.030,jnimell.spe, o U cUUILu y si.gna1s are periodically modulated at frequencies in excess of Hz, the frequency of modulation being increased for auditory signals of higher frequencies.

/2 13 1 2 and flip-flop 3 (in the present example) enables CF data 2 to be strobed into the first channel, the next step.

3 Once the desired CF data has been put onto bits 1 to 4 12, and bit 13 has been set, the actual data can be strobed into flip-flops CF1 and CF2 by a second strobing of bit 14.

6 Strobing pulses are directed to the appropriate pairs of 7 data flip-flops by AND gates 4 to 9. The data for MF, DF, 8 DA, LA and RA are transferred similarly to their respective 9 pairs of flip-flops (not shown) and all the data remains latched into the appropriate flip-flops CF1 and CF2 etc.

11 until it is deliberately altered or the power supply turns o 12 off. The effective strobing pulse is independent of the 13 duration of the pulse on bit 14, being transformed by a 14 monostable circuit 10 to 10 us duration and slightly delayed in its onset by delay circuit 11, 16 When bit 15 is low, bits 1 to 12 are available to be 17 used for the other, "static", data transmission by means of 18 the circuit detailed in Figure 5. Bits 1 to 1, 'en control 19 sampling interval, input selection (EEG, zero or test), left and right ear level ranges and analog waveform selection for 21 input to an analog-to-digital converter (ADC) 12, and must 22 be held at thsir appropriate values by the computer. Simple 23 RC circuits (not shown) ensure that at appropriate points ini 24 the circuitry, the values cannot change during the brief duration of strobing operations. A programmable var:Lable 26 range amplifier (PVRA) circuit (not shown) associated with 27 left and right level ranges (bits 5,6 and 11,12) allows the 28 activation of relays within that circuit; to give different .9 degrees of attenuation 40 or 80 dB) to the output 881110,Itbspe 030,unimell.spe, ~LL. i_ I-I_ i L _IL i r C~ 14 o i 2 3 4 6 7 8 9 11 12 13 14 16 17 18 O" 19 o 20 21 22 23 24 26 27 28 "9 signals which will drive the left and right headphones.

Ihese level ranges allow a stimulus dynamic range of 130 dB, from 0 to 130 dBSPL.

Bits 3 and 4 allow selection, via an electronic switch 13, of analog input signal: EEG, zero or a test signal. Bits 7,8 and 9 allow selection, via two electronic switches 14 and 15, of any one of eight possible analog outputs to the ADO 12. These outputs include the outputs of final "Hanning" filters, yet to be described.

Selection of carrier frequency and modulation frequency (from the computer keyboard or automatically under program control) sets the frequency of clock oscillators 17 and 16 in the stimulus generation section of the audiometer, as shown in Figure 3. The outputs of the relevant flip-flops CF1, CF2 etc. provide digital inputs to multiplying digitalto-analog converters (DAC) 20 cd and 18, (as well ai 19, 36 and 37), the outputs of which are the products of their digital and analog inputs. The main outputs produced are currents, Ic and Im, which control the frequencies of two current controlled oscillator circuits 17 and 16. The square wave outputs of the oscillators 17 and 16 (approximately -14 to +14V) are modified by known signal shaping means (not shown) to give final square wave cl "aveforms of 0 to Voltages -Vin(c) and -Vin(m) proportional to Ic and Im, are also generated and are used as inputs to voltage-tocurrent conversion circuits (not shown), whooe outputs are proportional to Ic and Im, respectively. The currents are then used to control the operating frequencies of numerous a 881110,tbspe.030,unimelIspe, i i C~; 15 1 "scavenging" low pass filters, as described further below, 2 used to perform such tasks as anti-aliasing and removal of 3 clock noise from the outputs of switched-capacitor filters.

4 The clock oscillators 17 and 16 control the frequencies of numerous tracking filters, both switched capacitor and 6 "scavenging" in the stimulus generation circuit of Figs. 3 7 and 7 and the signal detection circuit of Fig. 4. The square 8 wave clock output of oscillator 17 (16) is divided in 9 frequency by 64 (256 and 128) by frequency counter or divider circuits 54 and 55 to provide the actual carrier 11 frequency CF (modulation frequency MF and its second 0 4 ao 12 harmonic), which are passed through switched capacitor o 13 filters 53, 21 and 22 to become sinusoidal and, in the case 0 ao 'o 14 of the modulation frequency, to generate quadrature cosine 04 Q 4 o 15 and sine components (see Figs. 3 and The quadrature 16 output components are simply taken from the lowpass (LP) and i 17 bandpass (BP) outputs of the switched capacitor filters 21 0 4no S 18 and 22 (Fig. The sinusoidal outputs 01 and 02 from the S 19 stimulus generation circuit, as best shown in Fig. 7, r44 4« 20 provide the actual carrier and modulation waveforms and 21 these have had residual clock noise filtered out by low pass 22 filters 57 and 58. The remaining outputs 03 to 06 provide 23 inputs for signal detection multipliers 31 to 34 inputting 24 to anti-aliasing filters 23 to 26 (Figures 4 and It is not necessary to remove clock noise from these outputs.

26 Amplitude modulation of the stimulus signal, of 27 variable modulation depth, is possible using analog 28 multiplicaticn of the carrier by a constant plus a term 29 proportional to the modulation waveform (Figure 3 and Figure 881110,ltbspe.030,unimellspe, i_ 16 1 The amplitude modulation depth is proportional to the 2 digital input to a multiplying DAC 35. A similar frequency 3 modulating signal (constant plus a term proportional to the 4 modulation waveform) is generated, with the frequency modulation depth determined by the digital input to the 6 multiplying DAC 19. This signal is taken to the analog input 7 of the multiplying DAC 20 via a mixer, resulting in a 8 modulated current input to the current controlled 9 oscillation 17, the output of which is frequency modulated.

The sound levels to the two ears are controlled by the o 11 two muliplying DACs 36 and 37, best seen in Figure 8. The S 12 carrier signal, which may already be frequency modulated, is 13 multiplied by the amplitude modulation signal from 35 using 14 multiplier 38. Switch means S 1 and S 2 (Fig. 8) allow the 0 00 0o o 15 amplitude modulation to be switched on or off separately for 16 the two ears, enabling binaural stimuli, with one ear 0 o o o 17 modulated and one not.

0 C 1 a 0 18 The stimulus signals are bffered by a unity gain a 19 circuit (not shown), which is able to drive the earphones to 20 levels of up to 130 dBSPL, in the case of TDH-39 earphones.

21 The computer controlled relays in the PVRA circuitry (not 22 shown) allow attenuation of 0 dB, 40 dB or 80 dB.

23 Accordingly, output levels are available from about 0 dBSPL 24 to 130 dBSPL, as the level control DACs 36 and 37 easily allow control over a range of 50 dB. The availability of 26 such high sound levels allows the use of the system in 27 measuring the hearing of profoundly deaf infants and 28 children. Protection against accidental overly loud stimuli 29 is threefold: software level limitation, preset hardware 881110,ttbspe.030,unimellspe, i 17 o o 0000 con 0 00 a on 0 0 0 0 00 0 0 0 0 0 0 00 0oo0 0000.

0 0 0 0 000000 0 07 cutout and disablement of the highest 40 dB of the system, each performed in a known manner not shown in the drawings.

Similarly, tonebursts or more general modulated waveforms may be generated by multiplication of the carrier waveform by the output of a simple periodic envelope generator circuit (not shown, but preferably with linear ramp waveforms at onset and offset of each burst and a plateau region) or, more generally, by a modulation waveform stored in a random access memory (RAM), for example a UM6116 device. The rate at which the RAM waveform is read is governed by the previously nentioned modulation clock frequency. The RAM circuitry is similarly not shown.

Good combinations of large response amplitude and frequency specificity can be obtained using two sti'mulus types. One is a band-limited toneburst, the envelope of which is calculated and stored in a RAM (not shown). The resultant toneburst retains the five central harmonics of a normal tone burst, and also retains a fairly rapid onset/offset and a substantial quiescent period during each modulation cycle. The other is a combination of amplitude and frequency modulations, which gives more "character" to the "onset" of the waveform (relative to AM) and gives a combination of the good performance of AM at high carrier frequencies and the good performance of FM at low carrier frequencies.

The frequency specificity of any stimulus can be enhanced by the use of high-pass filtered masking noise to ensure that a high frequency region of the basilar membrane is not spuriously excited.

881110,ltbspe.030,unimell.spe, 18 o 0 000 oo 0 00 0 °o I o o 0 0 0 0 Referring to Figures 4 and 9, the output from the EEG amplifier (Fig. 1) is passed through a programmable antialias filter 39 via a variable gain amplifier 40 and then through two preliminary band-pass (switched capacitor) filters 41 and 42 (one each for the modulation frequency and its second harmonic). The two signal channels (for the two harmonics) then pass through further amplifiers 43,44 (Fig.

9) and are multiplied by the sine and cosine components of the two harmonics by means of the multipliers 31 to 34. The resulting four signal channels pass through the four antialias filters 23 to 26, followed by the four low-pass switched capacitor filters 27 to 30. The operating frequencies of these last "Hanning" filters are determined by their clock inputs CH (Fig. which are equal to the modulation clock frequency divided by 16, 32, 64 or 128, for sample lengths of 16, 32, 64 or 128 modulation periods, as shown in Figure 5. The "Hanning" clock frequency is further divided by 256 by divider circuit 74 (Fig. 5) to obtain pulses for instructing the computer to initijte sampling of the "Hanning" filter outputs.

A wide range of low-pass filter shapes would suffice for this role, including a single pole low-pass filter, but here four pole filters are employed (each comprising four cascaded identical single pole filters), giving rise to an effective sampling time "window" which approximates a Hanning windk'w, namely [1-cos(2?t/tH)]/2, where tH is the width of the window, or sample length (and the function. is only defined for 0< t tHj). The width of the "window" is, typically, approximately 64 modulation periods. Typically, a 881110,tbspe.030,unimell.2pe, 19 1 2 3 4 6 7 8 9 11 0 0 0 4 12 13 00 4 4 4 14 15 16 S17 18 19 S 20 21 22 23 24 26 27 28 29 sampling pulse is generated every 32 cycles of the modulation waveform, resulting in some degree of sample overlap. For a given inter-sample interval, ts, there is an optimum value of tH such that there is essentially 100% efficiency in the data collection, but minimum overlap of samples. If there are large gaps between the samples, clearly much information will be lost, but if the samples overlap substantially, the information contained in a given sample will not be independent from that contained in adjoining samples.

For the window shape in question, the optimum value of tH is approximately 2ts. If a single pole low-pass filter had been used, if ts had remained the same and 100% efficiency had been required, the amount of sample overlap would have been considerably greater. Viewed another way, if both single pole f lters and 4 pole filters were arranged to have 5% sample overlaps, the sampling efficiencies would be 60% and approximately 80%, respectively. Bits 7, 8 and 9 from the computer control the selection of the four filter outputs in turn. The computer then controls the sampling of each filter output by the analogue to digital converter (ADC) 12 (Fig. 4) under the control of an output selector 46. A timing diagram for the ADC sampling is shown in Figure 6.

The outputs of the low-pass filters 27 to 30 provide Fourier analysis of the EEG at the modulation frequency and its second harmonic, both in-phase (cosine) and 90 0 -shifted (sine) terms. The computer analyses the samples (which contain amplitude and phase information at the two 881110,1tbspe.030,unimell.spe, er. r- l r( -r i i i. iiil-ii ;liz .iY-:i L- 20 frequencies) to obtain the mean amplitudes and mean phases.

It also calculates probabilities that the sample distributions might have arisen by chance (ie. in the absence of a phase-locked response). This provides a quantitative measure of the likelihood of a hearing response being present. A function F(N) is defined, such that: F(N) (2/N) [E cos 8i]2 sin Gi] 2 i=l i=l od 0 0 0 9 0 0 0 00 S o e S0 0 00 0 00 i0 00 00 0 0 0 0 0 9 i S0 o o o 0 0 J 0 0 0 0 4 0 3 0 where N is the number of samples and 9 i are the individual sample angles. This function follows a Chi-squared distribution with two degrees of freedom. The .omputer can readily calculate the probability, using the function P% 100 exp. (-0.5027F), a good approximation to the Chi-squar6d (distribution), that the given distribution of angles could have arisen from random noise background, in absence of a response. If F(N) is large, the probability of the angle distribution having arisen from random noise is very small and the presence of a response is very likely.

In the case of moderate sample overlap, the statistics are distorted and a value Q, less than 1.0, can be found such that N N F(N) Q x [C cos e 2 [r sin i] 2 i=l i=l also follows a Chi-squared distribution. If samples are required not to overlap, as when tH=ts, the sampling efficiency is about 80%, which means that detection of a response takes times as long.

881110,ltbspe.030,unimell.spe, .I~-iib.l 21 0 ?0 0 00 0 oa 0 0 e n a 000 0 00 0 Su 0 oQ o 0 a 0 0 U o U o 0 000 0 J 00000U O* r> Thi. statistical processing is done in real-time between samplings of the output filters 27 to 30 and the computer stops sampling as soon as the required statistical criteria for the presence of a response have been satisfied.

The statistics to this point accurately describe the distribution of probability P (derived from that one would obtain from say 100 runs in the absence of a response, each run being sampled just once during its data collection.

For example, we would find, on average, that only 1 run would have P However, because we wish to identify the presence of a response as soon as possible during a run, we measure P continually, not just once during a run. This distorts the statistics and, indeed, the probabil, y of a run having P 1% at some stage of its data collection may be as high as 10% if the total number of sample points in a run is 256.

Numerous approaches to this problem could be taken, each with rather similar results. The approach taken in the present form of the invention was to require P 1% for 0.07N or 2 sample points, whichever is larger. Hence, the more points have been collected, the longer P must be 1%.

This gives a false hit rate of about 5% after 256 sample points. In addition, depending on the circumstances including the size of response expected, a response may not be allowed before a certain number of points have elapsed.

This number has been 64 with infants. This further reduces the percentage of false hits, by 1 or 2%.

A hearing threshold may be estimated by varying the loudness of the stimulus in, say, 1OdB increments until the 881110,1tbspe.030,unimell.spe, 1 I C 22 0 o0 4 00 4 o o 0o i 0 0

I

*4 4_ 44 sof:est sound is found that elicits a response.

Alternatively, several measurements may be made above threshold and the intercept of amplitude vs. level graph may be estimated. The detection of a response would usually require the collection of from 20 to 256 samples.

Figure 10 shows average response amplitudes (at the modulation frequency) vs. modulation frequency for five carrier frequencies in awake adults in response to amplitude modulated stimuli at 55 dBHL, binaural. There is a strong peak at 40 Hz modulation frequency for all CFs, followed by a minimum at 60-70 Hz and reduced activity above 70 Hz.

These results are representative of what is found with a wide variety of modulated tones, for example tone bursts.

Even pure FM gives quite similar results, especially at low carrier frequencies.

Figure 11 shows a similar curve for awake adult subjects at 4 kHz, 30 dBSL binaural AM. Again there is the Hz peak followed by diminished activity, not much above the noise floor of the measuring system (not shown). Also shown in this Figure is a comparable curve for sleeping adult subjects. The 40 Hz region, represented by the circle, is dramatically reduced compared to the waking state. The amplitude at higher modulation rates is also reduced but not nearly as much as in the 40 Hz region.

The "detectability" of a response depends primarily on amplitude and noise, in this case the amount of random background EEG noise that is detected by the very narrow band filters implicit in the Fourier analysis. Figure 12 plots the noise as a function of modulation frequency for 881110,1tbspe.030,unimell.spe, 23 1 awake and sleeping subjects. In sleep the noise is reduce' 2 at all frequencies shown, but dramatically 3o at higher 3 frequencies. Accordingly, the signal to noise ratio 4 actually increases with sleei at higher modulation frequencies. We can calculate a Detection Efficiency 6 Funct, which is proportional to the reciprocal of the time 7 taken to detect a response. It is given by: 0

(S/N)

2 X Fm 9 where S is the amplitude of the response, N is the noise level and F. is the modulation frequency. The results at 4 11 ktlz from Figure 11 give rise to Detection Efficiencies as 12 shown in Figure 13. Not only are the higher modulation 0 13 rates during sleep dramatically better than 40 Hz, but we 14 can expect to detect a response during sleep in about half 0 15 the time it would take with the optimum modulation frequency S 16 in the awake state.

17 Similar curves are shown in Figure 14 for five U~s with 18 combined AM and FM stimuli (30 dBSL, binaural). 4 kHz o19 benefits most from the higher modulation rates but even at 500 Hz frequencies in the 80-10Q liz region are a Ii t tIe S21 better on average than ot 40 Hz. 40 11z amplitudes are not 22 only small but very variable in sleep, Clearly, different 0 23 U~s will be best suited by different modulation rates.

24 In Figure 1.1 Detection ',-fficiencies are shown for averages over 10 neonates (less than I week old), for 26 monaural stimulation with combined AM and FM stimuli at 27 dBH-L at three carrier frequencies, A g a in, op t imu m 28 modulation rates are found to be in excess of 60 Hz.

29 The claims f~orm part of the disclosure of this 881111, 1 tbspe .030, uttimelau spe, 24 1 specification.

0 0 0 0 881l1ltbspe.30ttnmeiu~spe,

Claims (11)

1. An evoked response audiometer comprising means of supplying to a sleeping patient an auditory signal consisting of a carrier frequency which is periodically modulated such that the stimulus is at least substantially frequency specific, said auditory signal being presented for a sufficiently,extended period of time to enable phase- locked steady-state potentials to be evoked in the brain, means for sampling the brain potential signals evoked by said signal, and means for analysing said brain potentials to determine whether phase-locking of said brain potentials to the modulated auditory signal has occurred, said auditory signal means being controlled so that said auditory signals are periodically modulated at frequencies in excess of 1: Hz, the Frequency of modulation being increased for auditory signals of higher frequencies.

2. The audiometer of claim I, wherein said frequency of modulation is about 60-115 Hz for auditory signals having frequencies less than or equal to 1.5 kHz, and said frequency of modulation is about 65to at least 200 Hz for auditory signals having frequencies in excess of 1,5 kHz,

3. The audiometer of claim 1, wherein said frequency of modulation is selected according to the following table: Normal s1ee.ing neonates Audio SiJLnalI Modulation Frequency 500 Hz: about 60-140 Hz, kiz: about 60-165 Hiz, 4 4 kltz: about 65 to at least 200 Hz, 9002016, pdspe.004,unimell.spe, S26 Normal sleeping adults 250 Hz: about 70-130 Hz, 500 Hz: about 70-180 Hz, 1 kHz: about 70-200 Hz, 2 kHz: about 75 to at least 200 Hz, 4 kHz: about 75 to at least 200 Hz,

4. The audiome.ter of claim 3, wherein for maturing infants the modulation frequency for each auditory signal is o a gradually increased with age towards the modulation 0 0 0 O.0 frequency specified fcr adults. 0 0 0o o 5. The audiometer of claim 3 or 4, wherein said auditory .O 0 S signal supplying means is controlled by a computer programmed to select the most appropriate modulation frequency for each stimulus frequency depending on the S"156 nature of the patients and their state of arousal, 0

6. The audiometer ot claim 1, wherein said sampling and analysing means comprise means for multiplying said brain 000I potential signals by said modulation frequency waveform and its quadrature component and by the waveform of the second 20 harmonic of the modulation frequency and its quadrature component to produce product waveforms, means for low-pass filtering said product waveforms to provide a time window which samples the brain potential for a predetermined interval to provide sets of Fourier analysis samples containing amplitude and phase data in narrow bands centred on the modulation frequency and its second harmonic.

7. The audiometer of claim 6, wherein said analys.ng means further comprises means for analysing said Fourier analysis x samples to extract mean values of the amplitudes and phase CE V 9002016,tpdspe.004,unimell.spe, L 27 angles of said signals, means for extracting from said mean values of said phase angles the probabilities that the distributions of said phase angles could have occurred by chance, whereby the existence of said phase locking can be determined.

8. The audiometer of claim 7, wherein said Fourier analysis is performed by filters onerating according to an approximation of the R-irning fun tion [l-cos(2?Wt/tH)]/2 0 where tH is the width of said time .indow. 0.,10 9. The audiometer of claim 8, wherein said time window has o 0 o 0 a width of approximately 64 cycles and the computer samples 0000 a the signals every 32 cycles to provide sample overlap. The audiometer of claim 9, wherein said probability is extracted according to the tunc'ons: N N F(N) Q x cos @i12 sin ei] 2 i=l i=i A A where Q is a constant 1.0, being approximately 0.625 and P% 100 exp. (-0.5027F)

11. A method of testing the hearing of a sleeping patient comprising the steps of supplying to said sleeping patient an auditory signal consisting of a carrier frequency which is periodically modulated such that the stimulus is at least substantially frequency specific, said auditory signal being presented for a sufficiently extended period of tir.c to .i enable phase-locked steady-state potentials to be evoked in 9002016,Ipdspe.004,unimel1.spe, 28 multiplication of the carrier by a constant plus a term 29 proportional to the modulation waveform (Figure 3 and Figure 8 8 1110,Itbspe.030,unimell-spe, r- I :I _i j :I ~I 28 o00sn 0 0 o 0 0 00 0 0 the brain, sampling the brain potential signals evoked by said signal, analysing said brain poLtntials to determine whether phase-locking of said brain potentials to the modulated auditory signal has occurred, said auditory signals being periodically modulated at frequencies in excess of 60 Hz, the frequency of modulation being increased for auditory signals of higher frequencies.

12. The method of claim 11, wherein said frequency of 0 modulation is selected according to the following table: 0 0 Normal sleeping neonates 0 o, Auditory Signal Modulation Frequency 0 500 Hz: about 60-140 Hz, 01 kHz: about 60-165 Hz, 4 kHz: about 65 to at 1 east 200 Hz, 1s Normal sleeping adults 250 Hz: about 70-130 Hz, 500 Hz: about 70-180 Hz, 1 kHz: about 70-200 Hz, 2 kHz: about 75 to at least 200 Hz, 4 kHz: about 75 to at least 200 Hz,

13. The method of claim 11, wherein said brain potential signals are multiplied by said modulation frequency waveform and its quadrature component and by the second harmonic of the modulation frequency and its quadrature component to produce product waveforms, low-pass filtering said product waveforms to provide a time window which samples the brain potential for a predetermined interval to provide sets of samples containing amplitude and phase data arrow bands centred on the modulation frequency and its second harmonic, ii 0 0 9002016,lpdspe.004,unimell.spes f

29- said low-pass filtering providing Fourier analysis of said product waveforms to produce mean values of the amplitudes and phase angles of said signals, and extracting from said mean values of said phase values the probabilities that the distributions of said phase angles could have occurred by chance, whereby the existence of said phase locking can be determined. 14. The method of claim 13, wherein said probability is 0 00- 00 o extracted according to the functions: N N 0 0 o F(N) Q x cos i]2 [3 sin 6i] 2 oo i=l i=l where Q is a constant 1.0, being approximately 0.625 and P% 100 exp. (-0.5027F) 15. The audiometer of any one of claims 3 to 10, wherein said modulation frequeicies are Normal sleeping neonates Auditory Signal Modulation Frequency 0 00 s *20 500 Hz: about 65 95 Hz kHz: about 75 110 Hz 4 kHz: about 85 110 Hz Normal sleeping adults 250 Hz: about 80 115 Hz 500 Hz: about 80 115 Hz 1 kHz: about 80 115 Hz 2 kHz: about 85 195 Hz 4 kHz: about 85 to at least 200 Hz A 9002016,!pdspeimell e E 9002016, I pdsp6.004,unimell.spe, 28 ensure that a high frequency region of the basilar membrane 29 is not spuriously excited. 881110,!tbspe.030,unimell.spe, K- 16. The audiometer of any one of claims 3 to 10, wherein said modulation frequencies are Normal sleeping neonates Auditory Signal Modulation Frequency 500 Hz: about 72 Hz kHz: about 85 Hz 4 kHz: .about 97 Hz Normal sleeping adults o 250 Hz: a.out 85 95 Hz 500 Hz: about 85 95 Hz 1 kHz: about 95 Hz o o o o 2 kHz: about 105 160 Hz 4 kHz: about 120 190 Hz 17. The method of any one of claims 11 to 14, wherein said frequency of modulation is selected according to the following table: Normal sleeping neonates Auditory Signal Modulation Frequency 500 Hz: about 65 95 Hz 'a o0 S0 1.5 kHz: about 75 110 Hz 4 kHz: about 85 110 Hz Normal sleeping adults 250 Hz: about 80 115 Hz 500 Hz: about 80 115 Hz 1 kHz; about 80 115 Hz 2 kHz: about 85 195 Hz 4 kHz: about 85 to at least 200 Hz 18. The method of any one of claims 11 to 14, wherein said frequency of modulation is selected according to the 9002016,!pdspe.004,unimell.spe, i 31 following table: Normal sleeping neonates Auditory Signal Modulation Frequency 500 Hz: about 72 Hz 1.5 kHz: about 85 Hz 4 kHz: about 97 Hz Normal sleeping adults 250 Hz: about 85 95 Hz 500 Hz: about 85 95 Hz ,10 1 kHz: about 95 Hz o 2 kHz: about 105 160 Hz o 4 kHz. about 120 190 Hz 19. An audiometer substantially as hereinbefore described with reference to the accompanying drawings. 1,5 20. A method of testing the hearing of a sleeping patient S' according to any one of claims 11 to 14, 17 or 18 o .P s substantially as hereinbefore described. DATED THIS 5th April, 1990 a SMITH SHELSTON BEADLE Fellows Institute of Patent Attorneys of Australia. Patent Attorneys for the Applicant THE UNIVERSITY OF MELBOURNE 9002016,Ipdspe.004,unimell.spe, af. i^