The dynamic range of TMJ sounds

Transcription

1 Journal of Oral Rehabilitation ; The dynamic range of TMJ sounds S. E. WIDMALM*, D. DJURDJANOVIC & D. C. MCKAY *Departments of Biologic and Materials Sciences, School of, Dentistry and of Electrical Engineering & Computer Science, College of Engineering, University of Michigan, Department of Mechanical Engineering, University of Michigan, and 9201 Sunset Blvd., Los Angeles, CA, USA SUMMARY It is of clinical interest to record the amplitudes of temporomandibular joint (TMJ) sounds. The aim was to test the hypothesis that sealing the meatus, when placing a microphone in the ear canal affects such recording by increasing the sound pressure level (SPL). Bilateral recordings of 249 TMJ clickings were made from three subjects, using sampling rates of 48 or 96 khz and 24 bits A/D conversion, with and without the ear canals sealed by Silicone putty. The peak-to-peak equivalent sound pressure level (pespl) was higher (P < 0Æ001) when the ear canal was sealed (range of mean differences was 8Æ3 24Æ9 db pespl). This means that the signal to noise ratio can be improved by sealing the meatus because the electronic noise level is not increased. Most important is that the dynamic range of the clicking sounds was 62 db that is larger than the effective dynamic range of a 16 bits sound card. Future studies are needed to establish normative pespl values. However, cards with at least 24 bits A/D conversion will be required, especially in patients with suspected disc displacement with reduction, where the difference in loudness between opening and closing clicking often is large. KEYWORDS: temporomandibular joint, clicking, sound pressure level, signal processing, dynamic range, disc displacement, signal to noise ratio Introduction Electronic recording of temporomandibular joint (TMJ) sounds has the potential to be a valuable diagnostic aid because it is possible to apply objective signal processing methods to such recordings. Amplitude is one of the parameters of interest when analysing TMJ sounds (Muhl, Sadowsky & Sakols, 1987; Barghi, dos Santos & Narendran, 1992; Wabeke et al., 1992). The recordings are often made with microphones placed in the ear canals *Sound pressure level (SPL) is an absolute value measured in db representing the physical intensity of a sound; 20 times the log of the pressure output over the pressure reference (0Æ0002 dyne cm )2 or 20 lpa). Peak equivalent SPL (pespl); equal to the amplitude of a 1000-Hz tone as if it were equivalent to the peak of a transient signal such as a click; used to determine the intensity level of a click (Mendel, Danhauer & Singh, 1999). The term peak-to-peak equivalent SPL can also be used and is synonymous to peak equivalent SPL. For definitions see also International Standard. Audiometers Part 3: Auditory test signals of short duration for audiometric and neuro-otological purposes. Reference number CEI IEC 645 3, pp 1 18, (Gay & Bertolami, 1988; Widmalm, Williams & Adams, 1996; Prinz & Ng, 1996; Leader et al., 1999). In some studies the orifice was completely occluded by the silicone putty used to seat the microphone (Sano et al., 1999; Djurdjanovic et al., 2000). Sealing the meatus can be expected to increase the sound pressure level (SPL)* and thus also the amplitude of the recorded signal. It is of clinical interest to be able to make repeated standardized measurements to evaluate the effect of treatment on the intensity of a patient s TMJ sounds. In this context it is important to know if the effect of sealing is significant but that aspect seems not to have been discussed in TMJ sound research. The aim was to test the hypothesis, that sealing the orifice of the auditory meatus when placing a recording microphone increases the peak equivalent sound pressure level (pespl) of the TMJ clicking. The method used was comparing the pespl of bilaterally recorded TMJ clickings between sessions, with and without complete sealing of the ear canals. ª 2003 Blackwell Publishing Ltd 495

2 496 S. E. WIDMALM et al. Methods Bilateral electronic TMJ sound recordings (n ¼ 249) were made from three adult subjects (two females, one male) who had non-painful reciprocal clicking. The subjects were fully informed of the experiments and agreed to take part. Small electret condenser microphones (diameter 5 mm), bandwidth Hz, were placed at the openings a few millimetre inside the external auditory canals and either held in place using Silicone putty (Fig. 1a and b) or manually by an assistant. Recordings were digitized using a 24 bits sound card with preamplifiers that had a frequency response at )3 db: 1Æ6Hz)200 khz. Most clickings can be recorded both on the side of origin and on the contra lateral side, often with the same polarity and similar amplitude. We differed between the original clicking and the contra laterally recorded sound (here called the ÔechoÕ) by comparing the time locations. The echo appears about 1 ms, or less, later in time (Widmalm, Williams & McKay, 2002). The sampling rate was 48 or 96 khz. The higher rate was needed to localize the clicking side when the time delay was about 0Æ2 ms or less. A sound level meter (SLM) was used to measure the level of external noise. The recordings were made in a quite office. With the computer (PC) on, the equivalent noise level was about 55 db pespl (C-weighting) as measured 1 m from the PC. With the PC off the level was about 50 db pespl. Amplitude calibration was achieved by recording a 94-dB pespl (RMS), 1000 Hz, signal (Fig. 2) with each microphone at the start of each session and again if gain settings were changed. Recordings were made in four sessions from each subject. Recordings were made with and without the ear canals sealed (Fig. 1). Session 1. Left side microphone placed using silicone putty.** Right side microphone held in place by an assistant. Sony ECM-77B, Sony Corporation, , Kita-Shinagawa, Shinagawa-Ku, Tokyo 141, Japan. Direct Pro Aardvark, Ann Arbor, MI, USA. Radio Shack Sound Level Meter, Tandy Corporation, Fort Worth, TX, USA. Sound Level Calibrator Type 4231, Brüel & Kjær, A S, Nærum, Denmark. **Coltoflax, Art. no Coltène, Altstätten, Switzerland. Fig. 1. These figures show how the microphones were wrapped in silicone putty before insertion (a) and the location of the microphone in the earplug (b) after removal from the ear. Fig. 2. Two TMJ sound recordings are displayed together with the calibration signals in separate windows to allow optimal gain. Horizontal axes are in samples (96 samples ¼ 1 ms). Vertical axes are in parts of the A D card s dynamic range (1 ¼ maximal positive and )1 ¼ maximal negative values). Upper window: Clicking recording plus inset of a 94-dB pespl (RMS), 1000 Hz calibration signal. The pespl of the clicking is 42 db larger than that of the calibration signal. Lower window: Low level clicking (74 db pespl) followed by a 94-dB pespl calibration signal. The difference in pespl between the clickings in the two windows is 62 db. This means that the upper clicking would have a peak-to-peak amplitude more than 1000 times larger than the clicking in the lower window if both clickings were displayed with the same gain.

3 TMJ SOUNDS DYNAMIC RANGE 497 Table 1 TMJ Sound pressure levels TMJ sound Recording condition Ear canal n pespl (db) mean s.d. Difference sealed v. open (db) mean s.e. P Subject no. 1. Clicking Sealed Æ3 3Æ37 Clicking during opening Open Æ4 3Æ26 24Æ9 0Æ78 <0Æ001 Echo Sealed Æ3 2Æ29 Open 37 99Æ6 4Æ80 24Æ7 0Æ88 <0Æ001 Subject no. 2. Clicking Sealed Æ3 1Æ45 Clicking during opening Open Æ8 5Æ37 16Æ5 0Æ83 <0Æ001 Echo Sealed Æ8 4Æ25 Open Æ7 3Æ87 21Æ1 0Æ88 <0Æ001 Subject no. 3. Clicking Sealed Æ4 2Æ83 Clicking during opening Open 23 95Æ5 11Æ00 23Æ9 2Æ37 <0Æ001 Echo Sealed Æ0 2Æ52 Open 26 86Æ1 5Æ15 20Æ9 1Æ16 <0Æ001 Subject no. 3. Clicking Sealed 21 95Æ2 2Æ39 Clicking during closing Open 25 84Æ5 6Æ50 10Æ7 1Æ40 <0Æ001 Echo Sealed 20 91Æ4 2Æ38 Open 26 83Æ1 4Æ70 8Æ3 1Æ06 <0Æ001 This table gives the peak equivalent sound pressure levels (pespl) measured in db from recordings with versus without sealing the ear canal where the microphone was placed. Bilateral (two-channel) TMJ sound recordings were made using two microphones, one in each ear canal. ÔOpenÕ means here that the microphone was held by hand inside the auditory meatus at the orifice of the ear canal without using any sealing putty. Clicking ¼ sound recorded on the side of the clicking joint. Echo ¼ sound recorded on the side contralateral to the clicking joint. n ¼ number of sound recordings. s.d. ¼ standard deviation. s.e. ¼ standard error. Session 2. Both microphones placed using silicone putty. Session 3. Right side microphone placed using silicone putty. Left side microphone held in place by an assistant. Session 4. Both microphones placed without silicone putty and held in place by an assistant. Bilateral TMJ sound recordings were made while the subjects performed wide jaw opening-closing movements at a rate of 1 1Æ5 cycles per second. They were asked to perform about cycles if that was possible without feeling pain or discomfort. The peak equivalent sound pressure levels (pespl) were measured in db by comparing with the calibration signals and compared between sides within each opening-closing cycle. We were guided by listening to audio play back when searching the sound recording graphs for waveforms of clicking sounds. The recordings were analysed using Cool Edit Pro, Matlab software and application programs written in Matlab. The amplitudes of the clicking sounds during opening in the three subjects plus the clicking during closing from subject no. 3 were compared. Clickings were included only if they were always well above the noise level and could be reliably localized to side. The amplitudes were compared between sides (clicking side and the contra lateral ÔechoÕ side) using Paired Samples T-test and between sessions (with and without sealing of the meati) using an ANOVA General Linear Model for Repeated Measures. The null hypothesis that the samples came from populations with the same variance was tested using the Levene test (Norušis, 2000). Results (details are given in Table 1) Sealing the ear canal increased the peak equivalent sound pressure level (pespl) of the recorded TMJ clicking sounds. The range of mean differences between having the canals sealed instead of open was 8Æ3 24Æ9 db. The differences were significant (P < 0Æ001) both in recordings from the clicking side and in Syntrillium Software Corporation, Phoenix, AZ, USA. The MathWorks Inc., Natick, MA, USA. SPSS V10Æ0, SPSS, Chicago, IL, USA.

4 498 S. E. WIDMALM et al. recordings from the side contra lateral to the clicking joint. No significant effect was observed on the pespl of the sounds recorded on one side by sealing the contra lateral ear canal. The range of the individual clicking amplitude values was 62 db ( db pespl). Discussion The results of this study showed that sealing the ear canals improved the signal to noise ratio by increasing the TMJ sound s pressure level (SPL) when recording with a microphone placed in the ear canal. The SPL values were higher (P <0Æ001) when the opening of the ear canal was occluded. This means that complete occlusion of the ear canal is preferable when standardizing the method for recording because the sealing does not increase the level of electronic noise. Weak sounds, otherwise buried in the noise, such as faint closing clicking in a patient having disc displacement with reduction (DDR), may thus be amplified to raise above the noise level and become visible in the graphical display from an electronic recording by careful sealing of the ear canal. When asking a patient to describe the TMJ sounds it also makes a difference if he she occludes the auditory meatus by pushing the tragus against the ear canal. However, sealing the ear canal may also increase the low frequency internal noise from the patient, such as blood flow in the close temporal artery. Further studies are needed to test if electronic filtering applied before the digitization is needed to solve such problems. Another important result of this study was that a 16 bits A D card does not have a large enough dynamic range to always allow simultaneous recording of both the high and the low level TMJ sounds in a patient. The theoretically maximal resolutions in 12 and 16 bits A D conversion are 72 and 96 db but the effective dynamic range is much lower because of the noise level of the recording system, which may be anywhere from 30 to 50 db depending on the quality of the electronic components, the internal noise from the patient, and the external noise in the room where the recordings are made. Most commercial microphones have an internal equivalent noise level of about 30 db pespl and the external noise can hardly be < 20 db pespl even if a silent proof room (anechoic chamber) is used. Clinical TMJ sound recordings are usually made in a dental office where the equivalent noise level can be expected to be at least about 50 db pespl (C-weighting). There has also to be a safety zone of about 10 db above the joint sounds peak levels to avoid clipped recordings. The cards dynamic range has therefore to be about 10 db higher than the sum of the noise and the dynamic range of the TMJ sounds. Clicking during opening and closing (reciprocal clicking) is considered to be an important sign in diagnoses of disc function (Eriksson, Westesson & Rohlin, 1985). The findings in this study indicate that many previous reports about absence of closing clicking in electronic recordings from subjects with DDR may have been misleading. That is the sound cards used may have had too small a dynamic range to always record both the opening and the closing clicking sounds. A difference of 60 db is equivalent to 1000:1 in a linear plot. Huge differences in loudness between audible TMJ sounds may be a significant source of error when evaluating linear plots. That is weak sounds are difficult to observe and distinguish from the baseline if plotted together with loud sounds as illustrated in Fig. 3. Linear plots thus make it difficult to identify both high and low intensity sounds in the same graph. Unfortunately the waveforms cannot be plotted on a logarithmic scale because they are stored with both negative and positive values and it is not possible to take the log of a negative value. The absolute values of the voltage levels can be plotted on a logarithmic scale but such pictures would look very unfamiliar to most examiners. When searching the plot for low-level sounds such waveforms are easier to find if the examiner is guided by listening to an audio playback of the recording. The db values are ratios obtained by comparing with a calibration signal. They may not reflect true intensity levels. If they do, the reason that the high level clickings observed in this study were tolerable is most probably that the clicking sounds had a very short duration, usually only 2 10 ms. A-weighting is most common when measuring environmental noise. It gives lower noise levels than C-weighting because the signals are adjusted by attenuating the low frequency parts to approximate the loudness as perceived by the human ear. As well known the human hearing sensitivity is down about )60 db at 20 Hz as compared with its sensitivity for frequencies between 1000 and 2000 Hz. We used, however, C-weighting to get a fair estimate of the noise level in the electronic recordings because the microphones had a frequency response that was flat between

5 TMJ SOUNDS DYNAMIC RANGE 499 Fig. 3. This figure shows the differences in peak-to-peak amplitudes between seven copies of one and the same TMJ sound when they are displayed with decreasing gain. The difference is 10 db between each consecutive copy. Our results show that TMJ sounds may differ more than 60 db in pespl, which is the difference between the first sound copy in Section 1 and the copy in Section 7. The waveforms in sections 5 6 are barely visible and the one in Section 7 is not possible to distinguish from the baseline. This illustrates that linear plots of recordings of waveforms that differ more than about 30 db are difficult to identify when selecting TMJ sounds for spectral analysis from linear plots. Horizontal axes are in samples (48 samples ¼ 1 ms). Vertical axes are in parts of the A D card s dynamic range (1 ¼ maximal positive and )1 ¼ maximal negative values). 150 and 2000 Hz, down )3 db at 40 Hz and only about )7 db at 20 Hz. We did not use audiotape recorders where, as pointed out by Christensen (1992), frequencies <1 khz and >5 khz are often amplified to compensate for the reduced sensitivity of the human ear. Such equipment-dependent differences in frequency response have to be considered when comparing the audio characteristics of TMJ sounds at auscultation with those at playback of electronic recordings. Comparisons are further complicated by the fact that human ears too have noise levels that may differ between individuals and be age-dependent. A good young human ear has for instance an apparent noise level equivalent to that of a microphone with an A-weighted noise level of about 20 db pespl (Killion, 1976). The difference in pespl between a clicking and its contralateral ÔechoÕ was often too small to be observed by auscultation even if it had been possible to listen to both sides simultaneously. As illustrated in Fig. 4 several clickings may occur very close in time and Fig. 4. This figure shows in the upper window a TMJ sound recording, using the sampling rate (SR) Hz, from five consecutive opening closing jaw movements with loud clicking occurring during each cycle. In the lower window the file was down sampled, using the Matlab function decimate, to 1000 Hz to illustrate the loss of information (reduced amplitudes) when using too low SR. Note that the clicking at no. 1 has disappeared in the lower window from the site indicated by arrow 1. The waveform no. 4 in the upper window shows several clickings close in time but some are lost in the corresponding area in the lower window (arrow cli). Horizontal axes are in samples. Vertical axes are in parts of the A D card s dynamic range (1 ¼ maximal positive and )1 ¼ maximal negative values). they may not all come from the same side. Even if they do it is well known that amplitudes vary during consecutive opening-closing cycles. It is therefore often impossible to decide about side of origin by listening with a stethoscope first on one side and then on the other. Electronic recording on both sides is the only reliable method for deciding in which TMJ the clicking occurs. It is, however, important to note that comparisons of TMJ sound amplitudes cannot be based on recordings with such low sampling rates as Hz, often used in diagnostic TMJ sound recording (Ishigaki, Bessette & Maruyama, 1993; Olivieri et al., 1999; Garcia et al., 2000). The reason is that the amplitudes are very much affected when high frequency components are lost as illustrated in Fig. 4. In conclusion the method for placing the microphones for TMJ sound recording has to be a controlled factor. Sealing the ear canal improves the signal to noise ratio by significantly increasing the TMJ sound pressure level, and thus the amplitude of the recorded signal,

6 500 S. E. WIDMALM et al. without raising the electronic noise level. This is of special importance in patients with suspected DDR, where the difference in loudness between opening and closing clicking often is large. The mean pespl values we found are not suggested to be normative. Further studies are required to establish such values but soundcards with at least 24 bits A D conversion are required because the dynamic range of the patient s TMJ sounds may be larger than that of a 16 bits A D card s effective dynamic range. Acknowledgments The authors thank Drs Elisabeth Hultcrantz and Stig Arlinger for reading the manuscript and giving valuable advice. References BARGHI, N., DOS SANTOS, Jr J.& NARENDRAN, S. (1992) Effects of posterior teeth replacement on temporomandibular joint sounds: a preliminary report. Journal of Prosthetic Dentistry, 68, 132. CHRISTENSEN, L.V. (1992) Physics and the sounds produced by the temporoman-dibular joints. Part I. Journal of Oral Rehabilitation, 19, 471. DJURDJANOVIC, D., WIDMALM, S.E., KOH, C.K., WILLIAMS, W.J. & YANG, K.P. (2000) Computerized classification of temporomandibular joint sounds. IEEE Transactions on Biomedical Engineering, 47, 977. ERIKSSON, L., WESTESSON, P.L. & ROHLIN, M. (1985) Temporomandibular joint sounds in patients with disc displacement. International Journal of Oral Surgery, 14, 428. GARCIA, A.R., MADEIRA, M.C., PAIVA, G.& OLIVIERI, A.N. (2000) Joint vibration analysis in patients with articular inflammation. Cranio, 18, 272. GAY, T.& BERTOLAMI, C.N. (1988) The acoustical characteristics of the normal temporomandibular joint. Journal of Dental Research, 67, 56. ISHIGAKI, S., BESSETTE, R.W. & MARUYAMA, T. (1993) A clinical study of temporo-mandibular joint vibrations in TMJ dysfunction patients. Journal of Craniomandibular Practice, 11, 7. KILLION, M.C. (1976) Noise of ears and microphones. Journal of the Acoustical Society of America, 59, 424. LEADER, J.K., BOSTON, J.R., RUDY, T.E., GRECO, C.M. & ZAKI, H.S. (1999) The influence of mandibular movements on joint sounds in patients with temporomandibular disorders. Journal of Prosthetic Dentistry, 81, 186. MENDEL, L.L., DANHAUER, J.L. & SINGH, S. (1999) Singular s Illustrated Dictionary of Audiology, pp. 199, 224, 239. Singular, San Diego, CA. MUHL, Z.F., SADOWSKY, C. & SAKOLS, E.I. (1987) Timing of temporomandibular joint sounds in orthodontic patients. Journal of Dental Research, 66, NORUšIS, M.J. (2000) Guide to Data Analysis, pp Prentice Hall, New Jersey. OLIVIERI, K.A.N., GARCIA, A.R., PAIVA, G.& STEVENS, C. (1999) Joint vibration analysis in asymptomatic volunteers and symptomatic patients. Cranio, 17, 176. PRINZ, J.F. & NG, K.W. (1996) Characterization of sounds emanating from the human temporo-mandibular joints. Archives of Oral Biology, 41, 631. SANO, T., WIDMALM, S.E., WESTESSON, P.L., TAKAHASHI, K., YOSHIDA, H., MICHI, K. & OKANO, T. (1999) Amplitude and frequency spectrum of TMJ sounds from subjects with and without other TMD signs symptoms. Journal of Oral Rehabilitation, 26, 145. WABEKE, K.B., SPRUIJT, R.J., WEYDEN, K.J. & NAEIJE, M. (1992) Evaluation of a technique for recording temporomandibular joint sounds. Journal of Prosthetic Dentistry, 68, 676. WIDMALM, S.E., WILLIAMS, W.J. & ADAMS, B.S. (1996) The wave forms of temporomandibular joint sounds clicking and crepitation. Journal of Oral Rehabilitation, 23, 44. WIDMALM, S.E., WILLIAMS, W.J. & MCKAY, D. (2001) Localization of TMJ sounds to side. Journal of Oral Rehabilitation, 29, 911. Correspondence: Sven E Widmalm, 1565 Kuehnle, Ann Arbor, MI 48103, USA. sew@umich.edu