Acoustic analysis of the closing sounds of bileaflet prosthetic valves in a sheep model

J

THORAC CARDIOVASC SURG 1991;101:1060-8

Acoustic analysis of the closing sounds of bileaflet prosthetic valves in a sheep model Previous investigations of mechanical valve sounds have shown that (1) frequency spectra of sounds produced by abnormal valves differ from those of properly functioning valves and (2) leaflets of normally functioning bileaflet valves do not close synchronously. This investigation studied effects of mechanical valve size, environment, and hemodynamic state on closing sounds. A single 25 mm bileaftet mitral mechanical valve was implanted in six sheep and a single 27 mm valve in four sheep. With digital signal processing, asynchronous leaflet closure and frequency spectra were assessed after alterations in animal position, respiratory phase, heart rate, afterload, contractility, and preload. Both asynchronous leaflet closure and frequency spectra varied among animals, and, except for a decrease in asynchrony with increasing contractility, were largely independent of valve size and hemodynamics. Baseline asynchrony ranged from 0.5 to 4.2 msec. Frequency spectra were characterized by the first three resonant peaks. Lowest resonant peaks ranged from 2.5 to 3.4 kHz, middle from 4.7 to 6.8 kHz, and highest from 7.2 to 9.6 kHz. These results indicate that accurate assessment of mechanical valve function with acoustic analysis requires baseline studies in aU patients by means of a system with a frequency response of more than 10 kHz.

Richard L. Donnerstein, MD,a William A. Scott, MD,a Andre Vasu, MD,b and Jack G. Copeland, MD,b Tucson, Ariz.

Mechanical valves constitute the majority of the more than 100,000 prosthetic heart valves implanted worldwide each year. Because long-term outlook for patients after valve replacement strongly depends on valve-related factors, early recognition of prosthetic valve dysfunction is critical.I-? Detecting abnormal function can be difficult because even a properly functioning mechanical valve may cause turbulent flow and have some degree of stenosis and regurgitation.': 3-6 Moreover, significant valve problems may not cause noticeable hemodynamic changes until late in the course, or, in the worst case, until sudden catastrophic failure occurs. Although echocardiography, cinefluoroscopy, and stanFrom the Department of Pediatrics (Cardiology) and Children's Research Center,' Department of Surgery (Cardia- Thoracic)," University of Arizona, Health Sciences Center, Tucson, Ariz. Supported by a grant-in-aid award from the American Heart Association, Arizona affiliate. Received for publication Nov. 21, 1989. Accepted for publication June 18, 1990. Address for reprints: Richard L. Donnerstein, MD, Department of Pediatrics (Cardiology), Arizona Health Sciences Center, 1501 N. Campbell Ave., Tucson, AZ 85724.

12/1/24523

1060

dard phonocardiography'"? can confirm advanced valve dysfunction, none of these techniques has been shown to reliably detect early valve degeneration. Analysis of sounds produced by mechanical valves has been used to identify changes associated with valve degeneration. 11-18 In theory, any significant alteration in valve properties (calcification, tissue ingrowth, clot formation, cracking, tearing, or sticking occluding elements) should cause changes in opening or closing sounds. Therefore analysis of these sounds should provide information concerning valve integrity. In a previous clinical study we analyzed closing sounds produced by normally functioning mechanical bileaflet valves and demonstrated that leaflets of these valves do not close synchronously.19,20 Other studies have shown that frequency spectra of sounds produced by malfunctioning prosthetic valves differ from spectra of properly functioning valves. II - 18 We theorized that sounds created by prosthetic heart valves are influenced both by cardiac hemodynamics and by intrinsic mechanical and acoustic characteristics of the valve. Certain properties of these sounds might vary with hemodynamic state while other qualities might remain stable and could, therefore, be used to characterize the valve and possibly detect early valve

Volume 101 Number 6

I 06 I

Closing sounds of bileaflet prosthetic valves

June 1991

LEAFLET IMPACTS

RELATIVE TIME

RELATIVE TIME

Fig. 1. Computer-generated examples of sounds created by hypothetic impacts. A, Schematic representation of a single impact of a valve leaflet on the valve cage. B, Theoretic sound created by a single impact is modeled as a damped sine wave. C, Example of two widely spaced impacts of different magnitudes. D, Sound created by impacts shown in C. E, Example of two closely spaced impacts of different magnitudes. F, Sound resulting from impacts shown in E. (See text for details.)

Table I. Baseline values of asynchrony and frequency peaks Experiment No.

Valve size (mm)

1 2 3 4 5 6 7 8 9 10

27 27

25 25 25 25 25 27

25 27

Time between leaflet impacts (msec) Mean ± SD

2.9 ± 0.5 ± 3.2 ± 4.2 ± 2.6 ± 2.0 ± 1.1 ±

1.2 ± 2.1 ± 2.4 ±

0.6 0.1 0.1 0.3 0.4 0.5 0.2 0.3 0.3 0.3

Range

2.00.4 3.03.92.3 -

3.7 0.6 3.4 4.5 3.0 1.5 - 2.5 0.9 - 1.3 1.0- 1.8 1.8 - 2.5 2.0- 2.8

Resonant frequency peaks (Hz) Fl

F2

F3

2728 2272 3352 3096 2992 3424 2448 3000 2944 2536

5816 4480 6928 6280 6480 6760 3872 5264 5350 5128

8240 7312 9864 9704 9496 9664 6776 8048 7528 7192

FI, Frequency of lowest spectral resonant peak; F2, frequency of second spectral resonant peak; F3, frequency of third spectral resonant peak.

degeneration. To test this possibility we used digital signal processing techniques to study intracardiac and externally recorded sounds produced by prostheticbileafletmitralvalves implantedina sheepmodel. Experiments were designed to determine how asynchronous leaflet closure and frequency spectra of valve soundsvary with valve characteristics, valve environment, and hemodynamic state. Methods The experimental protocol was approved by the Institutional Laboratory Animal Care Committee. All animals received humane care in compliance with the "Principles of Laboratory Animal Care" formulated by the National Society for Medical Research and the "Guide for the Care and Use of Laboratory

Animals" prepared by the National Academy of Sciences and published by the National Institutes of Health (NIH Publication No. 80-23, revised 1978). Experimental preparation. Under halothane anesthesia, sheep weighing 30 to 50 kg were intubated and their lungs ventilated with a volume ventilator. Fluid-filled catheters were placed in a femoral artery and vein for pressure monitoring and fluid and drug administration. The pericardium was opened through a left thoracotomy, and the animal was placed on cardiopulmonary bypass. The left atrium was opened after cardiac arrest with cold cardioplegic solution. The native mitral valve was removed and replaced with one of two (25 mm or 27 mm) St. Jude Medical prosthetic bileaflet valves, Previously calibrated, high-fidelity, catheter-tip pressure transducers (Millar Instruments, Inc., Houston, Tex.) were placed in the left ventricle via an apical incisionand into the left atrium via the left atrial appendage. Catheters were positioned to detect valve sounds of

The Journalot Thoracic and Cardiovascular

1 0 6 2 Donnerstein et al.

Surgery

DIGITAL PHONOCARDIOGRAM A

Me

Me

Me

"

I

VALVE SOUND

125 msec/div

IMPULSE RESPONSE

B

C

1~1Jv-'-~----------1 0.5 msec/div

0.5 msec/dlv

FREQUENCY SPECTRUM

LEAFLET IMPACTS

E

D

II

II

I I

~

I

I

I. II

J,

0.5 msec/div

1000 Hz/div

Fig. 2. Example of digital deconvolution analysis. A, High-fidelity digital phonocardiogram showing three valve closing sounds. B, First closing sound with horizontal axis expanded by a factor of 70. C, Theoretic acoustic response (impulse response) to a single leaflet impact. D, Leaflet impacts needed to create closing sound shown in B. E, Frequency spectrum of impulse response. Me, Mitral valve closing sound.

maximum intensity. Catheters and right atrial pacing wires were brought to the surface through skin incisions. After the surgical procedure cardiopulmonary bypass was discontinued and the chest wall was closed. After the animal was hemodynamically stable, the experimental protocol was performed. Animals' lungs were mechanically ventilated throughout the experiment. After the experiment the prosthetic valve was removed for reimplantation in another animal. Data acquisition. High-fidelity pressure signals from the intracardiac transducers were electronically processed to provide the rate of left ventricular pressure development (dP/dt) and acoustic signals. Because frequencies contained in sounds produced by mechanical valves are much higher than frequencies required to define hemodynamic pressures accurately, intracardiac valve sounds were obtained by passing left atrial and left ventricular high-fidelity pressure signals through 50 Hz high-pass filters. An external phonocardiogram was obtained by means of an accelerometer microphone placed near the apex of the heart. Internal and external valve sounds were recorded simultaneously with a high-fidelity, dual-channel tape recorder (Tandberg, model 12-41, Oslo, Norway) at a tape speed of7V2 in/sec. The frequency response of the tape recorder was calibrated and shown to be flat within 3 dB from 50 to 15,000 Hz. The catheter-tip pressure transducers and accelerometer microphone had flat frequency responses to more than 10,000

Hz. The electrocardiogram, left atrial and left ventricular pressures, left ventricular dP / dt, and acoustic signal were recorded on a multichannel chart recorder at a paper speed of 100 mm/ sec. Central venous and aortic pressures were recorded from fluid-filled catheters. Experimental protocol. The following parameters were altered to test their effects on sounds produced by the prosthetic valve: 1. Position: Measurements were made with the animal supine and in both left and right decubitus positions. 2. Respiratory phase: Measurements were made during held inspiration and held expiration. 3. Heart rate: After baseline recordings were obtained, the right atrial pacing wire was used to increase heart rate, in increments of 20 beats/min, to a maximum of 180 beats/min. 4. After/oad: After an intravenous atropine bolus of 0.02 mg/kg, afterload was altered with an intravenous infusion of phenylephrine used as a systemic vasoconstrictor. The phenylephrine infusion was titrated to increase systolic blood pressure, in increments of 20 torr, to a maximum of 60 torr above baseline. 5. Contractility: A continuous infusion of dobutamine was used to alter contractility. The initial infusion rate was 2 ~g/kg/min and was increased to 5, 10, and 20 ~g/kg/min at la-minute intervals.

Volume 101 Number 6

Closing sounds ofbileaflet prosthetic valves

June 1991

1 06 3

DIGITAL PHONOCARDIOGRAM A

Me

Me

Me

125 msec/div

VALVE SOUNDS B

D

C

0.5 msec/div

0.5 msec/div

0.5 msec/div

LEAFLET IMPACTS

E

I ['

I I

G

F

..

II

0.5 msec/div

I

0.5 msec/div

1 0.5 msec/div

Fig•. 3. Example of digital deconvolution analysis applied to three consecutive valve closing sounds. A, High-fidelity digital phonocardiogram. B to D, Closing sounds with horizontal scales expanded by a factor of 55. E to G, Leaflet impacts needed to create closing sounds. Me, Mitral valve closing sound.

6. Preload: Preload was increased with normal saline given as three rapid infusions of 10 ml/kg at 5-minute intervals. Data analysis and signal processing. Heart rate, pressures, and left ventricular dP / dt were measured from chart recordings. Recorded closing valve sounds were analyzed with a DSP-IOO digital phonocardiographic amplifier and software system (International Acoustics, Inc., Palatine, Ill.) interfaced with an IBM-compatible (Compaq) computer (Compaq Computer Corp., Houston, Tex.). This system has an analog-to-digital conversion rate of 8000 Hz with antialiasing filters set at 3500 Hz. The effective frequency response of the system was increased by reducing playback tape speed to l'Vs in/sec, thereby allowing unambiguous frequency analysis of the original signal up to 14,000 Hz (recording tape speed was 7 1/ 2 in/sec). Mechanical valve closing sounds were examined with a signal processing technique known as digital deconvolution analysis. This method is related to spectral analysis techniques such as the maximum entropy method and linear predictive coding that have been used extensively for evaluating other types of waveforrns.21-26 Our application of digital deconvolution analysis is based on the assumption that sounds created by a closing valve result from multiple discrete impacts of valve leaflets on the valve cage. 19 , 20 Each impact is considered to be instantaneous and cause an acoustic response that is independent of timing, but dependent on impact force and characteristics of valve leaflets, cage, and environment. The signal processing technique may be likened to analyzing sounds created by hitting a cymbal with a brush. In theory, if the acoustic response of a singlebrush hair hitting the cymbal were known, then the entire

history of individual bristles hitting the surface could be determined from the composite signal. Application of digital deconvolution analysis to three hypothetic computer-generated valve sounds is shown in Fig. 1. In these schematic examples, relative timing and amplitudes of leaflet impacts are shown in the left panels, with resulting valve sounds depicted on the right. It is assumed that a single impact (Fig. 1, A) will produce a sound modeled by the damped sine wave shown in Fig. 1, B. Therefore sounds created by multiple leaflet impacts would be composed of multiple damped sine waves, with timing and magnitudes determined by the timing and magnitudes of the impacts. If a significant time interval separates impacts of different amplitudes (Fig. 1, C), the sound in Fig. 1, D, would result. In this simplified example inspection of the sound waveform suggests that it was created by two damped sine waves (identical to that in Fig. 1, B) with amplitudes and timing as shown in Fig. 1, C. Fig. 1, E, shows a more complex example in which a second impact occurs shortly after the first. The resulting sound waveform (Fig. 1, F) represents the algebraic sum of the individually generated damped sine waves. In each case digital deconvolution analysis would demonstrate that sounds in the right panels were created by impacts shown in the left panels. Additionally, the analysis would establish that individual impacts result in a damped sine wave, as shown in Fig. 1, B, and would determine the frequency spectrum of this sound. Statistical analysis. The time interval between leaflet impacts and locations of the first three peaks of the frequency spectrum was determined for each valve closing sound. For each change in hemodynamic state, data from three to five valve

The Journal of Thoracic and Cardiovascular Surgery

Donnerstein et al.

1064

Table II. Effects of hemodynamic state on asynchronous leaflet closure Time intervalbetween leaflet impacts (msec) Position

Heart rate

Respiration

Afterload

Contractility

Preload

Exp.

a

b

c

a

b

a

b

a

b

a

b

a

b

3 4 5 6 7 8 9 10

3.6 2.6 1.6 2.0 l.l 1.2 3.1 2.3

3.2 4.2 2.6 2.0 l.l 1.2 2.1 2.4

3.0 3.7 1.3 l.l 1.8 1.6 2.5 1.9

5.5 3.0 1.3 1.0 2.2 2.3 1.9 1.7

4.3 4.7 1.6 1.2 1.7 1.0 2.4 2.4

3.2 4.3 1.3 1.2 1.0 1.6 2.2 2.4

3.0 3.2 1.9 2.0 0.9 l.l 3.0 2.6

3.7 3.1

3.9 l.l

1.7

1.4

3.7 4.5 2.0 1.5 1.3 1.0 3.0 2.4

2.4 1.0 0.8 0.8 I.2 0.9 2.2 1.7

2.9 5.7 1.3 1.6 l.l l.l 3.5 1.7

2.6 3.9 2.6 1.3 1.0 2.6 1.2 0.9

Exp., Experiment number; position: a, right decubitus. b, left decubitus, c, supine; respiration: a, inspiration, b, expiration; heart rate: a, baseline rate, b, maximum rate; afterload: a, baseline mean arterial pressure, b, maximum mean arterial pressure; contractility: a, baseline left ventricular dP jdt, b, maximum left ventricular dP jdt; preload: a, baseline mean central venous pressure, b, maximum mean central venous pressure.

Table III. Average resonant frequency peaks Experiment No. 3 4 5

6 7 8 9 10

F1 (Hz) 3304 3086 2964 3430 2508 3128 3161 2559

± 187 ± 203 ± 119 . ± 189 ± 158 ± 81 ± 242 ± 71

F2 (Hz) 6450 6421 6006 6769 4461 5601 6004 4803

± 470 ± 636 ± 410 ± 366 ± 525 ± 279 ± 508 ± 194

F3 (Hz)

9626 9290 9330 9565 7227 8225 8634 7455

± 618 ± 767 ± 299

± 196 ± 622 ± 275 ± 787 ± 231

Values are mean ± standard deviation. FI, Frequency oflowest spectral resonant peak; F2, frequency of second spectral resonant peak; F3, frequency of third spectral resonant peak.

closing sounds were averaged and used for further statistical analysis. Effects of valve size on the time interval between leaflet impacts and frequency spectra of valve sounds were assessed by unpaired t tests. Analysis of variance was used to assess effects of other hemodynamic alterations on asynchronous closure and frequency spectra of valve sounds.

Results Prosthetic St. Jude Medical bileaflet valves were successfully implanted in the mitral position in 10 sheep. The 25 mm valve was implanted in six sheep and the 27 mm valve in four. Two sheep died during the early experimental protocol, and only baseline studies were obtained from these animals. The influence of hemodynamic changes on sounds created by closing valves was assessed by means of the intracardiac recordings. An example of digital deconvolution analysis applied to the analysis of an implanted bileaflet valve is shown in Fig, 2. The high-fidelity digital phonocardiogram (Fig. 2, A) is dominated by the three intense sounds produced by the closing prosthetic mitral valve. Fig. 2, B, shows the first closing sound with the horizontal axis expanded by a factor of 70, Fig. 2, C, shows the theoretic acoustic response to an individual impact, the impulse response, as deter-

mined by this analysis. With this response for an individual impact, the composite closing sound (Fig, 2, B) may be shown to have been created by the series of leaflet impacts shown in Fig. 2, D. The two clusters of impacts most likely result from individual valve leaflets striking the valve cage at different times. In this case the time interval between impact clusters is approximately 2.2 msec. Smaller deflections within each cluster may be due either to small bounces of the leaflet on the valve cage or to the fact that not all of the leaflet surface contacts the cage simultaneously. The frequency spectrum of the impulse response (Fig. 2, E) has resonant peaks at approximately 2500, 4800, 7400, and 9800 Hz, Results of hemodynamic studies are summarized in Tables I to III. Time interval between leaflet impacts. As shown in Fig. 3, asynchronous leaflet closure varied somewhat from beat to beat. In this example the time interval between consecutive leaflet impacts ranged from 2.2 msec (Fig. 3, G) to 3.0 msec (Fig. 3, E). Baseline studies. Asynchrony immediately after valve replacement ranged from a minimum of 0.4 msec in study 2 to a maximum of 4.5 msec in study 4 (Table I). As shown in Fig. 4, the time interval between leaflet impacts varied among animals in which the same valve was implanted and did not significantly depend on valve size (p = 0.29). Hemodynamic variation ofasynchrony. Table II summarizes effects of hemodynamic changes on the degree of asynchronous leaflet closure. Because only baseline data were obtained from experiments 1 and 2, results from these experiments were excluded from further analysis. The time difference between leaflet impacts did not change significantly with animal position (p =0.66), phase of respiration (p = 0.89), or changes in heart rates ranging from approximately 115 to 180 beats/min (p = 0.80). Most animals did not tolerate an acute increase in

Volume 101

Closing soundsof bileaflet prosthetic valves 1 06 5

Number 6 June 1991

afterloadin the early postoperative period,and afterload was varied independently in only three animals. These animals showed no consistent variation in asynchrony with changes in mean blood pressure (Table 11). Blood pressure, however, also varied spontaneously throughout the experimental protocol. Spontaneous mean blood pressures in each animal increased an average of 105% (range 46%to 151%) during the experiments. No significant relationship was found between asynchrony and spontaneous changes in blood pressure. Contractility, as determined by left ventricular dP/ dt, increased an averageof 154%(range 11 % to 300%) during dobutamine infusions. As summarized in Table II, asynchrony decreased with increasing contractility (p < 0.03). However, the rate of change varied from animal to animal. Most of the decrease in asynchrony occurred with moderate increases in contractility, with lesspredictablechangesas contractilityincreasedfurther. Central venous pressureincreasedan averageof 308% (range 167% to 600%) after salinebolusinfusions. Asynchronydid not change significantly in response to changes in preload (p = 0.48). Frequency spectra Baselinespectra. Frequencyspectra immediatelyafter valve implantationdifferedfor eachanimal (TableI). The lowest resonant peak ranged from 2272 to 3424 Hz, the middlefrom 3872 to 6928 Hz, and the third from 6776 to 9864 Hz. As shownin Fig. 5, frequencyspectra varied widely among animals in which the same valve was implanted. Although resonant peaks tended to be of lowerfrequencies in the 27 mm valvethan in the 25 mm valve, this difference was not statistically significant (p > 0.05). Hemodynamic variation ofspectra. Only baselinedata wereobtained from experiments I and 2, and these data were excluded from further analysis. Table III summarizes the variation in spectral peaks for experiments 3 to 10.The samevalve implantedin differentsheepproduced different spectral patterns that did not change significantlywithanimal position, respiratoryphase, heart rate, afterload, contractility, or preload. Internally and externally recorded valve sounds. Closing valve sounds recorded simultaneously from the intracardiac catheter and external microphone demonstrated essentially identicaltime intervalsbetweenleaflet impacts (Fig. 6). Fig. 6, A and B, shows the simultaneously recordedinternal and external closing sounds. In each case impact clusters are separated by approximately 1.7msec (Fig. 6, C and D). As shownin Fig. 6, E and F, however, the frequency spectra of the sounds differ. Although resonant peaks of internal and external signals occurat similarfrequencies, the higherfrequencies of the externally recorded sound were markedly attenuated.

.

'0 5 - , - - - - - - - - - - - - - , - - - - - - - - , 25 mm Valve


27 mm Valve

.§. w 4 II:

::;)

III

g3 U III

52 Z

o

II: :1:1

U

Z

>

III c(

0

3

4

5

6

7

9

1

2

8

10

EXPERIMENT NUMBER

Fig. 4. Time interval between leaflet impacts determined at start of experimental protocol. Error bars = 1 standard deviation. 12000..-------------,--------,

27 mm Valve

25 mm Valve

N

10000

:I:

> u

8000

•

ffi

6000

II:

4000

::;)

oW

lL

F1

• F2 E3 F3

2000

3

4

5

6

7

9

1

2

8

10

EXPERIMENT NUMBER

Fig. 5. Frequency spectra of valve closing sounds determined at start of experimental protocol. F 1, Frequency of lowest spectral resonant peak; F2, frequency of second spectral resonant peak; F3, frequency of third spectral resonant peak. Error bars = 1 standard deviation.

Discussion

This invivo study has demonstrated that asynchronous leaflet closure and frequency spectra of normally functioning bileaflet prosthetic valves differ widely among animals. Except for a decrease in asynchrony with increasing contractility, we did not detect significant changes in either asynchrony or frequency spectra with changes in valve size, phase of respiration, animal position, or hemodynamics. Although the limited size of our study may have preventeddetectionof small statistically significant changes,it is unlikelythat these wouldbe clinicallysignificant under normal physiologic conditions. As shown in Table I and Figs. 4 and 5, asynchronyand frequency spectra did not change in a predictable manner after repeated reimplantation of the same prosthetic valve. Thereforedifferences in asynchronyand frequency

The Journal of Thoracic and Cardiovascular

I 0 6 6 Donnerstein et al.

Surgery

VALVE SOUNDS

r-----------------, r-----------------, B

A

1000 Hzldiv

INTRACARDIAC

1000 Hz/dlv

BODY SURFACE

Fig. 6. Example of simultaneous intracardiac and body surface recordings of valve sounds. A and 8, Individual closing sounds. C and D, Leaflet impacts. E and F, Frequency spectra of closing sounds.

spectra are probably not caused by changes in valves resulting from multiple reimplantations. In sheep, and presumablyin patients, frequencycontent and asynchronous leaflet closure appear to be characteristics of the valveas implanted in each individual and are minimally altered by hemodynamicchanges within clinical ranges. Although defining frequency spectra and leaflet closure patterns of mechanicalvalves may be usefulfor detecting early valvedegeneration,serial studieswillberequired in each patient becauseof relatively widevariationsin valve closure sounds among individuals. Asynchronous closure. Prosthetic valve leaflets close when captured by reversing blood flow and impact the valve cage with a force related to leaflet inertia and the energy of this reversing stream. Asynchronous leafletclosure is expected under normal circumstancesand results from minor differences in forces applied to individual leaflets. Demonstrationof asynchronyin a bileaflet valve shows that both leaflets are freely moving structures. Slight variation in valve orientation changes flow patternst-? and probably accounts for individual differences in asynchronous closure. Beat-to-beat changes in direction or energy of the reversing bloodstream will

of

influence the amount of asynchrony and force individual impacts. Decreases in asynchrony noted with greater levels of contractilityare probablydue to increased energy imparted to valve leaflets during closure. Marked asynchronous closure of bileaflet valves has previously been described in circumstances such as atrial fibrillation."? However, we have demonstrated that a lesser degree of asynchronous closure is a normal finding of properly functioning bileaflet mechanical heart valves and that the amount of asynchrony remains stable over time.19,20 Although further studies are needed, we can speculate that valve dysfunction will cause changes in patterns of leaflet closure. Frequency spectra. Analysis of frequency spectra of sounds produced by prosthetic valves has been proposed as a method for assessing valvefunction. I 1-18,23-28 Dominant frequencies of closing sounds producedby bioprosthetic heart valves are abnormally elevated in the presence of calcification or fibrosis. 23- 28 However, frequency spectra and failure modes of mechanical valves differ from those of bioprosthetic valves. Multipleinvestigators have shown a decrease in high-frequency content of soundsassociatedwithmechanicalvalvedysfunction. I 1-18

Volume 101 Number 6 June 1991

We found that frequency spectra of closing sounds produced by the same valve varied among animals, which suggests that factors in addition to the valve itself may determine spectral characteristics.16 Most previous studies of mechanical valvesounds limited their frequency range to less than 5000 HZI I-20 and some to less than 1000 Hz. 13, 16 81. Jude Medical bileaflet valves evaluated in our study had significantfrequencycomponentsgreater than 9000 Hz, a findingconsistent with studies showing similar high frequencies in sounds produced by other pyrolytic carbon valves. I I, 29 Significant information may be lost if spectra are defined only at frequencies less than 9000 Hz. Intracardiac and externally recorded sounds. Externallyrecordedvalvesoundsreliably predicted the amount of asynchrony when compared with intracardiac sounds. However, high-frequency contents of external sounds were much lessthan those recorded internally. In clinical studies we and others II, 29 have found significant energy in frequencies greater than 10,000 Hz in prosthetic valve sounds recorded from the body surface. The marked reduction of high frequencies found in our externally recordedsounds was likelydue to poor acoustic coupling of the accelerometer microphone to the sheep's skin. Signal processing. Although standard phonocardiographic techniques of evaluating sound intensity and timingare not sensitive to early changes in valvefunction, frequencycontent of sounds produced by mechanical and bioprosthetic valves is affected by valve degeneration.I I- 18, 23, 24, 27, 28 Computer spectral analysis of prosthetic valvesoundswas initially performed with fast Fourier transforms.P: 27-29 However, frequency resolution of fast Fourier transforms is directly related to sampling duration, which can lead to uncertainty when trying to define spectral peaks of very short duration valve sounds.23- 26, 30 Moreover, sounds produced by artificial valves are affected by many factors other than properties of the valve itself, and a great deal of information is lost if onlyspectra of compositesounds are analyzed. The signal processing method used for digital deconvolution analysis can improve approximations of true spectra of short duration soundssuch as those produced by mechanicalvalves. 16, 22-26 After spectra are defined,valvesounds are searched for discontinuities that suggest individual impacts.By separating these factors, deconvolution analysis can provide both specific information about valve integrityand insight into the dynamic functioningof normal and abnormal mechanical valves. In conclusion, asynchronousleafletclosure is a normal finding of properly functioning bileaflet mechanical valves. The time interval between leaflet closure and sound frequency spectra of bileaflet prosthetic valves

Closing sounds ofbileaflet prostheticvalves

10 6 7

depend on valve environment and, therefore, cannot be predicted by valve characteristics and hemodynamics alone. Acoustic analysisof sounds created by mechanical valves appears to be a promising means for detection of certain modes of valve failure. However, results of this investigation suggest that optimal use of this technique requires baselinestudies in all patients shortly after valve implantation by means of a system with a frequency response greater than 10,000 Hz. REFERENCES 1. Schoen FJ. Cardiac valve prostheses: pathological and bioengineering considerations. J Cardiac Surg 1987;2:65108. 2. Deuvaert FE, Le Clerc JL, Primo G, et al. Thrombosis of the Saint Jude Medical valve prosthesis in the aortic position: a diagnostic and surgical emergency. J Cardiovasc Surg 1986;27:622-34. 3. Yoganathan AP,Chaux A,GrayRJ, et al. Bileaflet, tilting disc and porcine valve substitutes: in vitro hydrodynamic characteristics. J Am Coli Cardiol 1984;3:313-20. 4. Chandran KB. Pulsatile flow pastSt. Jude Medical bileaflet valve: an in vitro study. J THORAC CARDIOVASC SURG 1985;89:743-9. 5. Gibbs JL, Wharton GA,Williams GJ. Doppler ultrasound ofnormally functioning mechanical mitral andaortic valve prostheses. Int J CardioI1988;18:391-8. 6. Jones M, Eidbo EE.Doppler color flow evaluation of prosthetic mitral valves: experimental epicardial studies. J Am Coli Cardiol 1989;13:234-40. 7. Chaux A, GrayRJ, Matloff JM, Feldman H, Sustaita H. An appreciation of the new St. Judevalvular prosthesis. J THORAC CARDIOVASC SURG 1981;81 :202-11. 8. Panidis IP, Ren J-F, Kotler MN, et al. Clinical and echocardiographic evaluation of theSt. Judecardiac valve prosthesis: follow-up of 126 patients. J Am Coli Cardiol 1984;4:454-62. 9. Feldman HJ, GrayRJ, Chaux A,et al.Noninvasive invivo and in vitro study of the St. Jude mitralvalve prosthesis: evaluation using two-dimensional and M-mode echocardiography, phonocardiography, andcinefluoroscopy. AmJ Cardiol 1982;49: 1101-9. 10. Kotler MN, Mintz GS,Panidis I, Morganroth J, Segal BL, Ross J. Noninvasive evaluation of normal and abnormal prosthetic valve function. J Am Coli Cardiol 1983;2:15173. 11. Gordon RF, Najmi M, Kingsley B,Segal BL, LinhartJW. Spectroanalytic evaluation ofaortic prosthetic valves. Chest 1974;66:44-9. 12. Beaudet RL, Schwartz L, StarekPJK,SohnY, Craige E, Colson MA. Spectral phonocardiographic assessment of mechanical prosthetic heartvalves [Abstract]. Circulation 1984;70(Pt 2):11263. 13. Suobank DW, Yoganathan AP, Harrison EC, Corcoran WHo Aquantitative method for theinvitro study ofsounds

10 6 8 Donnerstein et al.

14.

15.

16. 17.

18.

19.

20.

21.

22.

produced by prosthetic aortic heart valves. Parts I-III. Med BioI Eng Comput 1984;22:32-54. Hylen JC, Kloster FE, Herr RH, Starr A, Griswold HE. Sound spectrographic diagnosis of aortic ball variance. Circulation 1969;39:849-58. Kagawa Y, Nitta S, Satoh N, et al. Sound spectroanalytic diagnosis of malfunctioning prosthetic heart valve. Tohoku J Exp Med 1977;123:77-89. Koymen H, Altay BK, Ider YZ. A study of prosthetic heart sounds. IEEE Trans Biomed Eng 1987;34:853-63. Kagawa Y, Sato N, Nitta S, et al. Real-time sound spectroanalysis for diagnosis of malfunctioning prosthetic valves. J THORAC CARDIOVASC SURG 1980;79:671-9. Sato N, Mohri H, Kagawa Y, Horiuchi T, Tanaka M, Nitta S. Real-time sound spectral analysis for diagnosis of thrombosed prosthetic valves. ASAIO Trans 1988;34: 831-4. Donnerstein RL, Kien GA, Katz HV, Allen HD. Digital phonocardiographic evaluation of mechanical prosthetic heart valves:characterization of the bi-Ieaflet prosthesis. In: Stead W, ed. Proceedings of the Eleventh Annual Symposium on Computer Applications in Medical Care. New York: Institute of Electrical and Electronic Engineers Computer Society Press, 1987:575-9. Donnerstein RL, Allen H D. Asynchronous leaflet closure in the normally functioning bileaflet mechanical valve. Am Heart J 1990;119:694-7. Iwata A, Suzumura N, Ikegaya K. Pattern classification of the phonocardiogram using linear prediction analysis. Med Bioi Eng Comput 1977;15:407-12. Press WH, Flannery BP, Teukolsky SA, Vetterling WT. Numerical recipes: the art of scientific computing. Cambridge: Cambridge University Press, 1986:430-5.

The Journal of Thoracic and Cardiovascular Surgery

23. Foale RA, Joo TH, McClennan JH, et al. Detection of aortic porcine valve dysfunction by maximum entropy spectral analysis. Circulation 1983;68:42-9. 24. Joo TH, McClellan JH, Foale RA, Myers GS, Lees RS. Pole-zero modeling and classification of phonocardiograms. IEEE Trans Biomed Eng 1983;30:110-8. 25. Durand LG, De Guise J, Cloutier G, Guardo R, Brais M. Evaluation of FFT-based and modern parametric methods for the spectral analysis ofbioprosthetic valvesounds. IEEE Trans Biomed Eng 1986;33:572-8. 26. Cloutier G, Guardo R, Durand LG. Spectral analysis of closing sounds produced by lonescu-Shiley bioprosthetic aortic heart valves. Part 3. Performance of FFT-based and parametric methods for extracting diagnostic spectral parameters. Med BioI Eng Comput 1987;25:497-503. 27. Stein PD, Sabbah HN, Lakier JB, Kemp SR, Magilligan DJ. Frequency spectra of the first heart sound and of the aortic component of the second heart sound in patients with degenerated porcine bioprosthetic valves. Am J Cardiol 1984;53:557-61. 28. Stein PO, Sabbah HN, Lakier JB, Goldstein S. Frequency spectrum of the aortic component of the second heart sound in patients with normal valves, aortic stenosis and degenerated porcine xenografts. Am J Cardiol 1980;46:48-52. 29. Schondube F, Keusen H, Messmer BJ. Physical analysis of the Bjork-Shiley prosthetic valve sound. J THORAC CARDIOVASC SURG 1983;86:136-41. 30. Yoganathan AP, Gupta R, Corcoran WHo Fast Fourier transform in the analysis of biomedical data. Med Bioi Eng 1976;14:239-44.