IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 39,

NO.

3. MARCH 1992

307

Communications A New Heart-Sounds Gating Device for Medical Imaging Mark W. Groch, James R. Domnanovich, and William D. Erwin Abstract-A heart-sounds gating device has been designed and tested which identifies, individually, both the first (S1) and second ( S 2 ) heart sound from their timing relationship, providing two trigger points through the cardiac cycle for synchronizing medical images. The new heart-sounds gate utilizes dynamically varying timing windows to anticipate the occurrence of S 1 and 52. The heart-sounds gate has been initially applied to nuclear imaging of the cardiac bloodpool, but may be applied to any imaging modality requiring cardiac synchronization.

INTRODUCTION

microphone. The phonocardiogram is conditioned to aid in sound identification and trigger generation, as illustrated in Fig. 1. After the heart sounds are amplified and bandpass filtered (0-2 kHz cutoff, followed by 0-500 Hz cutoff), an absolute value amplifier folds the negative portions of the waveform onto the positive voltage axis, and the S 1 and S2 signals are enveloped. A squaring circuit is used to minimize noise, and the peak-signal amplitude drives an automatic gain circuit. Finally, the gating point for each heart sound is set by a variable threshold, using a comparator circuit to generate a trigger pulse when the threshold level, set to 30% of the peak value, is exceeded. A standard ECG isolation amplifier and gate are also included in the system for ECG reference or for use as a conventional R-wave trigger.

Digital Logic

It is important to accurately assess cardiac function serially in patients with coronary artery disease for treatment and follow-up [ I]-[3]. Nuclear medicine methods provide ventricular function noninvasively, using multigated blood pool imaging (GBP), where scintigraphic data from the heart is acquired in up to 32 sequentially digitized frames through the cardiac cycle [4], [5]. Typically, the R wave of the electrocardiogram (ECG) provides the trigger point within the cardiac cycle to synchronize computer acquisition. Limitations of GBP acquisition using the ECG R-to-R interval as the triggering cycle, lie in poor reproduction of the diastolic filling portion of the ventricular volume curve, and in inadequate triggering in patients with severe arrhythmias and those with pacemakers. Newer ECG gating methods, such as backward and forward-backward gating, have not overcome these limitations [6]-[9]. Heart sounds can provide an alternate physiologic parameter to the ECG for cardiac synchronization, and moreover, provide two reference points within the cardiac cycle, to uniquely define ventricular systolic ejection and diastolic filling. Heart sounds alone have not been heretofore used as a single or as multiple sources of cardiac triggers due to problems in distinguishing the first (S 1) and second (S2) heart-sounds components. We have developed a microprocessor-controlled heart-sounds gate (HSG), which automatically identifies the first and second heart sound from the phonocardiogram alone, using their timing relationship [lo]. We report the details of this new heart-sounds gate. DESCRIPTION

Analog Circuitry A phonocardiogram is obtained from a commercially available Hewlett-Packard system using a small, 2.5 cm diam, piezoelectric Manuscript received October 22, 1990; revised September 6, 1991. M. W. Groch and W. D. Erwin are with the Departments of Medical Physics and Diagnostic Radiology/NuclearMedicine, Rush Graduate College, Rush Presbyterian-St. Luke’s Medical Center, Chicago, IL 60612. J. R. Domnanovich is with Siemens Gammasonics Inc., Hoffman Estates, IL 60195. IEEE Log Number 9105599.

Using the heart-sounds trigger pulses, the time intervals between the S1 to S2 pulses and the S2 to S1 pulses are determined. The S 1 and S2 pulses are distinguished using their timing relationships with the assumption that, at least initially, the S1 to S2 interval will be shorter than the S2 to S 1 interval. The logic is illustrated in Fig. 2. After each heart-sound trigger, a lock-out pulse is generated which inhibits any further triggers for its duration. The lockout serves to eliminate double triggers, or false triggers due to murmurs or the third heart sound ( S 3 ) . Concurrently with the lock-out window, a timing, or “listening” window is opened for a duration only long enough for the S2 sound to occur, if the instigating pulse was an S 1 pulse. A second-trigger pulse is only allowed if it occurs within the timing window and after the lock-out interval. The timing window and lock-out window are continuously variable and are controlled by the digital logic monitoring the time intervals (Fig. 3). In general, the timing window varies between 200 and 400 ms, and is initially set to be one-half the longest sound-to-sound interval heard. After the initial setting, the S 1 to S2 listening window is subsequently varied using the value obtained by the equation for the estimated time between the R and T waves of the ECG: S 1 4 2 Interval (ms)

=

455 - 1.878 X HR

+ 0.0026 x HR2

(1)

where HR is the patient heart rate [l 11. The heart rate is obtained from the S 1-S 1 interval or, if needed, from the electrocardiogram. When the HSG is first applied, if the instigating pulse is S1, the S 2 sound will occur after the lock-out and within the listening window. If the instigating pulse happens to be S 2 then, since the S2 to S1 interval is longer, the listening window will have expired before the next pulse is seen and will not be accepted as a valid S 1 trigger. The next pulse will be an S 1, which will reinstigate the listening window, catching the following S2. Thus, within two beats, the circuit will identify and lock on to the proper heart sounds in sequence. A 16-bit ROM controls the window lengths for the lock-out and timing windows. The digital logic circuitry will vary the lock-out and listening windows to track increasing or decreasing heart rates. The digital circuit block is illustrated in Fig. 3. An LED display allows for monitoring of the heart rate, timing intervals, and of the trigger pulses.

0018-9294/92$03.00 0 1992 IEEE

IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 39,

308

ANALOG

NO. 3, MARCH

1992

BLOCK

H/S

H/S TRIG.

H/S

H/S

I

'

Fold

I

I

I

Fig. 1 . Analog portion of the heart sounds gate where trigger pulses are generated from significant sounds in the phonocardiogram.

for these 19 patients using commercially available software (Siemens MicroDELTA'" Rev 6.1 software). HS Trig L O TM WND

Is T @Trig

ECG

h J--;J--'i l---;J---j A---------I - - - - --: - L

-

r

i

fl

g

n

r

/ 1 -

Fig. 2. The logic utilized to identify the first (S 1) and second (S2) heart sounds from the phonocardiogram. Heart sound (HS) triggers (Trig) are identified as either first (1st Trig), second (2nd Trig) or extraneous by use of lock-out and listening time windows (TM WND). The lock-out window serves to eliminate murmurs or noise after each HS. The TM WND is set to the estimated S 1 to S 2 interval by ROM look-up table.

PATIENTAPPLICATIONS The patient phonocardiogram is obtained from the microphone, coupled with gel to the patient's chest, at the location of the loudest second heart sound, since S2 is typically softer than S 1, placing the microphone at the location of the maximal second heart sound usually provides near equal amplitudes of S 1 and S2. We found that the algorithm used to identify S 1 and S2 from the phonocardiogram performed well at patient heart rates of up to 200 beats per minute. Left-ventricular ejection fractions (LVEF's) were computed and ventricular volume curves were generated from GBP studies, using three ( S l , S2, R wave) trigger modes, in 19 patients referred to nuclear medicine for normally prescribed GBP imaging, who could be adequately gated by both ECG and heart sounds methods. The LVEF's from a) ECG gating with this device, b) S 1 gating and c) S2 gating were compared to the clinically reported LVEF values

RESULTS

The range of LVEF's computed from the 19 patients was 1776%. The LVEF computed using the ECG trigger mode of this device agreed well with the reported LVEF using a standard commercial system, with a correlation coefficient of r = 0.94, and a linear regression equation of y = 1.Oh - 1.21

(2)

with the conventional method represented as x. When the first heartsound gated LVEF was compared to the conventional method, an excellent correlation of r = 0.96 was achieved, with a linear regression of y = 0 . 9 3 ~- 1.31.

(3)

The correlation of S2 gated LVEF was r = 0.96 with a linear regression of

+

y = 0 . 8 0 ~ 6.38

(4)

indicating good performance of the device in patients who could be gated by both heart sounds and standard gated acquisition methods. The ventricular volume curves obtained using ECG R wave and S1 triggers appear similar in form, beginning with end diastole. The volume curve initiated with an S 2 trigger is, of course, displayed beginning at end-systole. The volume curves from the a) ECG gated and b) S2 gated studies for one typical patient are illustrated in Fig. 4. Ventricular volume curves obtained by the three gating modes illustrated, qualitatively, superior definition of the diastolic filling phase for both the first heart-sound gated study, and in particular, the second heart-sound gated study, compared to the ECG gated study, as illustrated in Fig. 4. DISCUSSION Accurate reproduction of the time course of ventricular contraction and relaxation is essential to the evaluation of cardiac ejection

309

IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING. VOL. 39. NO. 3, MARCH 1992

OIGITAL BLOCK R D TRIG. R TRIG.

I

1

++

lst HIS PULSE

SHAPER

*

DIGITALLY VARIABLE FILTER (LOCKOUT)

*

DIGITALLY

,

INPUT SELECT

VARIABLE

I

SYSTEM

WINDOW

CONTROL FILE

Fig. 3 . Digital block of the heart sounds gate.

Fig. 4. Ventricular volume curves and background curves (below volume curve) obtained from typical patient comparing the (a) ECG R-wave triggered volume curve (left) to the curve obtained using the (b) second heart sound (S2) (right), as the initiating trigger. The S2 gated ventricular volume curve displays a pattern shifted from the conventional volume curve display. S2 triggered curve results in superior definition of the diastolic filling phase, with atrial “kick” clearly discernible.

and, particularly, filling indexes. The need to estimate the R-to-R interval to synchronize gated image frames causes “time jitter,” distorting the ventricular volume curve near the end of the gating cycle, when the R-to-R interval varies from the initial estimate. With HSG, two reference points in the cardiac cycle are available to potentially reduce the effects of “time jitter.’’ It is reasonable to expect that S 2 initiation of the multigated acquisition sequence would result in superior reproduction of the filling phases of the cardiac cycle, as the “time jitter’’ normally associated with the end of the ventricular volume curve occurs during systolic ejection. First heart-sound gating should reproduce the systolic ejection phase at least as well as ECG R-wave gating. In fact, we noted an improvement, over ECG R-wave gating, in the definition of the ejection phase of the volume curve when S 1 was used for the initiation of the multigating sequence. Furthermore, HSG may per-

form better than ECG gating in rapidly changing heart rates, as S 1 and S 2 can define the beginning and end of systole without need for setting an acceptance window for the R-to-R interval. We also noted a slight improvement in the filling phase, as well, when S 1 triggering initiated cycle acquisition. This may be due to the documented 5-40 ms time delay between the R wave and the presence of the S 1 heart sound. This corresponds to well-known differences between electrical stimulation and the onset of mechanical contraction of the heart [12], [13]. Limitations of heart-sounds gating include patients with faint heart sounds, murmurs, clicks, and the presence of third (S3) and fourth (S4)heart sounds; excessive room noise, and other body noises. However, the digital logic employed with this new generation heart sounds processor effectively reduced problems due to S 3 and S4 as well as murmurs and clicks, by use of lock-out and

IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 39. NO. 3. MARCH 1992

310

listening windows and expectation logic for the heart-sounds timing intervals. CONCLUSION Using the HSG, multigated blood-pool imaging, or any other modality requiring cardiac synchronization, can be performed using a diastolic trigger point (SI),a systolic trigger point ( S 2 ) , or a diastolic trigger with a “midcourse correction,” S 2 trigger. Leftventricular ejection fractions correlated well in patients who could be adequately gated by both methods. Ventricular volume curves obtained with HSG appear qualitatively to be superior in definition to standard, single-point, ECG R-wave gated volume curves. Further clinical verification of diastolic filling parameters is required. ACKNOWLEDGMENT The authors would like to thank S. A. Trhlik, V. Sabeti, D. J. Schippers, and C. M. Groch for reference and secretarial assistance in the preparation of this manuscript. REFERENCES [I] R. 0. Bonow, S. L. Bacharach, M. V. Green et a l . , “Impaired left ventricular diastolic filling in patients with coronary artery disease: Assessment with radionuclide angiography,” Circulation, vol. 64, pp. 315-323, 1981. [2] G. B. J. Mancini, R. A. Slutsky, S. L. Norris et a l . , “Radionuclide analysis of peak filling rate, filling fraction, and time to peak filling rate: Response to supine bicycle exercise in normal subjects and patients with coronary disease,” Amer. J Cardiol, vol. 51, pp. 43-5 l , 1983. [3] L. R. Poliner, S. H. Farber, D. H. Glaeser et a l . , “Alteration of diastolic filling rate during exercise radionuclide angiography: A highly sensitive technique for detection of coronary artery disease,” Circulation, vol. 70, pp. 942-950, 1984. [4] H. W. Straws, B. L. Zaret, P. J. Hurley et a l . , “A scintiphotographic method for measuring left ventricular ejection fraction in man without cardiac catheterization,” Amer. J Cardiol, vol. 2 8 , pp. 575580, 1971. [5] D. S. Berman, A. F. Sadel, G. L. DeNardo et a l . , “Clinical assessment of left ventricular regional contraction patterns and ejection fraction by high-resolution gated scintigraphy,” J . Nucl. Med., vol. 16, pp. 865-874, 1975. [6] P. H. Murphy, “ECG gating: Does it adequately monitor ventricular contraction,” J . Nucl. Med., vol. 21, pp. 399-401, 1980. [7] S. L. Bacharach, M. V. Green, J. S . Borer et a l . , “Beat-by-beat validation of ECG gating,” J . Nucl. Med., vol. 21. pp. 307-313. 1980. [8] J. E. Juni, J. Froelich, J. McMeekin et a l . , “Effects of heart rate variability on scintigraphic measurement of diastolic function,” Circulation, vol. 68, p. 111-24 (abstract), 1983. [9] C. Chen and J. E. Juni, “Measurement of left ventricular diastolic function: Effect of gating modes,” J . Nucl. Med., vol. 27, p. 935 (abstract), 1986. [IO] M. W. Groch and J. R. Domnanovich, “Heart sound detector and synchronization for diagnostics,” U . S . Patent No. 4546777, Oct. 15, 1985. [ I I] R. A. Wolthuis, A. Hopkirk, N. Keiser, andJ. R. Fischer, “T-Waves in the exercise ECG: Their location and occurrence.” IEEE Trans. Bio-med. Eng., vol. 11, pp. 639-643, 1979. R . H. Wagner, J. R. Halama, R. E. Henkin et U / . . “Errors in the determination of left ventricular functional parameters, J . Nucl. Med., vol. 30, pp. 1870-1874, 1989. D. A. Turner, P. L. Von Behren, N. T . Ruggie, R. G. Hauser. P. Denes, A. Ah, J. V. Messer, E. W. Fordham, and M. W. Groch, “Noninvasive identification of initial site of abnormal ventricular activation by least-squares analysis of radionuclide cineangiograms,” Circulation, vol. 65, pp. 1511-1518, 1982.

Use of Impedance Ratio for the Continuous Measurement of Stroke Volume of a Valveless Pouch Used as a Cardiac-Assist Device L. A. Geddes, W. Janas, and S. F. Badylak Abstract-A technique is described for measuring the volume of a valveless compressible plastic pouch and its volume change when used as a cardiac-assist device. The method employs measuring the pouch impedance at high frequency with sleeve electrodes at both ends of the pouch. The use of an adequately high frequency eliminates the electrode impedance and the impedance measured is that of the resistance of the electrolyte in the pouch. By equating the compressible pouch to two truncated cones with their bases adjacent, an equation is derived that relates pouch impedance to volume. It is shown that by plotting the stroke volume ejected (AV) versus the ratio of systolic (R,) to diastolic (R