Mechanical Correlates of the Third Heart Sound

DONALD D. GLOWER, MD, ROBERT L- MURRAH, MD . CRAIG O . OLSEN, MD,JAMES W. DAVIS, MSEE, J . SCOTT RANFJN f, D

Durham, North Carolina

In seven chronically instrumented conscious dogs, micrnmaoma •eters measured left ventricular pressure, and erllrssonic dmensi totransducers measured left ventricular minor-axis diameter, tinelatter recording was filtered to examine data between 20 and100 Hz. Acceptable external heart sensors were recorded with _aphonocardiographic microphone in four of the seven dogs. Witheach dog sedated, incubated and mechanically ventilated, datawere obtained during hemodynamic alterations produced byvolume loading, phenylepbrine, calcium infusion and you avalocclusion.

Damped oscillations were noted consistently in the left ventric .nine diameter waveform toward the end of rapid ventricularfilling. These wall vibrations, assessed by the filtered diameter,

Since the appearance of Charles Potain's cur=ie essay in

1900 (1) describing the third heart sound (S3 ), scientificcontroversy has existed over its biologic mechanisms andthe associated hemodynamic and pathologic implications .Three principal theories of origin attribute S 7 to 1) valvularand subvalvular phenomena (2-5), Z) me impact of the heartagainst the chest wall (6-8), and 3) ventricular wall vibra.tions (9-16) . Most investigators. (9-16) now accept ventric-ular wall vibrations as the origin of S, The following studioswere designed to investigate the mechanical basis of S 3 in anintact animal model aad to examine the physiologic factors

that modify the characteristics of S 3 , Specifically, the ques-tions to be examined were I) does the energy transmitted tothe ventricular wall during rapid filling contribute to theamplitude of S3 , and 2) does the frequency content of S,behave according to set physical principles governing theoscillation of viscoelastic materials .

From the Depanmena of Surgery and Biomedical Engmeenng, DukeUniversity Medical Center, Durham, North Carolina . This study was sup-ported in pan by Grants HL-09315 and HL-29436 and SCOR Grant HL-17670fmm the National Heart, Lung, end Blood Institute . National Institutes ofHealth, Bethesaa, Maryland . Dr . Glower is nraipieat of ILe Job. F . Gibbon Jr.Research scholarship of The Amcncan Association lot Thoracic Surgery,Manchcstcr, Macvaehusens .

Manuscript manned May I. 1991 ; raised manuscript received Jury 3,1901, ii-pied July 24, 1991.

Address for reprints ; Donald D. Glower, MD, ilex 3893, Duke UniversityMedical Center, Durham, North Carolina 27710 .

01992 by the Amcd-., College of Cardiology

JAOC V.I. 1% No 2February 1992 :451-7

correlated well with the third heart sound (S3) on the phinqocar.diagram . The peak frequency of the wall vibrations increased withincreased diastolic pressure (p = 0.004), probably reflecting anincrease in myocardial wall stiQaess. In contrast, the amplitude ofthe vibrations varied directly with left ventricular filling rate (p =0 .0001) .

Thus, S,, seemed to be related specifically to ventricular wallvibrations during rapid filling, and the spectra of the amplimde-frequency relation shifted toward the audible range with increasesin diastolic pressure, wall stiffness or filling rate. Spectral analysisof S, may be useful in assessing pathologic changes in myawrdialwall properties.

IJAm Coif Cmdiol 1992;19:450-7J

MethodsExperimental preparation . Seven healthy adult mongrel

dogs (18 to 24 kg) were surgically instrumented for subs,-quent studies in the closed chest state . Each dog wasanesthetized with intravenous thiamylal sodium (25-mg/kgbody weight) and ventilated with a volume respirator. A leftthoracotomy, was perfoa:ted through the fifth intereosistspace by sterile technique, and epic rdiai ultrasonic dimen-sion transducers (17) were positioned across the left ventric-ular anterior-posterior minor axis to measure left ventriculardiameter and veatricularwall oscillations (Fig . 1) . During theprocedure, silicone rubber pneumatic occluders were posi-tioned around tha inferiorvena cava and descending thoracicaorta . A silicone rubber tube was implanted in the base ofthe left atrium, and a second rubber tube was left in thepleural space for later passage of microe^.anemaefs . lirepericardium was left widely open, the transducer leads andsilicone tubes were exteriorized through subcutaneous tun-nels dorsal to the thoracotomy, and the incision was repairedin multiple layers. All dogs received daily intramuscularinjections of procaine penicillin-G (2 x 105 U) and dihydro-streptomycin (250 mg), along with a single injection of ferrichydroxide (100 mg) .

At the time of study, the dimension transducers werecoupled directly to a modified, dual output sonomicrometerthat sampled data at a rate of 1,000 samples/s. One channelinterposed a two-pole filter to produce a 12-db/octave atten-uation above 100 Hz yielding an effective frequency re-sponse of 0 to 50 Hz . This channel was used to monitor gross

0775-1197/92/55 .00

IACC Vol. 19, No. 2February 1992: 4 50-7

Figure 1, The standard preparation employed in this study withultrasonic -ansducers pos tioned across the rifle-axis diameter cithe left ventricle and infetior versa cave and descending thoracicaorta pneumatic occluders . AA solid state pressure transducer waspassed into the left vrntrh-!e through the implamed left atria) tube . Aand P = antunor and posterior left-. :a: .-:cr.:tar dimension transducers, respectively .

cardiac dimensions with a resolution of approximately0.05 nun. The second channel interposed a 5-Hz high passfilter to remove the gross minor-axis diameter waveform,followed by two sequential bandpass filters to select frequen-cies between 20 and 100 Hz (Fig . 2). All filters were theButterworth type to minimize ringing, and a minimal number0f poles was used to provide a sufficient slope of attenuationoutside the desired bandpass (18).

Left ventricular and pleural pressure measurements wereobtained ,.,:a. n:iceomanometers )PC-350. Millar Instru-

Figure 2. Digital examples of the minor-axis diameter signal afterfiltering with the 50-Hz low pass filter (middle panel) and with the 20-to t03-Hz bandpass filter (lower panel) . The top panel shows thederivative of minor-axis diameter (dDld) used to determine therapid filling phase (see text) .

40

-401

Mina 11-K .-

VV'

GLOWER ET AL.

451THIRD HEART SOUND

ments) coupled to pressure amplifiers (8805-C, Hewlett-Packard) . Manometers were passed into the mid-left ventri-cle and pleural space to yield left ventricular transmittalpressure as the difference between left ventricular andpleural pressures. In four of the seven dogs external phono-catdiograms of acceptable quality were obtained on thechest wall overlying the ventricular apex with use of astandard atn!ilification system and an Elema-Scholandermicrophone (EMT-25B),

physiologic protocol. After recovery from implantation (7to 10 days), each dog was anesthetized (intravenous mor-phine sulfate, 0.5 to 1 mg/kg), intubated and ventilated at aminute volume of 6 liters, The ventilator was disconnected,and data were recorded during brief periods of apnea withthe animal paralyzed and the endotracheal tube open to air .Apnea eliminated respiratory variations in filling rate andpressure, standardized the physiologic data from beat to beatand minimized extraneous breath sounds . Bradycaraia wasmaintained by administration of propranolol (1I to 5 mg) topermit analysis of a sufficient number of digital data pointsduring rapid diastolic filling to achieve 2- to 3-Hz resolutionwith a Fourier analysis program developed in our iaboratory .All analog data were recorded on magnetic tape (model A,Vetter) for subsequent digital analysis . Filling rate wasassessed by the derivative of minor-axis diameter (dD/dt)(19-22), and mean idling rate was defined as the average rateof change of minor-axis diameter during the rapid fillingphase of diastole.

To facilitate Fourier analysis of third heart sound data,the rapid filling phase of diastole was defined as the periodfrom peak positive dD/dt to 100 ms after the first subsequentzero crossing of dD/dt (Fig. 2) . From each study, segmentsof steady state data were selected during baseline condi-tions, then during steady state control conditions of in-creased left ventricular end-diastolic pressure achieved bysimultaneous blood volume expansion, infusion of phenyle-phrine (251aglkg per min intravenously), and partial descend-ing thoracic aortic occlusion . These interventions routinelyachieved left ventricular end-diastolic oressu ;es o 20 to40 c-aa Hg. Next, with tie volume expansion, phenylephrineinfusion and partial aortic occlusion still present, data wereacquired during partial inferior versa ceva Occlusion toreduce ventricular diastolic pressure and diameter . Finally,data were recorded after a calcium infusion (15 nt /kgintravenously) that increased cardiac o . .put and filling ratewhile minimally changing other variables . All studies wereconducted in accordance with the Position of the AmericanHeart Association on Research Animal Use .

Data analysis . Relevant data were analyzed for six car-diac cycles at each steady state, and the filtered dimensionsignals were averaged with use of a Fourier analysis programmodified for this purpose. The resulting spectral analysisyielded calculated mean frequencies and mean amolitud&, ofwall vibrations that were representative of the events occur-ring during a given rapid diastolic filling period for eachhemodynamic steady state. Multivariate linear regression

432

GLOWER LT AL."!'HIRD HEART SOUND

35

J EE n

t- - 3 3 -- -4---t--

`~----

4

oN

IsetJ

Figure 3. Representative raw digital rerotdings of left ventriculardiameter (LVD). phonocardiogram (Phono), filtered diameter andleft ventricular pressure (LVP) obtained during blood volume ex-pansion, phenylephrine infusion and partial occlusion of the de-scending thoracic aorta. The fine oscillations on the unfiltereddiameter signal were due to aliasing of the sonomicmmeter samplingrate and analog to digital converter sampling rate' these frequencieswere Well beyond the physiologic range and were eliminated by thefiltering procedures . All four heart smmds (I to a) ar^ demortr :rerrdon the phononrdiogram . A marked similarity was observed be-tween the waveforrns of S; on the phenonrdiegram and the filteredventricu!ar wall oscillations during

f ring-

analysis was used to detect relations between vibratoryamplitude or frequency and the independent variables of leftventricular end-diastolic diameter, systolic and diastolic leftventricular pressures and mean rapid filling rate . The effectsof vena caval occlusion or calcium infusion en individualvariables were tested by analysis of covariance .

ResaltsDemonstration of S3. After bandpass filtering of the TTY

minor-axis diameter signal in the veiume-loaded state, theeIse!Ur r;;,;,, of the left ventricular wall during the rapid fillingphase of diastole became readily apparent (Fig. 2). All fourheart sounds were discerned in the four dogs with anacceptable phonxaediogeam. Although there was limitedcorrelation between the filtered diameter and the phonocar-diogram for the first, second and fourth heart sounds, astriking similarity was evident between the waveform andtemporal relations of ventricular wall vibrations during rapidfilling and the phonoeordiographic S ., (Fig. 3).

Volume loading, phenylephrine infusion and partial oc-clusion of the descending thoracic aorta produced abnor-mally high left ventricular diastolic pressures (Table 1) . Thethird heart sound was not present on the phonocardiogram inthe control state, and it became apparent only after leftventricular diastolic pressure was increased by the loadinginterventions . Inferior vena cava occlusion decreased theminimal diastolic pressure and end-diastolic pressure (p <0.05) (Table 1), whereas the rate of ventricular filling (asevidenced by the slope of the minor-axis dimension curveduring rapid filling) remained essentially unchanged (Fig . 4) .The decrease in ventricular filling pressure was accompanied

at-,

na-

o-

U

a

u

U

m ata

m

8

2E

m

S

E

^,~ Rr v:~ X1 0.

•

m

•• m

IACC Vet, to . Nn. 2FehueOt, 1992 45e-7

IACC .'oI 19, Nn 2February 1012-4 ,5-7

too-

Figure 4 . Superimposed high speed plots during ra, ,.d ventricularfilling after phenylephrine infusion, volume loading and aortic oc-clusion (Cartrol, rolid line) and subsegecot steady state inferiorvena cava occlusion (VCO, dashed line) in a typical study. With thefall in filling pressure during vena caval occlusion, the frequency ofwail oscillations decreased (dashed lire, middle panel) . Similarresults were obtained in all seven dogs .

by both a decrease in the amplitude of left ventricular W, 11oscillations (p < 0 .05) and a trend toward a decreasedfrequency of wall oscillation (p = 0 .09).

Mechanical correlates of S3 amplitude and frequency .Fourier analysis of the filtered minor-axis dimension and thephonocardiogram quantified the change in frequency ofventricular wall oscillations d..-:-r-.̀ c

vennons (Fig . 5) ."With decreased filling pressure during vena caval occlusion,the total vibratory energy, represented by the area under theamplitude Versus frequency curve, usually changed mini-mally; however, the amplitudes generally shifted to the left,so that less energy was present in the audible range above20 Hz .

Infusion of calcium increased both the velociry of icetricular filling and the amplitude of wall oscillations,whereas diastolic pressures and oscillation frequency did notchange significantly (Fig . 6 and Table 1). As the rate ofdiastolic filling increased, total vibratory energy (again rep-resented by the area under the amplitude versus frequencycurve) increased for both the filtered diameter and thephonocardiogram, whereas the fundamental frequencies re-mained approximately the same (Fig . 7) . Throughout thestudies, spectral analysis of the phonocardiographic dataseemed to be somewhat limited by a greater amount ofextraneous noise in the phonocardiographic signal as com-pared with the filtered diameter.

Multiple regression analysis demonstrated a strong cor-relation between the mean frequency of wall oscillations andboth minimal ventricular diastolic pressure (r = 0.60, n =0 .004) and end-diastolic pressure (r = 0 .47, p = 0.03) (Fig .8). Higher diastolic pressures were uniformly related tohigher mean vibratory frequencies. Similarly, the high mooramplitude of wall oscillations correlated strrng)y both with

GLOWER ET AL .

453TUIRD HeAna SOUND

r .t u,ntN IX+~--~ CaMrnl

FREOtUEWY (He)

Figure 5. A, Fourier analysis of the filtered dimension data for sixcardiac cycles during rapid venniewar filling in a typical study with

vied filling pressure. B, Corresponding Fourier analysis of thephonocardiogram for the same cardiac ccles . Frequencies <20 Hz,shown here with shading, were considered inaudible . With higherfilling pressure and wall stiffness during phenylephrine infusion,volume expansion and aortic occlusion (Control, solid line), thetrequcncies shifted to the right into the redible rouge and were

so phonoeardiogram by a similar increase in soundintensity in the audible frequencies as compared with vena gnatocclusion NCO, dashed line) . Similar data were obtained in all dogs .

filling rate (r = 0 .80, p = 0 .0001) (Fig . 8) and to a lesserextent with end-diastolic diameter (r = 0 .46, p = 0 .03) . Withall of these relations, however, pare interactions could notbe cytnieed in the biologic setting because both filling rateand pressure tended to change to some extent with allinterventions (Fig. 8).

In examining Table 1, it should be noted that partial venacaval occlusion decreased left ventricular end-diastolic pres-sure, end-diastolic diameter, peak systolic pressure andminimal diastolic pressure only slightly despite the signifi-cant decrease in amplitude and frequency of SP Theseotherwise small changes were necessary because any signif-icant decrease in left ventricular filling by full vents cz -valocclusion caused S3 to not be detectable. Similarly, in analready distended ventricle, the effect of calcium was pri-marily to increase peak systolic pressure and mean fillingrat, without a large increase in end-diastolic diameter orpressure .

DiscussionEtiology of Ss . Shortly after Potain (1) attributed S 3 to

sudden tensing of the ventricle during eapid ventricular

454 GLOWER ET AL.IHIIte HEART SOUND

30

E

:00

o

0 mua

Figure 6. Snprnumpose't high speed plots of loft vontoeafan diameter(LVD), phonoeardiogram (Phone), filtered diameter and left ventric-ular pressure (LVP) during rapid diastolic filling in a typical study.With increased filling rate demonstrated by the increased slope ofthe niinoraxis diameter curve (Calcium, solid line), the amplitude ofthe rapid oscillations of the filtered diameter and phonecardiogramincreased (middle panels) while the fundamental frequencies did notchange significantly . Similar results were obtained in all dogs.

filling, alternative theories associated gallop sounds with themitral valve apparatus (2-5). Although many investigators(2-5) have postulated that either tensing or cloaure of taemitra) valve might produce an S„ the -ale of valvular and

Figure 7. .A, Average Fourier analysis of the filtered dimension forsix cardiac cycles during rapid venlnet lar filling i.t a typical studyunder control conditions and during calcium infusion . B, Corre-sponding Fourier analysis of the phonocardiogram for the samecardiac cycles, The amplitudes of wall vibration increased signifi-cantly in both the phonocard :ogram and filtered dimension withcalcium infusion, while p- : fundamental hegccnnies were un-changed . Again, the inaudible frequencies (<2il Hz) ate indicatedwith shading Similar data were obtained in all does .

-CalmCallrul

FREQUENCY (He)

-c.race .

JACC via . Iq 11., 2Fehreosy 1942459-7

yI v28

c 23

5065

25

35

a5

Ia7

no

15

is

M."' 11`9 R'""

M n mat Diastahe Pr0SwraA

torsion

B

(neSq)

Figure B. A, Relation bclwt: .n an left ventricular filing rate andmean amplitude of venlnico .r V. all vibrations for all seven experi-ments. B, Relation between left ventricular minimal diastolic pres-sure and mean frequency of ventricular vibrations fur all sevenexpenimrnts. Open mangle - vcna caval occlusion ; open circle =control ; open square = - :aleium infusion .

subvaPndar structures has been disputed Sy several clinicalreport ; (23-25). In fact, C, gallop sounds have been observedin patients with an incompetent mitral valve heterograft (25),and patients have been described with a persistent S 3 afterresection of both atrioventricular valves (23,24) . Thus, avail-able evidence suggests that the mitral apparatus contributeslittle to S 3.

In 1942, Boyer (6) proposed the theory that early diastolicmotion of the heart caused an impact on the chest wallproducing an S3 . Reddy et al . (,,8) noted that recordings ofS3 within the left ventricle were of lower amplitude thanrecordings of S, from the chest wall, thereby supporting thetheory that S3 results from impact of the left ventricle on thechest wall or from coupling of abrupt change in left ventric-ularvelocity to the chest wall . Nonetheless, Craige et al. (12)dispute the theory of chest wail impact because they wereable to record S 3 vibrations from the epicardial surface withthe chest wall removed . Furthermore, other studies (13,14)demonstrated that S3 vibrations were initiated with a markeddeceleration (leaned negative jerk) measured from both theepicardial surface and the chest wall, whereas Craige et al .(12) spectfiated that impart of the he... : 2:rrsi th st waitwould produce an outward acceleration or positive jerk .Intracardiac pressure fluctuations also appear to be inade-quate to account for the genesis of S3 because an intracar-diac S3 may become apparent only during the terminal stagesof heart failure, long after it is present by external cardiacrecordings (8,12,13,26) .

When the body of research concerning S 3 is reviewed inlight of the present study, one may conclude that Potain (1)was correct in suggesting that S 3 is produced as the rapidlyexpanding ventricle reaches the point where the fibroelasticnature of the myocardium limits ventricular distension (9-16). The resulting sudden resistance to further chamberexpansion under tension seems to generate a vibratory wavefront that is detectable as a low frequency evet,t by both thesonomicromeiet and the phonocardiogram. Our data fromthe current experiments are consistent with this ventriculardistension mechanism .

JACC Vul, 19,N,.2February l592 :aIl-7

Mechanical correlates of S, amplitude and frequency .Despite general acceptance of the ventricular distensiontheory, data regarding the actual determinants of S, ampli-tude and frequency are limited to a few clinical studies . Inchildren, Ewing et al. (15) found that S, energy <15 fizdecreased as filling velocity decreased with Increased mitralvalve orifice area, and they also observed that the highfrequency component of S 3 increased with age (and possiblyincreased with ventricular stiffness) . Porter et al . (27) notedthat S, was most prominent in patients with increased peakventricular filling rate or with increased wall stress at thetime of peak left ventricular filling . Van de Werf et al. (28)noted that the presence of S, correlated with an increasedrate of left ventricular filling (dVidt) in patients with moralregurgitation, and S3 correlated with impaired left ventricu-lar relaxation in patients with cardiomyopathy or constric-

tive pericardais . Longhini et al. (i,~) demonstrated that heenergy content of S 3 shifted toward higher frequencies inpatients with increased ventricular wall thickness and de-creased left ventricular diameter .

The current study supports the predictions of the ventric-ular distension theory with the frequency of S 3 influenced byventricular diastolic pressure or wail stiffness and with theamplitude of S3 deterrnined by left ventricular filling rate(Fig. 8). These mechanical determinants of the third heartsound may be understood using a simple acouOc model ofthe left ventricle as an elastic sphere impacted by a mass ofblood during ventricular filling, with resultant vibration inthe left ventrirtdar wall. The left vent riculor wall w aald thenbe represented as a spring with elastic constant (k) and mass(ml,.), and the blood impacting the left ventricle during fillingmay be considered to have mass (m s) and velocity Byconservation of momentum, the resultant initial velocity(vl„) of displacement of the left ventricular wall would herelated to the blood velocity (29)

v 1 , = vs (mslml,) .

III

The amplitude (A) of vibration of the wall will be (29)

A = v,,N/(2]

or by substituting equation I into equation 2,

A = yr (metm v )

13]

Thus, the amplitude of vibration will be roughly proportionalto blood velocity (or rate of left ventricular filling) andinversely proportional to the square root of the left ventric-ular wall stiffness constant . Using this elastic spring model ofthe ventricular wall, the trquency of oscillation (f) would be

(29)

2rf = \/i o .

141

If the left ventricle has significant dampening with dampen-ing constant (c), the frequency of oscillation will be slightlyreduced to (29)

2rrf = V/(kJmm) - (el2mw)' .

)5j

GLOWER FT AL .

455THIRD IIEART SOUND

Note that the frequency of wall oscillation will be roughlyproportional to the square root of the wall stiffness constant(k). Note also that, barring any change in wall stiffness,increased wall mass (as in left ventricular hypertrophy) mayor may not produce an audible S 3 because S3 frequency will

decrease (equation 4) while S 3 amplitude increases (equation2) . Although modeling of the left ventricle as a linear springand modeling of left ventricular filling as an impact betweenblood and the ventricular wall is clearly a simplification, thedata presented in Table I and Figure 8 are consistent w ithsuch modeling.

Delertion of S3 by sonumicroWdry. Sonomicronactry iswell suited for studying wall oscillations related to S 3 . Thistechnique has maximal sensitivity to the most subtle ventric-ular vibrations over the frequency range of 0 to 250 Hzwithout the disatw:ttages of phonocardiography, whichmay pick up extraneous noises . Dynamic analysis of ven-tricular geometry and wall motion are possible with highresolution . stable electrical calibrations and negligible driftor temper .tture sensitivity. Additionally, the method is nothindered by the dampening effects of intervening tissues orextraneous noises that plague external monitoring tech-niques (20) . Thus, sonomicrometry is ideal for the experi-mental application described in this study and could be

extended to Ibe clinical realm through intraoperative orpostoperative studies of cardiac surgical patients (35] .

Throughout tl:e picuent stud, , results of sonomicrometryand pboroc.rdicgsaphy were is agreement oil the presenceor absence of S3 before and after volume loading and venacaval occlusion. Moreover, vibrations measured by epicar-dial sonomicrometry and S3 measured by standard phono-cardiography correlated both in amplitude and frequencyover a range of ventricular filling rates and ventricularvolumes and pressures (Fig . 3, 5 and 7). These correlationssuggest that the vibrations measured by sonomicrometry didrepresent the S3 detected by phonocardiography.

The studies of Ishimitsu e' _1 . (31) suggested that themajor axis of the ventricle played a more dominant role inthe genesis of S, than did the minor axis . In a comparison ofventricular wail dynamir .T in Win byperdyndmic and hypo-dynamic canine models, these authors noted that the peakmajor-axis dimensional velocity invariably was greater thanthe corresponding minor-axis velocity when an S 3 occurred .The diastolic differences between the minor and major axesare not surprising and are consistent with the anisotropienature of the intact canine myocardium with greater stiffnessin the major axis than in the minor axis (32) . AlthoughIshimitsu et al . (3t) noted better correlation between S 3 anddeceleration in the major than in the minor axis, major- andminor-axis deceleration differed primarily in amplitude andnot in their relation to S3. The minor axis was chosen forinvestigation in our current study because of the relativefacility of minor-axis transducer placement and because thechanges in the minor axis with the appearance of an S 3 weresufficient to test the hypotheses outlined in the introduction .

456

GLOWERETAL.THIRD HEART SOUND

Timing of S3. In the current study (Fig. 3), S3 followedrelatively early after S 2 under control conditions . This rela-tiveiv early timing of S 3 may have resulted from the high leftveturicular filing roe produced by volume loading andelevated afterioad in the control state . Earlier occurrence ofS3 with volume loading has been reported by Van de Werfetal. (33) . In addition, the minor-axis diameter tracings ofFigures 3, 4 and 6 might suggest that peak filling occurredslightly before the nadir in left ventricular pressure, whileother investigators (9) have observed peak filling rates muchcloser to the ventricular pressure nadir. Much of the appar,ent discrepancy may result from minor-axis lengtheningoccurring earlier in diastole when the left ventricle is full (asin the present study) than when the ventricle is emptier (as inother studies) because of the more spheric shape of the fullventricle (22) . In addition, the volurne and pressure loadingin the present study (mean left ventricular end-diastolicpressure 29 mmHg) might have caused the moral valve toopen prematurely, thus causing peak filling to occur beforethe nadir in left ventricular pressure.

Assessment of wall stillness . One limitation of the currentstudy was the indirect assessment of wall stiffness usingdiastolic left ventricular pressure, which could be moreeasily measured. Diastolic pressure, however, should be areliable indicator of changes in wall stiffness because of th,well defined exponential pressure-volume relation of thediastolic left ventricle (21) and because none of the interven-tions used to alter cardiac load or inotropic state affect thebasic relation between left ventricular pressure and volume(and thereby pressure and stiffness) (21) .

Because of the exponential pressure-volume relation ofthe left ventricle (21), an increase in ventricular diastolicpressure is associated with increasedT%ll stiffness defined asthe slope (dP/dV) of the pressure-volume relation . Thisincreased wall stiffness would be expected to produce anincrease in the frequency of ventricular vibrations for anygiven wall displacement induced by rapid filling (equation 4) .A higher vibratory frequency would cause a ecr espondingshift of more acoustic energy into the audible range above20 Hz, which might be perceived as S3 becoming audible . Inthe intact heart, the relation between S 3 frequency and wallstiffness is supported by the observed rightward shits infrequency spectra of ventricular vibrations with increasedfilling pressure (Fig. 4 and 5, Table 1). The similar correla-tion between filling pressure and S 3 frequency observed withthe external phonocardiogram in four dogs corroborates thisobservation (Fig. 4) .

Assessment of ventricular filling rate. A second limitationof the present study was the use of left ventricular diameterto estimate relative left ventricular filling rate . AlthoughStinker alai Glantz (34) demonstrated the potential inaccu-racy of ass.ssing instantaneous volume or rate of change involume from left ventricular dimensions, left ventriculardiameter was used in the pressnt .:!:iy only to detectrktive changes in left ventricular filling rate. Other inves-tigators have shown good correlation between change in

25CC VoL 19 . No. 2Feb,na,y 190:45U-7

ventricular volume and change in left ventricular dimensions(20,35), es,ccially when high frequency resolution is notneeded (34) . Because the assessment of directional changesin mean rate of ventricular filling requires relatively lowfrequency information, the observations of Stinker andGlantz (34) are •.css rekvaui W the present study.

In these experiments, calcium was used to increaseventricular stroke volume in a volume-overloaded systemwhile comparable diastolic pressures were maintained de-spite more efficient ventricular emptying . Because rapidventricular tilling intervals remained similar, mean fillingrates increased (Fig. 6) . At higher filling rates, the amplitudeof wall vibrations increased significantly (','able 1, Fig . 8),but mean frequency changed little. Although it was impos-sible to vary filling rate or pressure in a completely indepen-dent fashion in the intact heart, the spectral data suggest thatthe amplitudes of S3 and ventricular wall oscillations duringrapid filling are proportional to the filling rate as predicted byequation 3. Higher filling rates would increase S 3 energy andcould be perceives as an S3 becoming audible.

Clinical implications for 53 analysis . The observed rela-tions between filling pressure, wall stiffness and the fre-quency of wall vibrations may have clinical relevance to thepathologic situations of congestive heart failure or ventricu-!2r hypertrophy, where an S 3 may become evident becauseof increased diastolic pressure or wall stiffness . The corre-lation between filling rate and amplitude of wall vibrationsmay have a clinical analog in high output states, such asthyrotoxicosis, or with lesions such as valvular incompe-tence that increase ventricular filling velocity. Finally, onemight hypothesize that the development of an abnormal S3 isdependent on an increase in either myocardial stiffness orrate of ventricular filling. t lihough similar principles mightapply to the right ventricle, caution should be exercised inapplying observations from the current study to a rightventricular third heart sound because the current studyexamined only the left ventricle.

It is reasonable to propose that different types of S 3 couldbe defined by spectral .nalysis of frequency and amplitudecharacteristics and that the data could lie used for patho-physiologic interpretation of distinct clinical entities . Cur-rently, the presence of an S3 gallop plays a qualitative role inthe management of patients with cardiac disorders (36-38)but spectral analysis could transform the phonocardiograniinto a quantitative clinical tool. As Robert Hooke (39) soastutely stated in 1705, "Who knows, I say, but that it maybe possible to discover the motions of the internal parts ofthe bodies . . . by the sound they make, that one maydiscover the seeks performed in several offices and strops ofa man's body and thereby discover what instrument orengine is out of order ."

We are grateful for the assistance of t .ouisr. Leighlon in preparing themaeu,aei,t and of loan loner in preparing the illuvations .

001105 Peat'P+'411UI0!1lu3sgale 10 "oua11n00H "3H YI'w5•30110000'SZ'8-10150ZL61 f 003(11 06"!1310

comes 1ivjn OaiILLLS11

'-3V 12 a3MOYJ

-6at'I15i5561 f 00054010001100agn155100l01dI10510051100010!0024000u' pu40spea4 01140 m0olm054d 001'3 0swglvl' j 1pPuup'003021'p

-5, Z.1Aa1"'q'd7e!:'61'0A,331f

e alua!Ied)00H 011001ploppe;a' 5 fIIq'oe!d•0171114sug0(•~a3'l-BE'8-091:3911861 05)!50!"110TH

-0805590113110!11 uI dq,A Punas Pewl501470015 fo50110;005a!u~-U,:pcW

-lot-Ss100 0060 0013 0_) -lop x..:000- 1010u! umpwa0m1011001010

-,ca

"Sf11111003'd 05111100.3'M1 slneCI'M!)aa!tw'S!)uusAj'(j 0-10'61pus S !OSpUP!s OO!u!p-I N opaue3•y5 ugu3'IW qu5+d'WV sll"PIV "LE

Oi Sfs,46Z!LL61 P°W fPugN '-pa3old PS®+nss 3t!plnuou u! Is" atlplea

'81-LZ1:5:L961 8.3158 P155 "Agderao!potruoud001 50aq!A'l-q-0'H 013!13 'p3f -IN-El'A3uawyualloA-81

'09-LSRH0L0Z1p961Jo5apufP!a0100J:(np; le 00'35 wnsg-N'10 HOPI)10000109'9E

-6-96p:9E'sL61 paP:A43IddV I'A1Pr 0 ei.plwn 2401 sup. pen!0011 Jo 0alaurz!p 1001! wnlo0wnpsgelo 0000555x51"N-&S 'Ii el°S '9C

-0L-191:95'S86[ sap >n,)'00500116 u-S a 11.x500

I-Agd f 1011--I'Mid--Ii-!usuual-8uol 1-JnwP-1 u0!suaw!Pae!p150 Pale- Appemauua04 V '1s 10.00nmo10'fl111!V'000x10'41

'EL-SS•Z:CE'0361 f P53H udJ 'P--OS Leaypnyl lesiolms6yd 004110 s!sau-l 04110 0 ii iAq 1000108uuds-sroV "300 004/11'3 105-_4'00100 '30000-/5'S 0!901'3'00(/11101 -91 01 papawcioI0'0v'nJo 1aoanbroj oql uo spuadap uo!suaw,p 500J

wvlon0e1nau01vw11008100 )0010 Aaunaav agl'VS-19'P3.Nu04S 'tl'I-OOpUCL!P9611103011013 f'UPop

'tS-Ior:6Eb861 Igp003 0111'A4de58ap0ewgsaput auoud Aq umppuo UI00005 unq P1140 05! Jo4.j0V'3u00wyua0oAu5A:5(001-3'305190100'1depwmaw'I)

'51 0! Apnl2 MqJ-!455004d-Spa- Peat 50001 put poly, 141 Jo 5,50002'000-06£:19'E861 00901000!'spafgns 00009 w

sa'01115'II'Punosueaypn,y005110u!8p03I9!uy'0g5wS'Ae00et0 41'8-E6E19lf861 uonslmn7'Sop u,

'A aqiH10!0193) 'H JsaU9M 'd 00'P+u+-3 'f "'5000J0MaP u0A "EE'8L-591:69'1661

sap 3014-lopsnapsu00 aup 0! Uppw-n 301 asp 1o saondold a!dolios!ue50!01100'I'P00000001100041001)0 0800 3080)'3 40+1S 'A0030'10'S-E9LSE'datl 1'11!0003

3goloH0-5fuqueL'VA11024010"f%-I'00 -10.03 0x10 TE'ZL-59R S:S86[ 10!500311)uR'vi---0!-0031 "PwwO 00004 NIT 14110 0500 "0 41-S 'A t- '_T 00000_) ZI Pa. --5P-U4 u! sswsoAp 11000000.1.00 Jo-..d.-'00-509-613961

u0!1tp0o11-)'UP4onl11Ee4duns)00001141 all 10-0t-SSJOwl!u .41 .000000lnagP+!41Jo'ono,3111!u~•g°I-U*(G•;m0•ynsnwlgUJ-lc'52-91-1:(1 I0dos)97Z961541100 N1. '1,00 001601'.'31110a1'af uoapM'MI wo4M'"100U010003 11 00~01oS110 8uP1e0HssedX.g 4Jaofo Rl 0000! 02Jit 100110;:01. ,0011901'Z0-SEOl:9C1L56, 0501101 093 I--pod1o 550500011 04'10100!

'01 p 03 01WS00510!c 100051'5U U'MO0a! 14.0)00¢10'+159U05A1

z-ml'os'LIdoge9 0,III3 'AV IezuyM '5M Nowagelg'S90ewm1'1d0051'01

'L-0000110.115 i' o' 11yj ('5905 'wOOn!pnd

'd :2456!

3bW(1.A51,0mdiujp.s

--13 6Z'9-OL0I:69108610510!0053

m Apl!s h+op.unccvd V"spunos 0000/101101P-P,."Iq,;0--1 501 'H -)III-1'11 15000 a0 'd 1 .!j1010,-)'1 uau!w'A 10005 ap -,A '6 spunos 100 .q png1 0!M5700!70/1un pae soInp, 101010 m apa...I£ £ 121115961 p!P+ J f011-510050044 pig, /01'5/1 hppu)d S

2401 ay11urap5adeid0!IoouIep'151aup7'Vi 4I'dI00M0P 11".1'III'L-QLC6!IL61 Paw f wt' "spun-

ooogdollu3ggm51ualard ul sa0lgl1salwnu6P Put suo15uawip 1r(na!nuan

ZCZZ6:C9a86100501110!) Agdwlolp0n5ade pue "ouogd anoalouan6 Ay-w2nSanw500905 /101105 to 5!5000814.1 01010010' 13 551013'4 0000'5/1 55005 'L

U S! 87.0!,003'w1w1"3'+!0011/WA'VM d70RH '01300000'(0ZAP'p-ZH'E£!CS6l 1240001.lay•6gdn08

o,piooouogd lN1oiN006 A4 spunos 10019 10011)51 pug o2'p1We+wl10 Apnss pwawp(dri 'Sd•(0003 '3 0103 'Dg SAUPO'300uagnV'9Z

-!6163¢61 f ueaH u051pun0s 000'9 PI-41N100s0!PSIS'HNOaAug-9

PnY0001001110000016-M W6%1 f ueaH 024'pun0s WIN

04110 w8uo0101001011'00+0103005'03'5(90001013 '5005-L6t 2£'OL51 11000H 05 ,gdoioaapq I .0. -Id...! wPS

'6C:4JLC'PJaJfeM Put 3P:011"0.1•;a n;UP CWOPut salnpa-l'L-116410965SoldPad0011543s!NauwieWO3 a!OUH asiq p Jo s3OM sn0wngls+d 0011-00010011gal pe/w000fI0000 O0d105050000 0:1015024: a1-Mafps.10-s1MUf'Ip3gwagPusrnHn3I,- g9000000Jospugpuaay7ggm-Oapuv'ma01 /51'IIH3/P1e00P3'5M8011'3300010101'51mque4'IZ'fuP--!pSI Agdow1e1d 1w01EU jo Aawiyzp Wxaid 024,1,'01400H'6E

SOL'£i 20£1609161 SIN 011, $OP 500:750 agl

!app an50j00730AOUwoo20750-'0.5111euWSuaw[p-0001ayj'M3u05upOUV'1)3101060003.0Po111'3juslu00V'VA00100W'Sf0051003 -66490E!Z961pawf IJu9N -midsauna ulmoH!p 101ep1,nnwopuuoz

mefdal a,Jl, IsPyw01510 005000045!4113u!.3 'NP-YcI4^j1Z'Z-WI ;i_ 9!61

-5110-!8:0'6061 Paw000101 9NV 'PunosWINpug?ay1 as 000! 030x50 n'Snd'SM 000451!!'C

lslg3'saoltop!ds40ut PutT,uwINltu 011115 Ug9O W al niJ50444'. boll'f: 0 51'00061100053 oslnd01141('91 f0ON(165+0d'51901um04310)51100004'0UN)'I1M'dumulL0000841017 m 10000 00410000200 oua40(t JoOIUC,Plu8151(1'91000019 'Z

"55-p&6T1961 sap 00!0Sop,egat0001a1I.:OIW;00IJOIonJo s350/03nweg0 I0OUgu!wupP'91 Mauovap1V Vd -THIN 'SM50!3'300007-01'Sf -!5153'03 5'510'7

"9-SLI:OZl0061 PIN wftwaS 'dofie9 a5 s11Nq 5x1'3010003'I