British Heart journal, I971, 33, 702-706.

'Pixie' cardiographyAccelerometer applications to phonocardiographyand displacement cardiography in childhood

Frederick E. Bew, Douglas Pickering, Peter Sleight, and Frank D. StottFrom The Radcliffe Infirmary and Medical Research Council External Staff, Oxford

The technique of deriving phonocardiogramsfrom the 'pixie' accelerometer record by appropriatefilters is described and typical examples are shown. Velocity and displacement traces have also beenobtained by successive integrations of the praecordial acceleration trace, which compare closelywith those recorded by conventional photoelectric methods.

An accelerometer is a small transducer whichsenses the acceleration of any object to whichit is attached. It usually consists of a small bobweight supported on a cantilever spring (Fig.i): upward acceleration of the base causes a

downward movement of the bob relative tothe base, of a magnitude dependent on themass of the bob, the stiffness of the spring,and proportional to the acceleration appliedto the base. The resonant frequency of thespring bob combination is a function of themass of the bob and the stiffness of the spring.If a high resonant frequency is required, thespring stiffness must be high in relation to themass, and the deflection is correspondinglysmall. When adjusted to give a resonant fre-quency of i kc/sec, the spring mass combina-tion is such that the movement of the bobweight for an acceleration of i g (98I cm/sec2)is about 02 ,Lm.

FIG . I Basic components of 'pixie'accelerometer.

CANTILEVERSPRING

4~~~

Received 2 December 1970.

MethodsTwo different methods of detecting the movementand converting it to an electrical output have beenused.

(a) A semi-conductor strain gauge The type usedis designed specifically for use as a cantilevermember, and is consequently very easily incor-porated in an accelerometer. It is supplied byEndevco Ltd., under the trade name 'Pixie'.This transducer gives an output of 40 mV/g withioV across the strain gauge.

(b) An optical system using a subminiature lampand photo-transistor (Mullard BPX25). This givesan output of about IV/g.

Both transducers have a resonant frequency ofapproximately i Kc/sec, but the advantage of theoptical system is not nearly as great as mightappear at first sight, as the phototransistor isinherently far more noisy than the strain gauge.All the clinical work reported here has in factbeen carried out with the strain gauge type, andthe optical transducer has been used only forexperimental comparison in the laboratory (seebelow). Both types of transducer show negligiblehysteresis and non-linearity within the workingrange.When sound is transmitted through a solid or

quasisolid medium, vibration will be detected atthe surface of the solid, along an axis normal tothe surface. The movements of the chest wall dueto the events of the cardiac cycle will thereforecontain lower frequency components correspond-ing to the gross mechanical events, and also higherfrequency components due to any events thatgenerate sound, such as turbulent flow.The amplitude of the vibrations due to events

generating sound will obviously be very small,but for a given amplitude of sinusoidal vibrationthe acceleration is proportional to the square ofthe frequency, so that vibrations in the audible

PLASTIC BASE

UPWARD MOVEMENT ACCELERATION

on Decem

ber 26, 2020 by guest. Protected by copyright.

http://heart.bmj.com

/B

r Heart J: first published as 10.1136/hrt.33.5.702 on 1 S

eptember 1971. D

ownloaded from

'Pixie' cardiography 703

frequency range which cannot be detected by a

displacement measuring device may be detected

by an accelerometer. These vibrations are detected

indirectly by the conventional type ofphonocardio-

graph microphone, by conversion into changesof air pressure in the closed cavity formed when

the microphone is applied to the chest wall; these

changes of pressure in turn produce mechanical

movements of the diaphragm of the microphone,which in turn produce an electrical signal.When an accelerometer is applied to the chest

wall, the vibrations at that particular spot are

convented directly and quantitatively into an

electrical signal, the praecordial acceleration

trace.

The accelerometers mentioned above respondto all frequencies from zero upwards, and can

therefore be used in conjunction with appropriatefilters to display low-frequency accelerograms or

heart sounds at choice. It is also possible to

integrate the accelerometer output electricallywith respect to timne to produce a velocity record,

and to integrate a second time to derive a dis-

placement record. Absolute recording of velocityor displacement, which are relative quantities,can only be obtained by using some form of fixed

reference point or framework; this is not necessary

for recording acceleration. There is of course an

arbitrary constant of integration introduced at

each stage in deriving velocity and displacement:these are simply assumed to be zero, i.e. the

subject as a whole is assumed to be stationary and

so himself provides the frame of reference.

Velocity records are easily obtained in this way,

and appear to be easier to interpret intuitively in

terms of cardiac events than accelerograms. Dis-

placement records still present considerable

FIG. 2 Accelerometer attached to chest wall

of an infant.

technical difficulties except on well-trained co-

operative subjects, due to the large magnitude of

the respiratory movements compared with the

cardiac displacements. There is not enoughdifference in frequency to separate the two byelectrical filters without introducing gross dis-

tortion of the cardiac displacement record.

In an attempt to assess cardiac defects without

recourse to catheterization, a number of workers

FIG.- 3 Characteristics of the band-pass filters used to produce phonocardiograms from 'raw'acceleration signal.

C~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~. I ......W..

1 4~~~~~~~~~~~~~~~~~~~~~~~~~~~F77r

...T...~~~ 1-t EILT#kSt~~~~~~~~~~~~~~~~.......6 .I.....

.....

4uF#4c,y ~ ~ ~ -

3

I

." i. lt-.-...&

on Decem

ber 26, 2020 by guest. Protected by copyright.

http://heart.bmj.com

/B

r Heart J: first published as 10.1136/hrt.33.5.702 on 1 S

eptember 1971. D

ownloaded from

704 Bew, Pickering, Sleight, and Stott

ECG

FIG. 4 'Pixie' phonocardiogram of infantwith secundum atrial septal defect (usingfilters A + D).

have examined the possibilities of obtaining thenecessary information from measurements ofdisplacement or from the low-frequency com-ponents of the phonocardiogram.Eddleman, Hughes, and Thomas (i959) and

Eddleman and Thomas (i959) have used the dis-placement cardiograph to differentiate betweenright ventricular pressure and volume loads.A volume load gives rise to an early systolic peakwhich is prolonged into the last third of systolewith a pressure load. Similar applications havebeen used by Mounsey (I967), Gillam, Deliyannis,and Mounsey (I964), Deliyannis et al. (I964), andBenchimol, Wuh, and Dimond (1966); the sub-ject has been reviewed in detail by Luisada (I962).The accelerometer appeared to have important

advantages, particularly in children and infants.Many conventional displacement transducers arelarge and heavy; they cover an appreciable areaof the infant praecordium. This could seriouslydistort the signal recorded and also make it

FIG. 5 'Pixie' phonocardiogram of 3-monthinfant with ventricular septal defect showingseparation of diastolic murmur (DM)from2nd sound (see text). SM= pansystolicmurmur.

E

* ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~. .. .. ....~~~~~~~~~~~~~~~~~~... .. .. .... .. ...................................................................*\~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.:.........

AR ~~~~APA P I.

difficult to record simultaneously from two ormore areas of the praecordium. The small size ofthe accelerometer transducer (Fig. 2) thus becomesof increasing value in infants. Furthermore, onehas the ability to derive both displacement andsound recordings from the same small transducersimultaneously.A further advantage is that this does measure

absolute displacement not, as many techniques do,movement relative to the chest wall or microphonemounting.The first stage of the development work was to

obtain, by appropriate filtering of the accelero-meter output, a trace which was directly com-parable with the trace produced by a conventionalphonocardiograph. Phonocardiograms of goodquality were produced by introducing a series ofband-pass filters (Fig. 3) which almost completelyremove low-frequency components of the prae-cordial acceleration trace signal; by adjustingthese filters we were able to produce a series ofphonocardiograms as illustrated below. Theaccelerometer is easily stuck to the chest wall bymeans of a small piece of adhesive tape. Inneonates, electrocardiogram contacts were madeby means of silver discs o 5 cm in diameterapplied to the shoulders and the right costalmargin. These were found by trial and error to bethe sites where they were least likely to be movedby the patient.

ResultsTypical 'pixie' phonocardiograms recordedin an open ward are shown in the Figures.

Fig. 4 is a record from the pulmonary areaof a baby with an ostium secundum atrialseptal defect. There is a grade 2/6 ejectionsystolic murmur and fixed splitting of thesecond sound in the same site.

Fig. 5 is a recording from a 3-month-oldinfant with a ventricular septal defect. Thepansystolic murmur is well shown and thediastolic murmur in this child proved of somediagnostic importance as it was suspected thatthe child might have a prolapsed aortic cuspgiving rise to the murmur of aortic incompe-tence. After seeing the record it was apparentthat the murmur began some time after thesecond heart sound and thus saved the childfrom an unnecessary aortogram.

Fig. 6 shows a continuous (Gibson) mur-mur in a i-year-old child with a persistentductus arteriosus.

Fig. 7 shows a recording made from a 4-year-old boy who had a severe tetrad of Fallotand his pulmonary valve was replaced with ahomograft aortic valve. A few days afteroperation the valve developed a slight leakand a grade I to 2/6 early diastolic murmurwas heard at the lower right sternal border.An ejection systolic murmur grade 2/6

on Decem

ber 26, 2020 by guest. Protected by copyright.

http://heart.bmj.com

/B

r Heart J: first published as 10.1136/hrt.33.5.702 on 1 S

eptember 1971. D

ownloaded from

through the graft aortic valve can also be seenin this record.

Fig. 8 shows a recording from a 3-week-oldchild with a hypoplastic left heart syndrome,confirmed at thoracotomy at St. George'sHospital, London. He was later transferred tothe care of the Radcliffe Infirmary at Oxford.Our accelerometer derived trace (Fig. 8) isshown above the conventional phonocardio-gram. (Courtesy of Dr. Aubrey Leatham atSt. George's Hospital, London.)

Displacement records We hope to de-velop methods to overcome the difficulty inseparating cardiac from respiratory chest walldisplacement records.

DiscussionThe first stage of development of this tech-nique, using an accelerometer to recordphonocardiograms in infancy and early child-hood, is now reliable and well established, asshown in the figures above.The accelerometer is a satisfactory alterna-

tive to the usual microphone for all phono-cardiographic work, and for use on infants itis much to be preferred, because of its smallsize and ease of fixation.The accelerometer also quantitatively re-

cords the low frequency components of theheart beat, but this is only a potential ratherthan an actual advantage at present, since theinformation contained in such records cannotyet be properly interpreted.

Records of displacement rather thanacceleration are much easier to comprehend,and we believe that it is essential to solve theproblems involved in the double integrationof the accelerometer signals to give displace-ment recordings; but the presence of respira-tory and other body movements, especiallywhen recording from children, adds to thedifficulties. That the method does holdpromise enough to justify further develop-ment is, we feel, shown by Fig. 9 which showssimultaneous displacement records obtaineddirectly (using a fixed reference and an opticalsystem) and by double integration of theacceleration tracing. Fig. IO shows thepraecordial acceleration trace recorded with

F I G. 8a 'Pixie' phonocardiogram of neonatewith hypoplasia of ascending aorta. ESM=ejection systolic murmur.FIG. 8b Conventional phonocardiogram ofinfant shown in Fig. 8a. RESPIRATION=inspiration upwards.

FIG. 6 Phonocardiogram showing a persistentductus continuous Gibson murmur (GM) in ai-year-old child.

FIG. 7 Phonocardiogram of 4-year-old withpostoperative pulmonary incompetence (EDM)after total correction of tetrad of Fallot.ESM= ejection systolic murmur.

ESM 2:I EDM:.ti ': 1. U.-ha

PCG ffOV. > _

a

ECG A^__esES

. 1.... ... 1 .I 11t I It I1 I I t I I I i

.C Ii $&1

bi I I I I i I I

KJ~~~~~~~~~~~L~ ~.sN~~ ..6:WIWISa

RESPIRATON

ECGOI A A~~---

'Pixie' cardiography 705

G.M.

2

i. L

'I'1 '-W-1r ' T

L..

11t 6 1.t'. 8 7 Tl- gsI. Tr_

I

--AvddbwL

on Decem

ber 26, 2020 by guest. Protected by copyright.

http://heart.bmj.com

/B

r Heart J: first published as 10.1136/hrt.33.5.702 on 1 S

eptember 1971. D

ownloaded from

706 Bew, Pickering, Sleight, and Stott

A

FIG. 9 Simultaneous optical (D1) and accelerometer derived (D2) displacement records.Outward displacement downwards. A =PACT trace. (UV recording retouched forreproduction.)

the 'pixie' accelerometer and below a tracederived by double differentiation of the photo-electric record. These records were taken on aco-operative adult subject, and neither methodof recording displacement would be workable

FI G. IO Simultaneous praecordialacceleration trace recorded directly (A1) andalso obtained by double differentiation ofphotoelectric displacement record. (A2)outward displacement (D) downwards. Thisshows that it is possible to derive accelerationfrom a displacement transducer by differentia-tion in the same way that the accelerometerrecord can be processed to produce displace-ment (by double integration).

ECG

Al

-oL1.ovement

on a baby; we expect however to be able toobtain displacement records on any patientafter further technical improvements havebeen made.

The authors wish to acknowledge the financialsupport given to us by the Board of Governors ofthe United Oxford Hospitals.

ReferencesBenchimol, A., Wuh, T.-L., and Dimond, E. G.

(I966). Apex cardiogram in the diagnosis ofcongenital heart disease. American Journal ofCardiology, 17, 63.

Deliyannis, A. A., GUam, P. M. S., Mounsey, J. P. D.,and Steiner, R. E. (I964). The cardiac impulse andthe motion of the heart. British Heart Jrournal, 26,396.

Eddleman, E. E., Hughes, M. L., and Thomas, H. D.(1959). Estimation of pulmonary artery pressureand pulmonary vascular resistance from ultra lowfrequency precordial movements (kinetocardio-grams). American Journal of Cardiology, 4, 662.

Eddleman, E. E., and Thomas, H. D. (I959). Therecognition and differentiation of right ventricularpressure and flow loads. American Journal ofCardiology, 4, 652.

Gillam, P. M. S., Deliyannis, A. A., and Mounsey,J. P. D. (I964). The left parasternal impulse.British Heart Journal, 26, 726.

Luisada, A. A. (I962). Cardiology, Vol. 2: Methods,pp. 3-64. (Suppl.) McGraw-Hill, New York.

Mounsey, J. P. D. (I967). Inspection and palpation ofthe cardiac impulse. Progress in CardiovascularDiseases, IO, I87.

on Decem

ber 26, 2020 by guest. Protected by copyright.

http://heart.bmj.com

/B

r Heart J: first published as 10.1136/hrt.33.5.702 on 1 S

eptember 1971. D

ownloaded from