Share this article on:

Acoustic Spectral Analysis for Determining Pump Thrombosis in Rotary Blood Pumps

Kaufmann, Friedrich; Hörmandinger, Christoph; Stepanenko, Alexander; Kretzschmar, Alexandra; Soltani, Sajjad; Krabatsch, Thomas; Potapov, Evgenij; Hetzer, Roland

doi: 10.1097/MAT.0000000000000097
Adult Circulatory Support

Implantation of rotary blood pumps is an established treatment for end-stage heart failure. Anticoagulation is routinely used for reducing pump thrombosis, but this severe complication still occurs in 5–8% of patients. In 105 patients, the sound produced by the left ventricular assist device HeartWare HVAD was prospectively and sequentially recorded. The acoustic spectra calculated by runtime fast Fourier transformation were investigated. Results were compared with respect to patients without thrombotic complications (97 patients) and those with signs of pump thrombosis (8 patients). All eight pumps exchanged were opened and investigated. Thrombus formation was confirmed in all eight exchanged pumps. The existence of a sound peak with the threefold frequency of the pump speed correlated with the presence of thrombi inside the pump (p < 0.001). An increase in the sound amplitude of the rotary frequency of approximately 75% compared to previous measurements also denotes pump thrombosis. Analysis of the acoustic spectrum of the HVAD pump can be used as a reliable method to detect pump thrombosis. The appearance of third harmonics indicates the presence of thrombotic particles independently of previous baseline measurements. The goal of this study is to define and validate such a method.

From the Department of Cardiothoracic and Vascular Surgery, Deutsches Herzzentrum Berlin, Berlin, Germany.

Submitted for consideration October 21, 2013; accepted for publication in revised form April 28, 2014.

Disclosure: The authors have no conflicts of interest to report.

Correspondence: Friedrich Kaufmann, Dipl Eng, Deutsches Herzzentrum Berlin, Augustenburger Platz 1, 13353 Berlin, Germany. Email:

Today, mechanical circulatory support is an accepted temporary or permanent alternative to heart transplantation.1–3 Pump thrombosis is a rare but severe complication requiring surgical intervention. Longer waiting times due to the shortage of donor organs for bridge-to-transplant patients and the growing numbers of patients on permanent support have led to a steady increase in the duration of support with implantable rotary blood pumps in recent years.4 Approximately 5–8% of supported patients have to undergo pump exchange due to thrombosis.5,6 Late pump thrombosis is associated with all clinically available systems for various reasons which are not the subject of this article. Thrombus formation inside a left ventricular assist device (LVAD) imposes the risk of embolization or sudden loss of function. Pump exchange or thrombolysis is mandatory, not only to prevent these high-risk complications but also to prevent further organ damage.

Diagnosis of a thrombus inside the pump is challenging. Hemolysis indicated by high lactate dehydrogenase (LDH) and free hemoglobin (fHB) levels or hemolytical urine,7,8 as well as signs of sudden drop of pump output or increased power consumption, is indicative for pump thrombosis.5,9 Echocardiographic data is able to support the diagnosis.10,11 However, the presence of thrombi inside the pump cannot be detected by imaging methods because there is no “window to the pump.” Only indirect effects of pump thrombosis, for example, the results of diminished pump flow, can be measured by these methods.

When there is suspicion of thrombotic deposition development in the pump, indicated by elevated power consumption but without clinical signs, additional technical information would be valuable for decision making. Obtaining reliable information about concomitant thrombus ingestion into the pump is important especially in patients presenting neurologic complications like intracerebral bleeding. The decision to exchange the pump has to be carefully weighed against attempting to solve the problem with, for example, thrombolytic therapy.12–14

The operating sound of the blood pump assessed by auscultation contains such information. So far no strategy of a standardized method has been reported. In vitro studies describe techniques to detect thrombi on centrifugal blood pumps.2,15 However, the large amount of artificial thrombus material used in the experimental setting displays the irrelevance for clinical use. Hubbert et al.16 investigated the sound patterns of Heartmate-II axial flow pumps (HMII; Thoratec Corporation, Inc., Pleasanton, CA) in an experimental model. Previous recordings of the sound pattern (“fingerprint”) are necessary. Also changes of speed and flow conditions produce alterations of the sound pattern, which probably limit the applicability in the acute clinical situation.

The investigation of VADs by acoustic analysis has been utilized earlier17 by focusing on changes of the acoustic spectrum caused by mechanical wear or destruction. The proposed method is designed to detect thrombotic deposits prior to any permanent damage to the device. The aim of this study is to isolate and validate the acoustic parameters that indicate pump thrombosis. The clinical relevance, reproducibility, and accuracy of this method resulted in its introduction into clinical routine in our institution. Further refinement may produce information about thrombus formation before clinical or technical signs appear.

Back to Top | Article Outline


The HeartWare HVAD LVAD System (HeartWare, Inc. Miami Lakes, FL) has been used at the Deutsches Herzzentrum Berlin since September 2009. Up to March 2013, the HVAD system was implanted in 308 patients. Pump thrombosis leading to pump exchange was suspected in 26 cases. The patients were readmitted to the hospital if high-power alarms occurred or hemolytic urine was observed. Readmitted patients were examined thoroughly for standard technical and laboratory parameters including auscultation and echocardiography. Prospective data was collected since June 2012. We routinely record the acoustic spectrum of all patients visiting the outpatient department. In all patients readmitted in this period due to changes in clinical, laboratory, or technical parameters indicating pump thrombosis, also an analysis of the acoustic spectrum was performed.

Back to Top | Article Outline

The Acoustic Spectrum of the HVAD

The acoustic spectrum emitted by hydraulic pumps can be divided into different components.8

Recognizable peaks (Figure 1) are produced by:

1. The turning of the rotor (e.g., 2,700 rpm corresponds to a 45 Hz peak) and its harmonics

2. Blades or flow channels passing the outflow volute

3. Peaks of different origin like electric interference (50 Hz in Europe—60 Hz in the USA, and their multiples)

Back to Top | Article Outline

Data Acquisition

The acoustic spectrum emitted by the running pump is calculated by fast Fourier transformation and displayed with a standard laptop computer connected to the data acquisition device (PicoLog USB DrDAQ; Pico Technology Ltd., St. Neots, UK) using the included software package (PicoScope 6—PC Oscilloscope software version: Recordings were performed with the onboard microphone of the DrDAQ, which was removed from the circuit board and attached to a conventional stethoscope head to pick up the pump sound from the patient. Considering the relevant low-frequency range the use of a high-end microphone did not prove to be necessary.

Set parameters are the following: spectrum range: 1 kHz, spectrum bins: 1,024, window function: hamming, display mode: averaging. We used the averaging function during 30 recordings to reduce stochastic noise to a minimum, producing an acoustic spectrum with significant characteristics of the examined pump.

Back to Top | Article Outline

Data Processing

After recording, the four described frequency peaks are identified and their amplitude measured. To normalize the peaks, the most intense fourth harmonic amplitude was set to 100% and the respective amplitudes of the normalized first to third harmonics were calculated. If clinical data (fHB >40 mg/dl and/or LDH >600 IU) and technical parameters (power increase of more than 1.5 W beyond average power consumption or calculated flow values of more than 10 L/min) consistently indicated pump thrombosis, pump exchange was performed. The explanted pumps were opened and evaluated for deposits.

Back to Top | Article Outline

Statistical Analysis

Data were analyzed with Predictive Analytics SoftWare Statistics 18 (Statistical Package for Social Sciences for Windows Release 18.0; SPSS Inc., Chicago, IL). Comparisons of the amplitudes between the two groups (no suspicion of pump thrombosis = control group; confirmed pump thrombosis = pump thrombosis group) used a Mann–Whitney U test. The level of statistical significance was set at a p value of less than 0.05. All statistical comparisons were two sided.

Back to Top | Article Outline


Between September 2009 and March 2013 344 HVAD pumps were implanted in 308 patients. Since June 2012 in all 105 patients visiting the outpatient department acoustic spectra were recorded and analyzed (control group). In eight patients readmitted to hospital with signs of pump thrombosis, pump exchange was performed (pump exchange group). In all pumps, thrombus formation inside the pump was visually confirmed after the pump was removed and disassembled. The acoustic spectra of these patients were compared with those of inconspicuous patients recorded during routine outpatient visits. In all patients in the pump exchange group, the intensity of the first and second harmonic normalized to the fourth harmonic (light gray = pump thrombosis group) is higher than in the control group marked dark gray (first harmonic +75%, p = 0.002 and second harmonic +75%, p = 0.004). Most indicative of pump thrombosis is the existence of a third harmonic as described above, which is not present in the control group (p < 0.0001) as shown in the boxplots (Figure 2). The values obtained immediately prior to pump exchange are shown in Table 1. The increased LDH of more than double normal value (higher than 600 IU) and fHB above 40 mg/dl point toward pump thrombosis according to international standards.18 Elevated power also showed pump thrombosis in all but one patient (Table 1, patient No. 7). In this case, the thrombus was occluding the pump inflow, and the subsequent low flow resulted in lower power consumption until thrombus buildup on the hydrodynamic bearings again elevated the power consumption to overcome friction. Maximal power consumption reached the recommended alarm level (1.5–2 W above mean power consumption) in only five of eight patients. For this reason, we use a tighter alarm threshold of only up to 1 W beyond average power, which results in earlier hints of pump thrombosis.

Back to Top | Article Outline


Pump exchange of HeartWare HVAD is a rare but already routine procedure which nevertheless involves the risk of infection and bleeding complications. Therefore the decision criteria for pump exchange have to be clear and reliable. In our study, in all cases where the pump was exchanged, the presence of pump thrombosis was clearly detected by acoustic analysis of the pump noise. As an independent method, acoustic analysis has the potential to detect pump thrombosis at an early stage prior to the onset of known signs which already reflect secondary damage, either to the patient’s blood or to the pump. Acoustic analysis of the HeartWare HVAD sound is a low-cost method with easy-to-obtain equipment. The method is simple in its application and interpretation and yet provides reliable results.

Taken together, the amplitudes of the first and third harmonic frequency in the acoustic spectra, the motor power increase (mean power consumption before the onset of power increase compared to maximal power before pump exchange), and the hemolysis levels conclusively indicate pump thrombosis. Pump thrombosis was confirmed in all of these pumps by visual inspection of the demounted pump after explantation. Auscultation by human ear cannot be used as a reliable source of information for evaluating pump thrombosis. The position of auscultation strongly influences noise perception of the pump flow and flow pulsatility. Prior to recording the sound produced by the HVAD we determine the location of the best signal quality obtained by normal auscultation, which in general is the location with the highest sound volume. Depending on the orientation of the implanted pump a perpendicular position may yield better signal quality. If in doubt, recordings of both positions—ventral and lateral—should be analyzed. Pump position and patient body size lead to huge variations in pump noise perception, mainly on the absolute sound level. To eliminate these factors, baseline noise recordings are helpful for comparison with actual measurements. The final goal, of course, is to develop a method for identifying thrombi inside the HeartWare HVAD pump without the need for earlier data for comparison. The most distinctive change in the acoustic spectrum of the pump with a thrombus is the presence of a frequency peak in the third harmonic position of the basic frequency of the turning rotor (Figure 3).

Basically the rotational speed is a fixed value kept constant by the controller of the HVAD system. However, if the pre- and afterload conditions vary greatly, as is the case if high pulsatile flow conditions are present, the speed regulation circuit of the controller is not able to maintain a constant speed, though the resulting fluctuations in rotary speed are not displayed on the clinical monitor. Actual speed variations result in shifting of the frequency peaks, resulting in a broadening of the peaks because of the accumulating averaging period. The normal acoustic spectrum consists of several peaks representing the harmonics of the fundamental frequency. Of course the fundamental frequency (also referred to as the first harmonic) of the rotary speed of the turning rotor is apparent in the first peak. The second harmonic (at the double frequency) is recognizable as a second peak.19 The fourth harmonic reflects instationarity of flow inside the pump when the blades, or rather the blood channels of the rotor in the case of the HeartWare “wide-bladed rotor” design, pass the outflow volute. It is created by pressure pulsations and named “blade passing frequency” in the respective literature.20 This peak usually has the highest amplitude.

Normally an imaginary point on the rotor rotates in a circular orbit without any aberrance in radial direction around the central axis. The rotor is centered and balanced. A thrombus on the rotor disturbs the balance. The rotor moves eccentrically around the central axis due to the centrifugal force caused by the additional mass, accordingly raising the amplitude of the fundamental frequency and its harmonic. The increase in these amplitudes is highly significant (first harmonic p = 0.002 and second harmonic p = 0.004).

When the unbalanced part of the rotor passes each of the three paired coils of the stator winding, which represent a symmetry order of three, the electromagnetic force created to turn the rotor also pulls the eccentered part in radial direction toward the center. In this way the rotor oscillates radially three times per revolution, creating a peak at the threefold fundamental frequency (third harmonic, p < 0.0001; Figure 4). No third harmonic sound peak can be expected in a pump free of thrombotic deposits. Thus, this criterion can be used even if no previous measurements have been performed to exclude pump thrombosis.

This hypothesis includes that the excitation of the third harmonic is caused merely by the imbalance of the rotor and is not influenced by changes of the fluid dynamic behavior of the pump. The existence and the amplitude of the third harmonic thus would be independent of the system parameters and the physiologic state of the patient (rotational speed and resulting pump flow determined by preload and afterload). Mock loop simulations with an artificial thrombus created by a blot of silicone glue on one of the quadrants of the rotor demonstrate the independence of the third harmonics from variations of rotary speed and pump flow (Figure 5). The pump was placed into a plastic container lined by foam material to avoid resonance effects and to provide tissue-like dampening of the sound. The foam material divides the container into several sections, one for the pump body, with the inflow penetrating the wall to the second compartment, and a third where the outflow graft is placed. The sections are connected by holes the size of the inflow diameter. Thus, vortices are prevented which could be a source of misleading sound components. To investigate the sensitivity of the method a new thrombus model has to be created with gradual buildup of artificial thrombus material in the pump. Thus the extent of thrombotic material on the rotor could be correlated with gradual changes in the acoustic spectrum.

We speculate that with close monitoring of the acoustic spectrum observation of third harmonics will indicate thrombotic deposit in the HVAD pump before any other clinical or technical signs show changes severe enough that pump thrombosis can be concluded. Two ways may be possible for continued surveillance: developing a simple tool like a smartphone app, so that the patient is able to check for any changes of the emitted sound of the pump by himself or herself. This seems a bit far-fetched because not every patient has a smartphone, will be able to use it in a proper way to produce reliable acoustic data, or will be willing to do so. Also external hardware would have to be coupled to the device (microphone/stethoscope), which would diminish the convenience using a smartphone. The other possibility would be to integrate necessary circuitry and software into the controller of the HVAD using an accelerometer or other sensors integrated in the pump. Technically this should be feasible with the aim of automatically triggering an alarm. The question is whether the manufacturer’s priority in spending money is the detection of pump thrombosis or if they prefer to invest in strategies or design optimization to prevent pump thrombosis, or both.

This low-cost method has now been introduced into our routine outpatient survey protocol. However, this method seems to be limited to systems with nonmechanical bearings, because the most sensitive parameter showing thrombotic deposits on the rotor of the pump, the presence of the third harmonic in the acoustic spectrum, originates from imbalanced rotation. Also the diameter of the rotor may be of importance because the centrifugal forces developed by the imbalance depend on the distance of the additional mass to the center of rotation. Further studies with axial flow pumps featuring other types of bearings may reveal alternative ways of deriving information about pump thrombosis out of their specific acoustic spectrum. The study was limited to the extent that early detection of thrombus formation by the presence of third harmonics could not be proven by visual inspection because pump exchange was not performed until clinical and technical signs clearly indicated pump thrombosis.

Back to Top | Article Outline


Acoustic analysis of the pump noise is a noninvasive method to detect thrombus adhering to the rotor of the HeartWare HVAD. In all pumps exchanged because of acoustically detected thrombosis, the presence of thrombus formation could be confirmed by visual inspection.

Acoustic analysis has the potential to detect pump thrombosis at an early stage prior to the onset of known signs which already reflect secondary damage, either to the patient’s blood (hemolysis) or to the pump itself.

Back to Top | Article Outline


The authors thank Anne Carney and Anne Gale for editorial assistance.

Back to Top | Article Outline


1. Drews T, Dandel M, Krabatsch T, et al. Long-term mechanical circulatory support in 198 patients: Largest single-center experience worldwide. ASAIO J. 2011;57:9–16
2. Kawahito K. Transformation of vibration signals in rotary blood pumps: The diagnostic potential of pump failure. J Artif Organs. 2013;16:393–396
3. Capoccia M, Bowles C, Sabashnikov A, Simon AR. Recurrent early thrombus formation in HeartMate II left ventricular assist device. J Investig Med High Impact Case Rep. 2013;1:1–3
4. Garbade J, Barten MJ, Bittner HB, Mohr FW. Heart transplantation and left ventricular assist device therapy: Two comparable options in end-stage heart failure? Clin Cardiol. 2013;36:378–382
5. Najjar SS, Slaughter MS, Pagani FD, et al.HVAD Bridge to Transplant ADVANCE Trial Investigators. An analysis of pump thrombus events in patients in the HeartWare ADVANCE bridge to transplant and continued access protocol trial. J Heart Lung Transplant. 2014;33:23–34
6. Potapov EV, Stepanenko A, Kaufmann F, et al. Thrombosis and cable damage in the HeartWare pump: Clinical decisions and surgical technique. ASAIO J. 2013;59:37–40
7. Cowger JA, Romano MA, Shah P, et al. Hemolysis: A harbinger of adverse outcome after left ventricular assist device implant. J Heart Lung Transplant. 2014;33:35–43
8. Shah P, Mehta VM, Cowger JA, Aaronson KD, Pagani FD. Diagnosis of hemolysis and device thrombosis with lactate dehydrogenase during left ventricular assist device support. J Heart Lung Transplant. 2014;33:102–104
9. Santise G, Sciacca S, Baglini R, Clemenza F, Pilato M. Can learning to interpret pump messages help lead to an early diagnosis of HeartWare ventricular assist device thrombosis? ASAIO J. 2012;58:629–632
10. Catena E, Milazzo F, Montorsi E, et al. Left ventricular support by axial flow pump: The echocardiographic approach to device malfunction. J Am Soc Echocardiogr. 2005;18:1422
11. Estep JD, Stainback RF, Little SH, Torre G, Zoghbi WA. The role of echocardiography and other imaging modalities in patients with left ventricular assist devices. JACC Cardiovasc Imaging. 2010;3:1049–1064
12. Jennings DL, Cabrera R, Wang DD, Tita C. Successful treatment of a continuous flow left ventricular assist device thrombosis with eptifibatide. ASAIO J. 2012;58:633–635
13. Tang GH, Kim MC, Pinney SP, Anyanwu AC. Failed repeated thrombolysis requiring left ventricular assist device pump exchange. Catheter Cardiovasc Interv. 2013;81:1072–1074
14. Thomas MD, Wood C, Lovett M, Dembo L, O’Driscoll G. Successful treatment of rotary pump thrombus with the glycoprotein IIb/IIIa inhibitor tirofiban. J Heart Lung Transplant. 2008;27:925–927
15. Nakazawa T, Makinouchi K, Takami Y, Glueck J, Tayama E, Nose Y. Vibration assessment for thrombus formation in the centrifugal pump. Artif Organs. 1997;21:318–322
16. Hubbert L, Sundbom P, Loebe M, Peterzen B, Granfeldt H, Ahn H. Acoustic analysis of a mechanical circulatory support. Artif Organs. 2013 [Epub ahead of print]
17. Slaughter MS, Feldman CM, Sobieski MA, et al. Utilization of acoustic signatures to identify HeartMate XVE device end-of-life. J Heart Lung Transplant. 2007;26:579–583
18. Goldstein DJ, John R, Salerno C, et al. Algorithm for the diagnosis and management of suspected pump thrombus. J Heart Lung Transplant. 2013;32:667–670
19. Yuan S, Yang J, Luo Y, Pei J. Experimental Investigation on the Flowinduced Noise under Variable Conditions for Centrifugal Pumps. Chinese J Mech Eng. 2012;25:456–462
20. Birajdar R, Patil R, Khanzode K. Vibration and noise in centrifugal pumps—Sources and diagnosis methods. July 20–24, 2009 Porto/Portugal 3rd International Conference on Integrity, Reliability and Failure

HeartWare HVAD; left ventricular assist device; pump thrombosis; acoustic spectrum

Copyright © 2014 by the American Society for Artificial Internal Organs