Detecting Suspected Pump Thrombosis in Left Ventricular Assist Devices via Acoustic Analysis

• Published in: IEEE journal of biomedical and health informatics 2020-07, Vol.24 (7), p.1899-1906
• Main Authors: Semiz, Beren, Hersek, Sinan, Pouyan, Maziyar Baran, Partida, Cynthia, Blazquez-Arroyo, Leticia, Selby, Van, Wieselthaler, Georg, Rehg, James M., Klein, Liviu, Inan, Omer T.
• Format: Article
• Language: English
• Subjects: Abnormalities; Acoustics; Aged; Algorithms; Biomarkers; Blood; Centrifugal pumps; Correlation; Correlation analysis; Feature extraction; Female; Harmonic analysis; Heart; Heart failure; Heart Sounds - physiology; Heart-Assist Devices - adverse effects; Humans; Learning algorithms; Left ventricular assist device; Machine learning; Male; Middle Aged; Patients; pump thrombosis; Pumps; Regression analysis; Regression models; Remote monitoring; Signal Processing, Computer-Assisted; Sound Spectrography; Stethoscopes; Thromboembolism; Thrombolysis; Thrombosis; Thrombosis - diagnosis; Ventricle; Ventricular assist devices
• ISSN: 2168-2194; 2168-2208
• DOI: 10.1109/JBHI.2020.2966178

Objective: Left ventricular assist devices (LVADs) fail in up to 10% of patients due to the development of pump thrombosis. Remote monitoring of patients with LVADs can enable early detection and, subsequently, treatment and prevention of pump thrombosis. We assessed whether acoustical signals measured on the chest of patients with LVADs, combined with machine learning algorithms, can be used for detecting pump thrombosis. Methods: 13 centrifugal pump (HVAD) recipients were enrolled in the study. When hospitalized for suspected pump thrombosis, clinical data and acoustical recordings were obtained at admission, prior to and after administration of thrombolytic therapy, and every 24 hours until laboratory and pump parameters normalized. First, we selected the most important features among our feature set using LDH-based correlation analysis. Then using these features, we trained a logistic regression model and determined our decision threshold to differentiate between thrombosis and non-thrombosis episodes. Results: Accuracy, sensitivity and precision were calculated to be 88.9%, 90.9% and 83.3%, respectively. When tested on the post-thrombolysis data, our algorithm suggested possible pump abnormalities that were not identified by the reference pump power or biomarker abnormalities. Significance: We showed that the acoustical signatures of LVADs can be an index of mechanical deterioration and, when combined with machine learning algorithms, provide clinical decision support regarding the presence of pump thrombosis.