Early Pump Thrombosis Detection Algorithm in Patients Implanted with Heartware Pump

Abstract

Purpose Continuous-flow left ventricular assist devices (CF-LVADs) have been extensively used in a strategy of bridge to orthotopic heart transplant due to increasing number of patients with end-stage heart failure and worldwide shortage of heart donors. However, usage of CF-LVADs is not free from limitations such as device-related adverse events, including pump thrombosis (PT). The purpose of this study was to develop an early PT detection algorithm based on the basic parameters monitored by the implanted device. Methods We analyzed the power consumption recorded by pump controller, normalized with respect to impeller speed. The signal was transformed to Open-High-Low-Close form with the averaging window of 4 hours, which enabled calculation of Typical Price (TP) technical indicator. We then developed an algorithm that learns patient s usual TP values over two consecutive days. Later such an algorithm was able to analyze each subsequent data point and indicate whether the probability of thrombotic event meets predetermined threshold. Results We retrospectively reviewed log files of 101 patients implanted with HeartWare pump (HVAD). Seventeen had suspected or confirmed PT (21 events) according to INTERMACS definitions. Our algorithm classified correctly all cases with 100% sensitivity and 100% specificity. Based on abnormally high power consumption (Fig 1A) and TP evaluation, the algorithm alerted at least 40% probability of pump thrombosis (Fig 1B) preceding the onset of PT by 5.4 days on average. In the worst-case scenario, the threshold was reached 16 hours prior to the event. Conclusion Our study shows that it is possible to predict PT events with use of a dedicated, custom-made algorithm. Automatic event detection and alarming could enable earlier medical management of PT and improve prognosis in patients implanted with HVAD.