We have recently demonstrated that accelerometer-based pump thrombosis and thromboembolic events detection is feasible in vitro. This article focuses on detection of these conditions in vivo. In an open-chest porcine model (n = 7), an accelerometer was attached to the pump casing of an implanted HeartWare HVAD. Pump vibration was analyzed by Fast Fourier Transform of the accelerometer signals, and the spectrogram third harmonic amplitude quantified and compared with pump power. Interventions included injection of thrombi into the left atrium (sized 0.3-0.4 ml, total n = 35) and control interventions; pump speed change, graft obstruction, and saline bolus injections (total n = 47). Graft flow to cardiac output ratio was used to estimate the expected number of thrombi passing through the pump. Sensitivity/specificity was assessed by receiver operating characteristic curve. Graft flow to cardiac output ratio averaged 66%. Twenty-six of 35 (74%) thrombi caused notable accelerometer signal change. Accelerometer third harmonic amplitude was significantly increased in thromboembolic interventions compared with control interventions, 64.5 (interquartile range [IQR]: 18.8-107.1) and 5.45 (IQR: 4.2-6.6), respectively (p0.01). The corresponding difference in pump power was 3 W (IQR: 2.9-3.3) and 2.8 W (IQR: 2.4-2.9), respectively (p0.01). Sensitivity/specificity of the accelerometer and pump power to detect thromboembolic events was 0.74/1.00 (area under the curve [AUC]: 0.956) and 0.40/1.00 (AUC: 0.759), respectively. Persistent high third harmonic amplitude was evident at end of all experiments, and pump thrombosis was confirmed by visual inspection. The findings demonstrate that accelerometer-based detection of thromboembolic events and pump thrombosis is feasible in vivo and that the method is superior to detection based on pump power.
