Left ventricular assist devices (LVADs) can significantly improve survival rate and quality of life for patients suffering from end-stage heart failure. Several promising strategies to control LVADs are being developed, some being focused on the end-diastolic pressure (EDP). For those, the problem of EDP estimation in real-time has to be solved. In this work, a deconvolution-based method to identify features in cardiac signals is presented. This method is applied to the estimation of the EDP from the left-ventricular pressure (LVP) signal and evaluated on animal trial data. In 11 trials with adult sheep, a myocardial infarction was induced and an LVAD was implanted. A total of 37.6 hours of LVP data was annotated by a medical expert. Compared to the annotations, a root mean square error of 11.6 ms / 4.1 mmHg was achieved using the proposed deconvolution method.
IntroductionAccording to the American Heart Association, about 5.7 million Americans live with heart failure. For patients suffering from its end-stage manifestation, left ventricular assist devices (LVADs) can significantly improve survival rate and quality of life [1]. For the control of LVADs, several strategies exist [2], one promising approach being the control of the end diastolic pressure (EDP). EDP is defined as the end-diastolic (ED) value of the left ventricular pressure (LVP). Thus, to implement an EDP control strategy, its value has to be determined in real time.In [3], the estimation of the ED time point was approach by an analysis of the peak curvature of the LVP signal and evaluated in terms of temporal accuracy. Here, a deconvolution-based approach is presented and its accuracy is evaluated both in temporal as well as amplitudinal accuracy. In previous work we have demonstrated that deconvolution methods have a great potential in the processing of (multimodal) cardiac signals. It was shown in [4] that blind deconvolution can be used to estimate a virtual source signal and linear filter coefficients to analyze and represent measured multimodal signals (e.g. photoplethysmography and ballistocardiography). In [5] it was further shown that a desired, measured signal (e.g. reference ECG) can be approximated by filtering measured multichannel signals (e.g. capacitively coupled ECG) with estimated linear filters. In this work, we expand the concept of blind deconvolution towards the estimation of the EDP. In particular, a measured signal and a desired, virtual signal are used to estimated linear filter coefficients in a training phase. These filter coefficients can subsequently be used to efficiently locate ED time points via convolution and peak detection.
Deconvolution AlgorithmThe aim of the proposed algorithm is to find a feature, namely the ED time point, in the LVP signal. Thus, the observed signal x = x(t), with t ∈ 0 . . . T − 1, is the LVP signal, see Figure 1, top graph, solid line. The desired signal y = y(t) is derived from the temporal location of the ED time points. In particular, a medical expert has identified all points i...