Abstract 039: Hybrid Intrinsic Frequency Machine Learning Approach For Calculation Of Total Arterial Compliance And Aortic Characteristic Impedance From A Single Carotid Waveform In Heart Failure With Preserved Ejection Fraction

Abstract

Introduction: Previous studies demonstrated that heat failure with preserved ejection fraction (HFpEF) is characterized by abnormal arterial function and left ventricle (LV)-arterial coupling. Aortic characteristic impedance (Zc) and total arterial compliance (TAC) are well-established biomarkers for quantification of the arterial function. However, the routine clinical use of these biomarkers requires both pressure and flow measurements. Intrinsic Frequency (IF) method is a novel system-based mathematical approach for the analysis of the cardiovascular dynamics that can predict HF events (Cooper et al. Hypertension, 2021 PMID: 33390053). We hypothesized that a hybrid IF-machine learning (ML) approach can compute TAC and Zc from a noninvasive carotid waveform. Method: We used data from the Framingham Heart Study (FHS) (n=6168). The reference values of Zc and TAC were computed from carotid pressure and aortic flow waveforms. IF parameters of carotid pressure waveform were used in a fully connected feed-forward neural network model. The model was trained/validated/tested on 80% of randomly selected data from all FHS data. The final model was tested on the remaining 20% of the data blinded to all stages of ML development. Result: In blind test set data, our IF-ML models showed correlations of r=0.89 and r=0.79 with the reference value of TAC and Zc, respectively (Fig 1). More importantly, the TAC and Zc from our IF-ML models showed strong correlations among HFpEF patients (r=0.9 and r=0.77 respectively) (Fig.1) Conclusion: Since IFs can be measured noninvasively (e.g. an iPhone), this method can be used to facilitate routine and noninvasive evaluation of TAC and Zc in HFpEF.