Integrated super-resolution CMR-deep learning to deconvolute passive and active causes of impaired relaxation - SUMMARY Left ventricular (LV) diastolic dysfunction (LVDD) is a key contributor to the pathophysiology of several cardiac diseases, including diabetic cardiomyopathy, hypertrophic cardiomyopathy, and heart failure with preserved ejection fraction. LVDD has two common sequelae: restrictive filling and impaired LV relaxation (ILVR). The restrictive filling is caused by the passive stiffening of the LV, generally linked to fibrotic and hypertrophic events in the LV free wall myocardium (LVFW). ILVR, on the other hand, is thought to be driven by alterations in cross- bridge cycling kinetics impairing active force development and relaxation in the LVFW. However, the role of passive myocardial alterations in ILVR remains unknown, while its understanding is central to the treatment of ILVR as passive and active causes of ILVR demand different therapeutic choices. Moreover, the current characterization of ILVR remains limited to echocardiography and hemodynamic indices, such as mitral annular relaxation velocity and diastolic time constant. These indices quantify only the “net” manifestation of underlying active and passive causes of ILVR at the organ level, falling short of suitable measures to distinguish cellular- and fiber-level mechanisms, collectively, leading to ILVR. As such, a feasible technology that can segregate and quantify contributions of passive and active causes of ILVR in vivo remains elusive, while it is essential for simplifying the etiologic heterogeneity of LVDD. Our central hypothesis is that passive and active remodeling events are independent, but competing, contributors to ILVR, and integrated super-resolution cardiac magnetic resonance imaging-deep learning (CMR- DL) can isolate and quantify the contribution of each event to LVDD in diseased and treated subjects in vivo. We propose to test our hypotheses via three specific aims: 1. Determine contributions of passive and active remodeling events to LVFW slow relaxation in mice ex vivo. We will identify and separate the contributions of passive and active remodeling events to LVFW slow relaxation in diabetic murine models of LVDD and newly generated knock-in mice models. 2. Predict passive and active contributions to ILVR using CMR-DL in mice in vivo. We will use subject-specific inverse models integrated with DL to estimate the contributions of passive and active causes of LV impaired relaxation in vivo. 3. Dissect LVFW passive and active contributions to ILVR using common clinical metrics in humans. We will estimate the independent contributions of active and passive forces to echocardiography indices of the ILVR in four phenotypes of retrospective LVDD patients. The completion of these studies will: 1) determine passive fiber-level alterations in the LVFW that would be most deteriorating to LV relaxation, and 2) advance individualized treatment of LVDD based on identified contributors to ILVR and diastolic dysfunction through integrated CMR-DL.
