Derivation of in vivo pressure-volume loops for post-heart transplant patients using real-time 3D echocardiography and left ventricular catheterisation

Abstract Background Image-based methods that combine catheterisation with non-concurrent cardiac magnetic resonance (CMR) imaging data and echocardiography (echo) is gaining more interest than the conductance catheter method to derive pressure-volume loops (PVLs) due to improved accuracy and accessibility of LV volume quantification [1–3]. However, accurate temporal registration between LV pressure and volume is not well developed. Purpose We propose a framework for temporally registering invasive LV and aortic pressures (LVP and AOP) acquired during left heart catheterisation with real time 3D echocardiography (RT3DE) to generate in vivo PVLs in a group of heart transplant (HTx) recipients. Methods 25 orthotopic HTx recipients (mean age: 54±8 years and 7 female) indicated for routine coronary assessment were recruited for invasive hemodynamic measurement and RT3DE imaging. A fluid-filled pigtail catheter was used to measure LVP and AOP with simultaneous electrocardiogram (ECG) over several (9–15) heartbeats. Within an hour of catheterisation, single-beat transthoracic RT3DE of the LV was performed from the apical window in a left lateral decubitus position. Imaging parameters were optimized for each patient to maximize the temporal resolution (between 15–41 imaging frames per cycle). We developed a piecewise linear temporal scaling method based on cardiac events (end-diastole (ED), end of isovolumic contraction (eIVC), end-systole (ES), end of isovolumic relaxation (eIVR), and diastasis (DS)) of RT3DE and haemodynamic measurement to resample the LVP at the RT3DE imaging frames between the cardiac events to construct PVLs (Fig. 1a). Geometric LV models were manually fitted at ED and ES, followed by automatic tracking across intermediary frames to estimate LV volume over the entire cardiac cycle (Fig. 1b). The temporally aligned pressure values were further averaged to find the beat-averaged LV PVL (Fig. 1c). Results Based on the number of cardiac cycles selected for haemodynamic analysis, multiple in vivo PVLs were constructed for each patient. A beat-averaged PVL was also computed for each patient (Fig.1d). With the exception of one case, the beat-averaged PVLs exhibited classically representative shape with distinct isovolumic contraction and isovolumic relaxation phases. The individual diastolic PVRs for all patients are shown in Fig.1e, with beat-to-beat variation observed in most patients. For some cases, the variation manifested as an offset in LVP,