The end-systolic pressure-volume relationship (ESPVR) and end-diastolic pressure-volume relationship (EDPVR) are the essential load-independent measures of the heart (not influenced by preload or afterload of the heart), and it allows for a more in-depth understanding of cardiac function. Current ESPVR and EDPVR measurements are invasive; thus, their utility in clinical practice is limited. Recently, imaging modalities have developed non-invasive procedures to predict ESPVR and EDPVR. Still, they share the major disadvantage of requiring a trained operator and expensive capital equipment and also, these methods are not accurate under certain diseased conditions. To overcome the limitations mentioned above, we will use machine learning to estimate the ESPVR and EDPVR of the left ventricle using unintrusive, readily available standard clinical measurements. However, an extensive database of healthy and diseased subjects and their respective measurements are necessary to train the machine learning algorithm. We tackle these problems with a virtual database approach where training data is generated using a computational model of the cardiovascular system. This is the first unified approach to predict both ESPVR and EDPVR of the left ventricle non-invasively in the clinical setting.