We evaluated the performance of statistical shape models (SSMs) derived from short-axis (SAX) cardiac magnetic resonance imaging (CMR) to estimate pulmonary capillary wedge pressure (PCWP) in a cohort of 68 cases with heart failure with preserved ejection fraction (HFpEF) and dyspnea. We obtained a 5% improvement in performance (R2 = 0.36) using 3 principal component analysis (PCA) modes compared to the existing state-of-theart non-invasive method (R2 = 0.31), which uses a linear combination of left ventricular (LV) mass and left-atrial (LA) volume to estimate PCWP. We also evaluated the linear formula on our dataset and noticed that inconsistent LV mass-PCWP correlations limited predictive performance while the correlation between PCWP and LA volume calculated from 3D meshes was on par (R2 = 0.30). Furthermore, manually calculated LA volume using the voxel-calculation method had the highest individual predictive performance (R2 = 0.33) followed by PCA mode 1 (R2 = 0.32). These findings underscore the potential of SSMs derived from CMR for non-invasive PCWP estimation in HFpEF patients, suggesting avenues for further improvement and clinical application.