OBJECTIVE: To evaluate the feasibility and limitations of reconstructing ventricular action potential waveforms using non-invasive, unshielded magnetocardiographic mapping (uMCG), highlighting differences between healthy individuals and patients, even at the current level of precision. METHODS: Clinical uMCG was performed using a 36-channel DC-SQUID system. The mathematical reconstruction method developed by Kandori et al. was applied to derive reconstructed ventricular action potential waveforms (rVAPw) from uMCG data in 10 healthy volunteers and 12 patients with various cardiac abnormalities. In four cases, simultaneous recordings of uMCG and right ventricular monophasic action potentials (RVMAP) were obtained using an amagnetic catheter technique. RESULTS: Reconstruction of rVAPw from uMCG signals was feasible in all subjects. Waveforms derived from 90-s averaged uMCG signals were comparable to those obtained with 300-s averages. The rVAPw closely matched the simultaneously recorded RVMAP waveforms. Compared to healthy individuals, patients showed a significant prolongation of rVAPw phase-0 (p < 0.01) and a trend toward increased total duration (p = 0.06), demonstrating the method's sensitivity to electrophysiological abnormalities. CONCLUSIONS: While incomplete rVAPw at some MCG mapping sites reflects the current spatial resolution limitations of the uMCG array, the close alignment between rVAPw and RVMAP recordings suggests that 90-s uMCG acquisitions may suffice for reliable, non-invasive imaging of ventricular action potentials in clinical practice. These findings support further development of MCG technology as a medical device uniquely suited to bridge experimental and clinical applications by enabling non-invasive rVAPw mapping in patients. Future improvements in sensor technology, mathematical modelling, and multimodal imaging may allow for near-cellular spatial resolution.