Prediction of ventricular arrhythmias (VA) mostly relies on left ventricular ejection fraction (LVEF), but with limited performance. New echocardiographic parameters such as mechanical dispersion have emerged, but acoustic window sometimes precludes this measurement. Nuclear imaging may be an alternative. We aimed to assess the ability of mechanical dispersion, measured with phase standard deviation (PSD) on radionuclide angiocardiography (RNA), to predict VAs. This retrospective monocentric observational study included all patients who underwent a tomographic RNA from 2015 to 2019. Phase analysis yielded PSD and follow-up was examined to identify VAs, heart transplantation, and death. The study population consisted of 937 patients, mainly with LVEF ≤ 35% (425, 45%). Most had ischemic (334, 36%) or dilated cardiomyopathies (245, 26%). We identified 86 (9%) VAs. PSD was strongly associated with the occurrence of VA hazard ratio per 10 ms increase (HR10) 1.12 (1.09-1.16), heart transplantation HR10 1.09 (1.06-1.12), and death HR10 1.03 (1.00-1.05). The association between PSD and VA persisted after adjustment for age, sex, QRS duration, LVEF, global longitudinal strain (GLS), and echocardiography-assessed mechanical dispersion. The occurrence of ventricular arrhythmias was predicted by mechanical dispersion assessed by RNA, even after adjustment for LVEF and GLS.