Monitoring the activities of athletes with lower-limb amputations who use running-specific prostheses is essential for evaluating their training regimes, as well as the effectiveness and mechanical fatigue wear of their prostheses over time. Recent advancements in Inertial Measurement Units (IMUs) and activity detection algorithms offer new opportunities for objective assessment, but their application in Paralympic sports remains unexplored. The aims of this work were to design and implement an innovative protocol and analytical software for short-term and long-term activity detection of athletes with transtibial and transfemoral amputation and then test its validity on a sample of elite Paralympic runners and triathletes. Overall, the ability of the model to detect activities presented an accuracy of 98%, and the error in the stride counting for all activities fell within a 1% margin.