BACKGROUND:Wearable inertial sensors are increasingly used to monitor swimming performance; however, evidence regarding their validity remains limited and inconsistent, particularly in Para swimmers, whose alterations in coordination, symmetry and propulsion may challenge head-mounted inertial event detection. This study aimed to provide a preliminary evaluation of the validity and agreement of the Triton 2 Unit system (triaxial accelerometer, gyroscope and magnetometer) against video-based analysis in swimmers with and without disabilities. METHODS:Four male and two female competitive swimmers (three without disabilities and three Paralympic swimmers) performed a structured training protocol under controlled conditions. Performance metrics derived from the head-worn inertial sensor were compared with a video-based reference. Variables analysed included lap time, mean stroke rate, number of strokes, underwater time, and turn time. Agreement and accuracy were assessed using median bias, interquartile range (IQR), median absolute percentage error (MdAPE), Lin`s concordance correlation coefficient (*c), and non-parametric 95% limits of agreement. RESULTS:In swimmers without disabilities, Triton showed high agreement for lap time (*c = 0.968), stroke rate (*c = 0.962), and number of strokes (*c = 0.955), with low relative errors. Moderate agreement was observed for underwater time (*c = 0.829), while turn time showed poor agreement (*c = 0.228). In Para swimmers, agreement was generally lower and more variable, with wider limits of agreement across all variables. Lap time (*c = 0.869) and stroke rate (*c = 0.822) showed the highest consistency, whereas underwater time, number of strokes, and turn time showed reduced agreement. In one swimmer with severe impairment (S5), the device failed to correctly identify lap structure, leading to invalid outputs. CONCLUSIONS: Triton 2 Unit provides acceptable accuracy for selected metrics in swimmers without disabilities under ecological conditions. In Para swimmers, increased variability and occasional detection failures limit its applicability, particularly for event-based variables and severe impairments.
© Copyright 2026 XVth International Symposium on Biomechanics and Medicine in Swimming Proceedings. All rights reserved.