Does artificial intelligence feedback result in different kinematic and muscle excitation patterns compared to...

Lower-limb rehabilitation exercises often require supervised feedback to ensure correct technique and muscle engagement. Artificial intelligence systems could provide an alternative to physiotherapy supervision, offering real-time feedback. This study aimed to compare effects of artificial intelligence-based feedback with physiotherapy feedback on kinematic and electromyographic outcomes during lower-limb exercises in healthy participants.

A repeated-measures design was employed, with uninjured participants performing four hip and knee exercises under physiotherapy and artificial intelligence (Merlin Ltd) feedback conditions. Kinematic data of the hip and knee were collected using an active infrared marker system. Muscle excitation was measured using surface electromyography for seven lower-limb muscles. Paired t-tests and two one-sided t-tests were used to assess differences and equivalence between conditions.

Among 11 participants (45 % females, mean age 24.1 years ±4.1, height 173.0 cm ± 9.2, mass 69.3 kg ± 13.7, Tegner Activity Scale 5.7 ± 1.4) no consistent significant differences were observed between physiotherapy and artificial intelligence feedback across exercises for kinematic and electromyographic outcomes. Equivalence in range of motion was observed for 58 % of all hip angles and 67 % of all knee angles; however, significant variability existed for minimum and maximum joint angles. Peak and root mean square amplitudes were mostly non-equivalent between conditions.

While artificial intelligence feedback demonstrated potential for guiding rehabilitation exercises, it lacked consistency with physiotherapy feedback for certain electromyographic and kinematic parameters due to limitations in evaluating multi-planar movements. Despite these limitations, artificial intelligence could serve as a supplementary tool, enhancing adherence and technique between physiotherapy sessions.