Can AI Detect Improper Exercise Form During Rehab?

How AI-Based Form Detection Actually Works

Some AI-powered rehabilitation tools use computer vision, often through a smartphone or webcam camera, to track key points on a person’s body during exercise, effectively mapping a simplified skeleton-like model of the person’s movement. This tracked movement is then compared against an expected pattern for the specific exercise being performed, allowing the system to flag deviations that might indicate improper form, such as an incomplete range of motion, misalignment, or asymmetry between limbs. Other approaches rely on wearable sensor data, such as from accelerometers or gyroscopes, rather than camera-based tracking, to achieve a similar goal of comparing actual movement against an expected reference pattern.

This general approach — comparing observed movement data against a defined reference pattern — allows these systems to provide a degree of real-time or near-real-time feedback to a patient performing rehabilitation exercises, which can be genuinely useful for reinforcing correct technique.

Best Suited to Well-Defined, Structured Movements

AI form-detection tools tend to perform best on exercises with clearly defined, relatively consistent expected movement patterns, since this gives the system a well-characterized reference to compare against. More complex, nuanced, or highly individualized movement issues — the kind of subtle compensations or asymmetries that a trained physical therapist might notice through years of hands-on clinical experience — can be more challenging for current AI systems to reliably detect and interpret correctly. This means the real-world usefulness of AI form-checking varies considerably depending on the specific exercise and the complexity of what’s actually being assessed, rather than being uniformly reliable across every possible movement or condition.

A Support Tool Between Supervised Sessions

Given these capabilities and limitations, AI-based form-detection tools are generally positioned as a way to support and reinforce proper technique during independent practice between scheduled, supervised physical therapy sessions, rather than as a full substitute for direct, in-person correction and guidance from a licensed physical therapist. This kind of continuous, at-home reinforcement can genuinely help patients maintain better technique consistency throughout their rehabilitation, complementing rather than replacing the more nuanced, hands-on assessment and correction that in-person sessions provide, particularly for more complex conditions or movement patterns.

Bottom Line

Some AI-powered tools can detect improper exercise form during rehabilitation using computer vision or wearable sensors to compare movement against expected patterns, offering useful real-time feedback for well-defined exercises, but current capabilities have real limitations and generally work best as a support tool between supervised sessions rather than a full substitute for a physical therapist’s in-person assessment.

Go deeper

• How is ai used in robotic exoskeletons for physical rehabilitation?