From episodic therapy to continuous rehabilitation

01The between-sessions problem

Rehabilitation is one of the most data-poor corners of medicine. A therapist sees a patient for an hour, once or twice a week; the other hundred-odd waking hours — where the recovery actually happens or stalls — go unrecorded. Adherence is guessed at, plateaus are noticed late, and the exercises done wrong at home quietly undo the ones done right in clinic.

AI could transform recovery — personalizing protocols, predicting plateaus, catching regression early — but only if it can see the patient between sessions. Trained on twice-weekly snapshots, a model inherits the same blind spot the clinician has.

Recovery happens in the hours between appointments — exactly where no instrument is watching.

02Therapy that comes home, in a garment

Textile interfaces may extend selected sensing and stimulation functions beyond supervised clinic sessions, provided the garment, control system, protocol, and intended use are validated together. Functional electrical stimulation garments for recovery after stroke and spinal cord injury have been developed with rehabilitation clinicians — delivering stimulation from fabric rather than a rigid device strapped on for a single session.

Under defined laboratory protocols, the tested textile stimulation electrodes matched the comparison condition, remained functional after 30 wash cycles and 1,000 stretch cycles at 50% of break strain, and delivered comfortable responses for at least six hours of continuous wear. Laboratory durability is not the same as general real-world durability, but it is what makes wearing, washing, and re-wearing a therapy plausible.

03From monitoring to a closed loop

A future closed-loop rehabilitation system would coordinate sensing and stimulation. In one closed-loop neuromuscular stimulation study — run with the Feinstein Institutes for Medical Research — feedforward–feedback control with textile electrodes regulated individual finger grasp force in one participant with quadriplegia to under 15% steady-state error, with 0.67-second settling.

The controller adjusted stimulation in response to measured force within that experimental configuration. Translating the approach into unsupervised home use would require additional clinical, usability, safety, and regulatory validation.

04Built with clinicians

Rehabilitation technology fails when it ignores the people who use it. MyantX's FES-garment work was shaped by a qualitative study of nineteen patients and clinicians defining what a wearable-FES garment must do to be viable, and a standing-balance intervention was co-designed directly with spinal-cord-injury and stroke end-users alongside rehabilitation researchers, including the KITE Research Institute at Toronto Rehab (University Health Network).

Co-design helps identify practical requirements that signal performance alone cannot reveal, including comfort, donning, adherence, setup, supervision, and integration into clinical workflows.

05What AI does with a continuous recovery signal

Give a model the full arc of recovery — adherence, muscle activation, gait, balance, day after day — and rehabilitation stops being reactive. Protocols adapt to how a specific patient is actually responding. Plateaus and regressions surface early enough to change course. Remote and hospital-at-home programs get a signal they can act on, not a self-report.

The bridge to that future is not another clinic device. It is the garment itself, sensing and stimulating in a loop — a human-recovery data layer that adaptive rehabilitation AI could run on.