Physio: Edge AI Stroke Rehabilitation Sleeve

#Healthcare

#Rehabilitation

Event

LA Hacks

Team

Scott Chiang, Ethan Pham, Majock Bim, Yoda

Physio is an edge AI-enabled stroke rehabilitation sleeve built at LA Hacks 2026 that gives stroke patients clinical-grade movement monitoring at home, with no cloud required. Built by Scott Chiang, Ethan Pham, Majock Bim and Yoda, Physio tackles one of stroke recovery's biggest barriers: long-term hospital rehab sessions are too impractical and expensive for most patients, and existing medical devices are costly.

What it solves

A person in the US has a stroke every 40 seconds. Stroke rehabilitation is proven to help recovery but most patients cannot sustain the frequency of clinic visits needed. Existing home devices lack clinical-grade accuracy and send sensitive patient data to the cloud. Physio gives patients a cheap, private, home-based alternative that never sends personal data anywhere.

What it does

Physio lets patients select rehabilitation movements ranging from isolated range of motion exercises to complex daily living activities. A custom dual-sensor wearable sleeve continuously evaluates movement quality in real time, comparing the patient's trajectory against clinical stroke baselines and outputting a responsive 0 to 100 movement quality score that immediately flags spasticity or compensatory movements. ElevenLabs TTS provides encouraging real-time voice guidance based on current performance, gamifying the path to recovery.

How it works

The sleeve uses an ESP32-C3 microcontroller with two MPU6050 sensors on the bicep and wrist, streaming 80Hz sensor data over Bluetooth Low Energy to an iPhone. A custom PyTorch 1D CNN trained on the JU-IMU clinical dataset achieves 90.5% accuracy at classifying stroke movement quality. The model is compiled and deployed on the iPhone's NPU through ZETIC Melange, achieving sub-millisecond inference with zero cloud dependency and complete patient privacy.