StrokeChange

Home-based AI monitoring for stroke patients — detecting recovery patterns and flagging new neurological events.

Comment

Below is a Devpost-ready Project Story written in Markdown, structured for clarity and aligned with your StrokeChange research and mobile implementation. You can paste this directly into Devpost.

Inspiration

Stroke recovery does not end when a patient leaves the hospital. Many stroke survivors continue rehabilitation at home, where subtle changes in facial symmetry, motor control, or expression may indicate improvement — or signal a potential secondary event. However, continuous clinical monitoring outside the hospital is limited.

StrokeChange was inspired by the need for an accessible, at-home monitoring solution that leverages AI to detect meaningful neurological changes over time. The goal is to provide patients, caregivers, and clinicians with objective, continuous insight into recovery patterns and early warning signs of additional strokes.

What It Does

StrokeChange is a mobile AI system designed to monitor stroke recovery at home through facial and movement analysis.

Using a smartphone camera, the system analyzes:

  • Facial symmetry and asymmetry
  • Micro-movements and expression changes
  • Recovery progression patterns over time

The app detects deviations from a patient’s established baseline and flags significant changes that may indicate neurological deterioration or a potential additional stroke.

Rather than replacing clinicians, StrokeChange acts as a continuous monitoring layer — extending clinical insight into the home environment.

How We Built It

StrokeChange builds upon computer vision and machine learning techniques for facial feature detection and symmetry analysis.

The system:

  1. Captures video data using a mobile device.
  2. Extracts facial landmarks and geometric features.
  3. Computes symmetry and motion-based metrics.
  4. Uses machine learning models to analyze temporal changes.
  5. Establishes a personalized recovery baseline.
  6. Flags statistically meaningful deviations from prior patterns.

The architecture is designed to be lightweight and mobile-compatible, enabling real-time analysis directly from a smartphone interface.

Challenges We Ran Into

One of the primary challenges was distinguishing between normal variability in recovery and clinically significant change. Stroke recovery is non-linear, and patients may show fluctuations in strength, coordination, and facial control.

Key challenges included:

  • Handling lighting and camera variability in home environments
  • Establishing individualized baselines rather than generic thresholds
  • Reducing false positives while maintaining sensitivity to meaningful change
  • Designing an interface that is usable by elderly patients and caregivers

Balancing clinical sensitivity with usability was central to the system’s design.

What We Learned

We learned that continuous monitoring requires more than detection — it requires contextual understanding over time. Longitudinal analysis is far more informative than single-point assessments.

We also learned that AI-driven health tools must prioritize:

  • Explainability
  • Reliability
  • Accessibility
  • Ethical use of patient data

Designing for real-world deployment means building systems that are robust, interpretable, and user-centered.

What’s Next

Future work includes:

  • Clinical validation studies
  • Integration with telehealth platforms
  • Expansion to include speech and additional motor indicators
  • Enhanced caregiver alert mechanisms

Our long-term vision is to transform smartphones into accessible neurological monitoring tools that empower stroke survivors to recover safely at home while enabling earlier detection of secondary events.

Built With

Share this project:

Updates