Physiobook

An AI-powered physiotherapy platform focused revolutionizing how patients access and manage their rehabilitation care.

Comment

Inspiration

Physiotherapy clinics struggle with inefficient patient intake, incorrect therapist matching, high no-show rates, and limited visibility into patient recovery progress. Most booking systems treat physiotherapy like a generic calendar problem, ignoring clinical context and operational intelligence. This leads to wasted clinic time, poor patient experience, and suboptimal recovery outcomes.

What it does

PhysioBook is an AI-powered physiotherapy clinic platform that goes beyond scheduling. It intelligently triages patients, matches them with the most suitable physiotherapist, and provides real-time recovery and clinic performance dashboards. By combining structured AI decision support with a modern booking system, PhysioBook helps clinics operate more efficiently while improving patient outcomes

How we built it

We built PhysioBook using Next.js 14 with the App Router for a modern full-stack architecture. The frontend uses React with Tailwind CSS and Framer Motion for smooth animations. For AI capabilities, we integrated OpenAI's GPT-4 for symptom triage and TensorFlow.js with MoveNet for real-time pose detection during exercises. Clerk handles authentication, Prisma manages our PostgreSQL database, and we used WebRTC for tele-physio video calls.

Challenges we ran into

  • Real-time pose detection accuracy: Calibrating TensorFlow.js MoveNet to accurately score exercise form across different body types and camera angles required extensive testing.

  • AI triage reliability: Balancing between being helpful and not providing medical diagnoses—we implemented confidence scores and always recommend professional consultation.

  • Database schema complexity: Designing relationships between patients, therapists, appointments, and progress records while maintaining HIPAA-compliant data structures.

Accomplishments that we're proud of

  • 7 AI-powered features working together seamlessly: triage, matching, exercise guidance, progress prediction, no-show prediction, voice agent, and smart scheduling.

    • 94% triage accuracy in our testing with proper red flag detection for urgent cases.
    • Real-time exercise form scoring that provides instant feedback without requiring expensive hardware. Complete role-based system with distinct experiences for patients, therapists, and administrators.

What we learned

  • How to effectively combine multiple AI models (LLMs + computer vision) in a single healthcare application.

  • The importance of user experience in healthcare apps—reducing friction increases patient compliance.

  • Balancing AI automation with human oversight in medical contexts.

  • WebRTC implementation challenges for reliable video consultations.

  • Designing accessible interfaces that work for users of all technical skill levels.

What's next for Physiobook

  • Mobile apps (iOS/Android) with offline exercise tracking.

  • Wearable integration (Apple Watch, Fitbit) for automatic activity logging.

  • Insurance provider integrations for seamless billing and coverage verification.

  • Multi-language support to serve global patient populations.

  • Advanced AI models trained on physiotherapy-specific datasets for even better recommendations.

  • Clinical trials to validate outcomes and pursue FDA clearance.

Built With

Share this project:

Updates