MedBox AI

An intelligent "black box" for medical and dental procedures that uses computer vision to automatically record, segment, and analyze every step of an operation.

▶ PROJECT DATA LOADED...

Introduction

MedBox AI is an intelligent "black box" system designed for medical and dental procedures. Inspired by aviation black boxes, it provides a transparent, objective system to capture and analyze procedural data, enhancing accountability and supporting quality assurance in healthcare.

Project Overview

MedBox AI uses computer vision to automatically record, segment, and analyze procedure videos in real-time. It detects key surgical or dental steps, identifies tool usage, and flags deviations from standard protocols. The result is a secure, timestamped record that serves as a reliable reference for clinicians, insurers, and patients.

Demo Video

Key Features

  • Automated Segmentation: Uses computer vision to detect and segment key steps in medical procedures.
  • Tool Detection: Identifies specific surgical and dental tools used during operations.
  • Deviation Flagging: Automatically flags deviations from standard medical protocols.
  • Secure Logging: Creates a timestamped, immutable record for medico-legal documentation.
  • Auditable Dashboard: Provides a visual timeline of events for easy review by clinicians.

How It Built It

We integrated OpenCV computer vision models for step segmentation and tool detection with Chroma for vector storage and retrieval of procedural data. The frontend is built with Next.js and Shadcn/UI, providing a clean dashboard for reviewing annotated procedures. The backend manages real-time inference and report generation.

Technology Stack

  • Frontend: Next.js, Shadcn/UI, Tailwind CSS
  • Backend: Python, Flask
  • AI/CV: OpenCV, Chroma (Vector DB)
  • Video: NextVideo

Challenges & Learnings

Obtaining clean, annotated datasets for specific surgical tools was a significant challenge. We also had to balance the complexity of medical workflows with the need for an intuitive user interface. We learned that effective medico-legal documentation requires not just AI accuracy, but absolute data integrity.