CortexMD

Home page
Camera page
Health score results
Prevention Tips

FND Detect: Improving Diagnosis with ML and Real-Time Body Analysis

Functional Neurological Disorders (FND) are notoriously difficult to diagnose due to the complexity and overlap of symptoms with other neurological issues. FND Detect is a pioneering solution designed to improve patient outcomes by enhancing diagnostic accuracy and efficiency using machine learning and real-time body analysis.

Problem and Solution

Misdiagnosis in FND patients often leads to delayed or incorrect treatments, impacting patient safety and quality of life. FND Detect solves this by analyzing patient movements via video, using TensorFlow-powered node placement and cosine similarity to analyze key body points. Our app calculates a health score for each video, providing clinicians with objective data to make faster, more accurate diagnoses.

How We Built It

Frontend TypeScript: We use TypeScript for strict type checking, which minimizes data type errors and enhances development efficiency. Bun: This modern package manager is much faster than npm, accelerating the overall development cycle and ensuring smoother performance. Expo (React Native): Expo allows us to develop a high-performance cross-platform application with a single React codebase. The simplicity of React, combined with Expo’s optimization capabilities, means we can focus on creating an intuitive, responsive, and visually appealing user experience.

Backend Flask: This robust Python framework handles video files efficiently, processing video inputs quickly to enable real-time body analysis. SQLAlchemy: SQLAlchemy makes it easy to manage and manipulate our MySQL database directly in Python, without requiring manual database modifications. MySQL: This reliable database is where we securely store patient scores and related data, providing a structured and secure environment for our app’s sensitive information. Core Features Real-Time Body Node Placement: Using TensorFlow, we map nodes across the patient’s body, tracking key points to monitor movement and posture in real time.

Cosine Similarity for Enhanced Accuracy: Our use of cosine similarity ensures that differences in camera angles or perspectives don’t interfere with readings. This technology allows us to provide more precise measurements by focusing on the angles between body nodes, giving a consistent and accurate health score regardless of perspective.

Patient Safety Outcomes: The application’s primary benefit lies in improved patient outcomes. By providing more accurate diagnostic insights, FND Detect helps healthcare providers make informed decisions faster, reducing the risk of misdiagnosis and allowing patients to receive the correct treatment sooner.

Impact

The implications for patient safety and diagnostic accuracy are significant. FND Detect provides healthcare professionals with a critical tool for earlier and more reliable FND diagnosis, ultimately leading to better patient care and outcomes. In a field where time and accuracy are vital, FND Detect stands out as a solution that empowers clinicians to make faster, data-driven decisions.

What’s Next?

We aim to further refine our ML models and explore additional features such as integrating wearable device data and developing tailored recommendations based on the health score. As we expand FND Detect, we’re committed to leveraging technology to enhance patient safety and redefine what’s possible in FND diagnostics.