CASPER (Cardiac Auscultation Screening and Predictive System)
Client: Group Project
|Role: Fullstack Developer & AI Egnineer
Tech Stack: Python, ONNX, React, Next.js, Express.js, MongoDB (Docker & Atlas), MinIO (Docker), Cloudinary, Tailwind CSS, Docker
Project Overview
CASPER is a web application that enables preliminary cardiac screening through heart sound analysis. Users can upload or record heart auscultations, and the system uses machine learning to detect anomalies such as murmurs or irregular rhythms—providing a risk assessment for further medical consultation.
Problem Statement
Access to cardiologists is limited in rural or underserved areas. CASPER aims to democratize early cardiac screening by turning smartphones or laptops into diagnostic aids, empowering users to monitor heart health proactively.
How It Is Done
The app captures audio via the Web Audio API, preprocesses it (filtering, segmentation), and feeds it into a TensorFlow.js model trained on labeled heart sound datasets. Results are displayed with confidence scores and recommendations. Firebase handles user accounts and history. The UI guides users through proper recording techniques to ensure data quality.