BIOMEDICAL AI • MATLAB

Blood Clot Detector

MATLAB-based CNN model with GUI for detecting blood clotting disorders.
Focused on early detection using thermal imaging datasets and an interactive medical interface.

Overview

This project developed a Computer-Aided Detection (CAD) system for identifying potential blood clotting disorders. It was implemented in MATLAB, combining deep learning with a graphical user interface (GUI) to allow clinicians to upload thermal or diagnostic images, process them through a CNN, and receive a prediction in real time.

Technical Highlights

Results

The CNN achieved high classification accuracy on the testing dataset, successfully distinguishing clot-affected images from healthy ones. The GUI integration enabled smooth operation, providing clinicians a non-technical way to run predictions. This made the model usable outside of traditional machine learning environments.

Challenges & Solutions

One major challenge was working with thermal medical datasets, which were relatively small and imbalanced. I used augmentation and careful normalization to expand data diversity. MATLAB’s limited real-time GPU support was addressed by optimizing network size and using efficient layer pruning, allowing faster inference for clinical use cases.