Brain MRI Scans Classification

A brain tumor MRI image classifier built with Teachable Machine (transfer learning) and deployed as a TensorFlow.js web app for in-browser inference.
This project was created for the AI course and focuses on dataset preparation, model training/validation, and lightweight deployment on the web.

Demo

https://diegomalagrida.me/mri-tumor-classifier/app/

About

The project trains an image classification model to recognize 4 classes from brain MRI scans:

• Glioma
• Meningioma
• Pituitary
• No tumor

The workflow includes dataset balancing, image resizing/normalization, hyperparameter exploration (epochs, batch size, learning rate), and evaluation on an external test set.
The final model is exported to TensorFlow.js and integrated into a simple web interface that displays class probabilities and supports a confidence threshold for uncertain predictions.

Executing the project

If you want to run the project, you only need a modern web browser and a simple static server.

To run it locally:

1. Clone the repository.
2. Start a local static server (e.g., VS Code Live Server or python -m http.server).
3. Open the app in your browser and:
   • Upload an MRI image, or
   • Pick one from the sample library (if included).

To avoid CORS issues, do not open the HTML file directly. Always use a local server when loading the TensorFlow.js model files.

Authors

• Diego Malagrida González

• Andreu Massanet Felix