Brain Tumor MRI Classifier
A computer vision web app that classifies brain MRI images into glioma tumor, meningioma tumor, pituitary tumor, or no tumor using a trained deep learning model.
Project Overview
An educational Streamlit application that allows users to upload a brain MRI image and receive a model prediction with confidence scores. The app preprocesses the uploaded MRI image, optionally crops the visible brain region using OpenCV, resizes the image to the model input size, and uses a saved TensorFlow/Keras model to classify the scan into one of four tumor-related categories.
The Problem
Brain MRI interpretation requires medical expertise, and beginners in medical AI often need practical tools to understand how image classification models process scans and produce predictions. A clear, interactive demo can make medical image classification easier to explore while emphasizing responsible limitations.
The Solution
Built a Streamlit-based prediction pipeline that loads a trained Keras model, accepts JPG/JPEG/PNG MRI uploads, converts images into RGB format, optionally crops the visible brain region, resizes the image to 224x224, runs model inference, and displays the predicted class with confidence scores and class probabilities.
The Impact
Created a medical computer vision demo that transforms a trained notebook model into an interactive web application, showing end-to-end skills in image preprocessing, model saving, Streamlit deployment, and responsible AI communication for healthcare-related applications.
Date / Year
2026