A Hybrid Machine Learning Approach using MRI Scans
This project implements a hybrid interpretable machine learning pipeline for brain tumor classification and severity detection using MRI images. Unlike deep learning “black-box” models, this system focuses on feature-engineered, explainable, and computationally efficient diagnosis — making it suitable for real clinical decision support.
The system performs two tasks:
Tumor Classification → Glioma, Meningioma, Pituitary, No Tumor Severity Detection → Low, Mild, High, No Tumor
👉 BRISC2025 MRI Dataset
Contains MRI images + segmentation masks.
Fine-grained classes (4-class): 0: Glioma 1: Meningioma 2: Pituitary 3: No Tumor
Coarse classes (2-class): Tumor-present Tumor-absent
Masks allow extraction of ROI-based geometric and intensity features.
Masks allow extraction of ROI-based geometric and intensity features. 🔹 Hybrid ML Pipeline (ROI + Global Features)
Uses a combination of: Geometric features (Area, Perimeter, Solidity, Circularity) Intensity statistics (Mean, Std, Min, Max) Texture features (LBP, HOG, Gabor filters) Gradient & edge-based features
🔹 Random Forest–based Dual-Classifiers
Two ML models trained using extracted features: Model Task Accuracy Random Forest Tumor Type Classification 90.0% Random Forest Severity Classification 59.7%
🔹 Interactive Gradio Deployment
The user can upload MRI images and receive: Tumor Presence Tumor Type Severity Visual & color-coded explanation
1️⃣ Install Dependencies pip install -r requirements.txt
2️⃣ Run the Gradio App python app.py
3️⃣ Upload an MRI Image
You will receive: Tumor detection Classification Severity level Explanation
Python Scikit-learn (Random Forest) OpenCV (Image Processing) Albumentations (Mask Augmentation) Skimage (HOG, LBP) Gabor Filters Numpy, Pandas Gradio (Deployment UI)