muhammedpanchla/brain-tumor-detection

🧠 Brain Tumor Detection System

Overview

This is a Flask-based web application for brain tumor detection using a VGG16 model (transfer learning), trained on brain MRI images. Upload an MRI scan (PNG, JPG, TIFF) to get a classification: glioma, meningioma, notumor, or pituitary, along with confidence score.

• —Model: VGG16 (Transfer Learning)
• —Framework: PyTorch
• —Task: Multi-class Classification of Brain Tumors
• —Dataset: Based on Brain Tumor MRI Dataset (inspired by Kaggle datasets)
• —Classes: Glioma, Meningioma, No Tumor, Pituitary
• —Metrics: From notebook: High accuracy on test set (update with actual values from your notebook, e.g., Accuracy: 98%, F1: 97%)

Features

• —Upload MRI images and get instant classification results
• —Displays predicted tumor type and confidence percentage
• —Simple, clean UI for medical/demo use
• —Disclaimer for educational use only

How to Use

1. 1.Open the app
2. 2.Upload a brain MRI image
3. 3.Get prediction: Tumor type and confidence
4. 4.View results in JSON or integrate with frontend JS for display

⚠️ Disclaimer: This is for educational/research purposes only. Not for clinical use. Consult a medical professional for diagnoses.

Technical Details

• —Architecture: VGG16 with custom classifier (3 FC layers with dropout)
• —Input: 224x224 RGB images (resized)
• —Output: Class label + softmax confidence
• —Dependencies: See requirements.txt

From the notebook:

• —Training: Transfer learning from pretrained VGG16, fine-tuned on brain MRI dataset
• —Augmentation: Used transforms for robustness
• —Evaluation: Confusion matrix, accuracy, precision/recall on test set

Deployment on Hugging Face

This space uses Docker to run the Flask app.