Team Members: Hareeshravi, Dharun Kumar, Dhivinkumar, Naveen Prasath
Brain tumors are abnormal growths of cells within the brain, and their early detection is crucial for effective treatment. Unfortunately, manual analysis of MRI scans can be time-consuming, and the risk of human error is always present. To address this challenge, we have developed ResCortex, a custom deep learning model built upon ResNet-50, designed specifically to detect and classify brain tumors from MRI images automatically. By utilizing advanced preprocessing techniques and powerful neural networks, ResCortex extracts critical features from the scans, improving the accuracy and speed of diagnosis. This AI-driven system helps healthcare professionals identify tumor types quickly and with greater precision, ultimately supporting faster decision-making and better patient outcomes.
We used the Brain Tumor MRI dataset containing T1-weighted contrast-enhanced MRI images, which are categorized into:
- Glioma Tumor
- Meningioma Tumor
- Pituitary Tumor
- No Tumor
- Step 1: Image Preprocessing – Cropping, Augmentation, and Normalization
- Step 2: Data Serialization – Pickling the preprocessed data for memory-efficient loading
- Step 3: Model Design – Custom ResNet-50 architecture (ResCortex)
- Step 4: Training with Hyperparameter Tuning
- Step 5: Evaluation – Confusion Matrix, Precision, Recall, F1-Score, and Classification Report
- Step 6: Deployment – React-based frontend with Flask backend for prediction API
- Validation Accuracy: >94%
- F1 Score: ~89%
- Optimizer: Adam
- Loss Function: Categorical Crossentropy with Label Smoothing
- Features: Class Weights, Early Stopping, Model Checkpointing
- Fork the repository
- Create your branch
- Add new models inside
Contributions/Your_Name - Include a README file detailing your approach
- Dharun Kumar
- Dhivinkumar
- Naveen Prasath
- Python 3.7
- TensorFlow / Keras
- ReactJS
- Flask
- Google Colab (for model training)
- Anime.js (for UI animations)