🧠 Explainable Brain Tumor Detection using Deep Learning

An AI-powered medical imaging system that automatically detects and classifies brain tumors from MRI scans using a Convolutional Neural Network (CNN) integrated with Explainable Artificial Intelligence (Grad-CAM). The application combines deep learning with visual interpretability to provide transparent predictions, confidence analysis, probability visualization, and downloadable clinical-style reports through an interactive Streamlit web application.

🌐 Live Demo

🚀 Experience the Application

https://brain-mri-ai.streamlit.app/

🧠 Project Overview

Brain tumor diagnosis is a critical task in medical imaging where timely and accurate interpretation of MRI scans plays a significant role in clinical decision-making. Deep learning has demonstrated remarkable performance in medical image analysis; however, most models function as black boxes, making it difficult to understand the reasoning behind their predictions.

This project presents an Explainable Artificial Intelligence (XAI) framework for multiclass brain tumor detection and classification using Convolutional Neural Networks (CNNs). To improve transparency and interpretability, the model integrates Gradient-weighted Class Activation Mapping (Grad-CAM), enabling visualization of the image regions that contribute most to the model's predictions.

The application is deployed as an interactive Streamlit web application where users can upload MRI scans, receive real-time predictions, analyze confidence scores and probability distributions, visualize Grad-CAM heatmaps, and generate downloadable clinical-style PDF reports.

The project demonstrates the integration of medical image preprocessing, deep learning, computer vision, Explainable AI, and interactive deployment into a unified diagnostic support system.

Disclaimer: This project is developed solely for educational and research purposes. It is not intended to replace professional medical diagnosis or clinical decision-making.

🚀 Key Features

🧠 Deep Learning-Based Classification

Automatic Brain Tumor Detection
Four-Class MRI Classification
CNN-Based Prediction Engine
Softmax Probability Distribution

🔥 Explainable Artificial Intelligence

Grad-CAM Heatmap Generation
Visual Interpretation of Model Decisions
Explainable Predictions
Improved Model Transparency

📊 Intelligent Prediction Analysis

Prediction Confidence Score
Class Probability Distribution
Uncertainty Detection
Clinical Interpretation

📄 Automated Medical Reporting

Clinical-Style PDF Report Generation
Prediction Summary
Embedded Grad-CAM Visualization
AI Verification Stamp

🌐 Interactive Web Application

Streamlit-Based User Interface
Drag-and-Drop MRI Upload
Real-Time Prediction
Responsive and User-Friendly Experience

🧬 Supported Classification Categories

Tumor Type	Description
🟣 Glioma	Tumors originating from glial cells within the brain and spinal cord.
🔵 Meningioma	Tumors developing from the meninges, the protective membranes surrounding the brain and spinal cord.
🟠 Pituitary	Tumors affecting the pituitary gland located at the base of the brain.
🟢 No Tumor	MRI scans without detectable tumor abnormalities.

🛠 Technology Stack

Category	Technologies
Programming Language	Python
Deep Learning	TensorFlow, Keras
Computer Vision	OpenCV
Data Processing	NumPy, Pandas
Data Visualization	Matplotlib
Explainable AI	Grad-CAM
Web Framework	Streamlit
Report Generation	ReportLab
Development Environment	Jupyter Notebook, VS Code

🗂 Dataset Overview

The model is trained on a publicly available Brain MRI Dataset containing four different categories of brain MRI images.

Dataset Classes

Class	Description
Glioma	Brain tumors originating from glial cells
Meningioma	Tumors affecting the protective membranes of the brain
Pituitary	Tumors occurring in the pituitary gland
No Tumor	Healthy brain MRI images without tumor abnormalities

The dataset was divided into training and testing subsets to evaluate the model's generalization capability while maintaining balanced class representation.

🧪 Image Preprocessing Pipeline

Medical images require preprocessing before being used for deep learning to improve consistency and model performance.

The preprocessing pipeline includes:

RGB Image Conversion
Brain Region Extraction
Contrast Enhancement
Gaussian Noise Reduction
Image Resizing (224 × 224)
Pixel Normalization

These preprocessing steps improve feature extraction while reducing noise and variations present in MRI scans.

🧠 CNN Model Architecture

The classification model is built using a custom Convolutional Neural Network (CNN) designed for multiclass brain tumor classification.

Model Components

Convolutional Layers
ReLU Activation
Max Pooling Layers
Dropout Regularization
Fully Connected Dense Layers
Softmax Output Layer

The network progressively extracts low-level and high-level image features before classifying MRI scans into one of the four supported tumor categories.

🔥 Explainable Artificial Intelligence (Grad-CAM)

Traditional deep learning models often behave as black-box systems, making it difficult to understand how predictions are generated.

To improve transparency, this project integrates Gradient-weighted Class Activation Mapping (Grad-CAM).

Grad-CAM produces visual heatmaps that highlight the regions of an MRI scan contributing most to the model's prediction, enabling users to better understand the reasoning behind the classification.

Benefits of Explainability

Improved model transparency
Better prediction interpretability
Increased user confidence
Educational visualization for AI-assisted diagnosis

Note: Grad-CAM highlights influential image regions and should not be interpreted as an exact tumor segmentation method.

⚙ System Workflow

                MRI Brain Image
                       │
                       ▼
              Image Preprocessing
                       │
                       ▼
               CNN Classification
                       │
                       ▼
            Probability Distribution
                       │
                       ▼
              Confidence Analysis
                       │
                       ▼
           Grad-CAM Heatmap Generation
                       │
                       ▼
            Clinical Interpretation
                       │
                       ▼
            Downloadable PDF Report

📂 Project Structure

Explainable-Brain-Tumor-Detection
│
├── assets
│   ├── banner
│   │   └── banner.png
│   │
│   ├── icons
│   │   └── ai_stamp.png
│   │
│   └── screenshots
│       ├── ui_main.png
│       ├── prediction_output.png
│       ├── gradcam_output.png
│       ├── preprocessing.png
│       └── report_output.png
│
├── dataset
│   ├── Training
│   └── Testing
│
├── docs
│   ├── Project_Report.pdf
│   └── Research_Paper.pdf
│
├── model
│   ├── model_loader.py
│   └── multiclass_brain_tumor_cnn.h5
│
├── utils
│   ├── gradcam.py
│   ├── preprocess.py
│   └── report.py
│
├── app.py
├── requirements.txt
├── README.md
└── .gitignore

🎯 Core Functionalities

Brain MRI Classification
Image Quality Assessment
MRI Image Preprocessing
CNN-Based Prediction
Confidence Analysis
Probability Visualization
Grad-CAM Explainability
Clinical Interpretation
PDF Report Generation
Interactive Streamlit Interface

📊 Model Performance

The developed CNN model demonstrates strong performance in multiclass brain tumor classification while maintaining prediction transparency through Explainable AI.

Evaluation Metrics

Overall Classification Accuracy: ~91%
Precision, Recall, and F1-Score Evaluation
Confusion Matrix Analysis
Softmax Probability Distribution
Grad-CAM Visual Validation

The combination of quantitative metrics and visual explanations enables a more interpretable AI-assisted diagnostic workflow.

📸 Application Screenshots

🖥️ Home Interface

📊 Prediction Output

🔥 Grad-CAM Visualization

📄 Clinical Report

🧪 Image Preprocessing

📥 Trained Model

The trained CNN model is not included in this repository because of GitHub's file size limitations.

Download the trained model from the link below and place it inside the model/ directory.

Google Drive

https://drive.google.com/drive/folders/1J6zwcEmjOlWpcxnOJCGMR1g0edaYCM2G?usp=sharing

model/
└── multiclass_brain_tumor_cnn.h5

🚀 Getting Started

Clone Repository

git clone https://github.com/TanmayT134/Explainable-Brain-Tumor-Detection.git

cd Explainable-Brain-Tumor-Detection

Create Virtual Environment

python -m venv venv

Activate Virtual Environment

Windows

venv\Scripts\activate

macOS / Linux

source venv/bin/activate

Install Dependencies

pip install -r requirements.txt

Run the Application

streamlit run app.py

The application will automatically open in your default web browser.

🎯 Applications

This project demonstrates the practical integration of Artificial Intelligence and Medical Imaging in several domains, including:

AI-Assisted Medical Image Analysis
Brain Tumor Classification
Explainable Artificial Intelligence (XAI)
Medical Imaging Research
Computer Vision Applications
Educational Demonstration of Deep Learning
Clinical Decision Support Research

⚠️ Limitations

Although the model demonstrates promising performance, several limitations remain:

Performance depends on the quality of MRI images.
Trained on a limited publicly available dataset.
Grad-CAM provides visual explanations but not precise tumor segmentation.
Intended solely for educational and research purposes.
Should not be used as a substitute for professional medical diagnosis.

🚀 Future Enhancements

Potential improvements include:

Integration of larger clinical datasets
Transfer Learning using advanced architectures
3D MRI volume analysis
Tumor segmentation using U-Net
Multi-modal MRI support
Real-time clinical integration
Model optimization for faster inference
Cloud deployment with scalable APIs

📚 Documentation

The repository also includes supporting project documentation.

docs/
├── Project_Report.pdf
└── Research_Paper.pdf

These documents provide detailed information about the project methodology, implementation, evaluation, and research findings.

👥 Team

Member	Contribution
Tanmay Tawade	CNN model integration, Streamlit application development, Grad-CAM implementation, system architecture, report generation, and deployment
Aishwarya Kale	Dataset preparation, preprocessing, workflow design, documentation, and project validation
Sakshi Bedekar	Project planning, testing, performance evaluation, documentation, and presentation

The project was developed collaboratively, with all members contributing to the design, implementation, testing, and evaluation phases.

🙏 Acknowledgements

Special thanks to the following resources and communities:

Kaggle Brain MRI Dataset
TensorFlow & Keras
OpenCV
Streamlit
ReportLab
Research community working in Deep Learning and Explainable Artificial Intelligence (XAI)

👨‍💻 Author

Tanmay Tawade

If you found this project helpful or interesting, consider giving it a ⭐ on GitHub.

⭐ If you found this project useful, please consider starring the repository!

Building trustworthy AI requires not only accurate predictions but also transparent and interpretable decision-making.

Made with ❤️ using Python, TensorFlow & Explainable AI by Tanmay Tawade

Name		Name	Last commit message	Last commit date
.devcontainer		.devcontainer
assets		assets
model		model
utils		utils
.gitattributes		.gitattributes
.gitignore		.gitignore
README.md		README.md
app.py		app.py
requirements.txt		requirements.txt
runtime.txt		runtime.txt

Latest commit

History

Folders and files

Repository files navigation