🧠 Brain Tumor Detection System

An automated medical imaging system designed to detect and classify brain tumors from MRI scans using Convolutional Neural Networks (CNN).

📋 Project Overview

This project processes MRI images to classify them into one of 4 categories:

1. Glioma
2. Meningioma
3. Pituitary
4. No Tumor

We utilize a CNN architecture for feature extraction and classification, achieving high accuracy through K-Fold Cross Validation.

---

🛠️ Setup & Installation

1. Prerequisites

Ensure you have Python 3.8+ installed.

2. Install Dependencies

Run the following command to install all required libraries:

pip install -r requirements.txt

Key Libraries used: tensorflow, pandas, numpy, matplotlib, seaborn, scikit-learn

3. Dataset Setup

Note: The dataset is not included in this repository.

1. Place your dataset folder in the project root.
2. Ensure it has the standard structure:

   dataset/Training/glioma/
   dataset/Training/meningioma/
   dataset/Training/notumor/
   dataset/Training/pituitary/
   dataset/Testing/glioma/
   dataset/Testing/meningioma/
   dataset/Testing/notumor/
   dataset/Testing/pituitary/
   

3. Update the DATASET_PATH variable in train_model.py if your folder name is different.

---

🚀 How to Run

Step 1: Analyze the Data

Check the balance of the dataset to ensure fair training.

python data_analysis.py

Outputs: dataset_distribution.png (Bar chart of image counts)

Step 2: Train the Model

Train the CNN using 5-Fold Cross Validation. This will generate training graphs and save the best model.

python train_model.py

Outputs:

• tumor_detection_model.h5 ( The saved trained model)
• accuracy_fold_X.png & loss_fold_X.png (Training graphs)
• confusion_matrix_fold_X.png (Performance heatmaps)

Step 3: Interactive Evaluation

Test the model yourself! This script opens a window where you can click "Next" to see predictions on random test images.

python interactive_eval.py

(Requires tumor_detection_model.h5 to exist)

---

📊 Technical Details

• Model: Custom Sequential CNN (3 Convolutional Blocks + Dropout).
• Input: 150x150 pixel MRI images.
• Validation: 5-Fold Cross Validation to ensure robustness.
• Metrics: We monitor Accuracy, Loss (Sparse Categorical Crossentropy), and Confusion Matrices.