Neura - Brain tumor detection model

➲ Overview

Brain tumors are among the most critical health issues worldwide, requiring early and accurate diagnosis to improve patient outcomes. This project focuses on automatic brain tumor detection from MRI scans using a deep learning approach.

The model is based on a Convolutional Neural Network (CNN) architecture (Xception-based), trained to classify MRI images into tumor or non-tumor categories. The goal is to provide a reliable, fast, and scalable solution to assist radiologists and healthcare professionals.

Our trained model achieves:

• ~98% training accuracy
• ~96% validation accuracy
• ~95% test accuracy
• ROC-AUC: 0.97–0.99 across classes

➲ Prerequisites

Install the following packages before running the project:

• Python 3.x
• TensorFlow
• Keras
• NumPy
• Pandas
• Matplotlib
• Scikit-learn

➲ The Dataset

We used publicly available MRI brain tumor datasets containing labeled images for classification.

Classes: Tumor, Non-Tumor

Data Split: 70% training, 15% validation, 15% testing

Images were preprocessed (resizing, normalization, augmentation) to improve model performance.

Sample MRI images:

➲ Dataset Description

Tumor: MRI scans indicating the presence of a brain tumor.

Non-Tumor: MRI scans of healthy brains without visible tumors.

The preprocessing pipeline includes:

Resizing images to uniform dimensions

Normalizing pixel values

Data augmentation (rotation, flip, zoom)

➲ Model Architecture

We implemented a CNN model with transfer learning (Xception).

Key layers include:

Convolutional layers with ReLU activation

MaxPooling for feature extraction

Dropout for regularization

Fully connected dense layers

Softmax activation for classification

Libraries used import numpy as np import pandas as pd import tensorflow as tf from tensorflow.keras.preprocessing.image import ImageDataGenerator from tensorflow.keras.models import Sequential from tensorflow.keras.layers import Conv2D, MaxPooling2D, Flatten, Dense, Dropout from sklearn.metrics import classification_report, confusion_matrix import matplotlib.pyplot as plt

➲ Model Performance

Confusion Matrix & Classification Report confirm strong performance across both classes.

➲ Results Visualization

Training & Validation Accuracy/Loss curves

ROC curves for classification

Example predictions with MRI scan input vs. predicted label