About the Project

Detecting brain tumours early is crucial for better patient outcomes.

In this project, the team delves into the impact of various pre-processing techniques on the classification accuracy of brain tumour images obtained from MRI scans. The project compares and analyzes a few pre-processing methods. The employed dataset comprises MRI scans collected from three different sources: FigShare, Sartaj, and BR35 H. Through the investigation, the project highlights the importance of pre-processing by evaluating classification performance on noisy, original, and pre-processed images. The TV method emerges as particularly effective, demonstrating the highest Peak Signal-to-Noise Ratio (PSNR) and superior noise reduction capabilities.

The project further explores the performance of various machine learning classifiers for brain tumour classification. These include Support Vector Machine (SVM), Convolutional Neural Network (CNN), Transfer Learning using the MobileNetV2 base model, and an ensemble method combining CNN and Transfer Learning. The ensemble model achieves the highest accuracy. Evaluation metrics such as Area Under the Curve (AUC), F1-score, recall, and overall accuracy validate the significant improvement in classification accuracy facilitated by pre-processing techniques. By reducing noise and enhancing image quality, pre-processing plays a critical role in improving the accuracy of brain tumour detection, especially for early-stage tumours. This project offers valuable insights for enhancing diagnostic accuracy in clinical settings, ultimately contributing to improved patient care and outcomes.