Brain Tumor/Mass Classification Framework Using Magnetic-Resonance-Imaging-Based Isolated and Developed Transfer...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Machine learning and deep learning offer advanced tools for medical image analysis.
• Accurate and early diagnosis of brain tumors is crucial for effective treatment planning.
• Magnetic Resonance Imaging (MRI) is a primary modality for visualizing brain structures and detecting abnormalities.

Purpose of the Study:

• To develop and evaluate a novel transfer deep-learning model for the early diagnosis and subclassification of brain tumors.
• To assess the model's performance in classifying brain MRI images into specific tumor types: pituitary, meningioma, and glioma.
• To validate the model's adaptability and reliability across different MRI datasets.

Main Methods:

• Development of isolated convolutional neural network (CNN) models from scratch to evaluate performance on brain MRI images.
• Implementation of a transfer learning approach by re-utilizing a 22-layer binary-classification CNN model.
• Fine-tuning the model's weights for multi-class classification of brain tumors (pituitary, meningioma, glioma).

Main Results:

• The developed transfer-learned model achieved a high accuracy of 95.75% on brain MRI images from the same machine.
• Testing on an unseen dataset from a different MRI machine yielded an accuracy of 96.89%, demonstrating generalizability.
• The model effectively classified brain tumors into subclasses with high precision.

Conclusions:

• The proposed deep-learning framework demonstrates significant potential for early and accurate brain tumor diagnosis.
• The transfer learning approach enhances model adaptability and reliability for real-world clinical applications.
• This technology can serve as a valuable tool for radiologists and clinicians in improving patient outcomes.