Brain Tumor Classification from MRI Scans via Transfer Learning and Enhanced Feature Representation

2025-09-26

AI Summary

To address data scarcity, severe class imbalance, and poor generalizability in brain tumor MRI classification, this paper proposes a deep learning–based automated classification framework. Methodologically: (1) a Dense-Dropout sequential architecture is designed to enhance nonlinear feature representation and model robustness; (2) the clinically validated MMCBT multi-center brain tumor MRI dataset is constructed to alleviate data paucity; (3) leveraging a pre-trained ResNet50 backbone, the framework integrates global average pooling, linear projection, and targeted data augmentation to mitigate class imbalance. Experimental results demonstrate that the proposed framework significantly improves classification accuracy and cross-center generalization performance on the balanced MMCBT dataset. It achieves state-of-the-art performance while maintaining clinical interpretability and reliability, offering an efficient and robust AI solution for computer-aided diagnosis in neuro-oncology.

Technology Category

• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Computer Vision: Multi-modal Vision
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Brain tumors are abnormal cell growths in the central nervous system (CNS), and their timely detection is critical for improving patient outcomes. This paper proposes an automatic and efficient deep-learning framework for brain tumor detection from magnetic resonance imaging (MRI) scans. The framework employs a pre-trained ResNet50 model for feature extraction, followed by Global Average Pooling (GAP) and linear projection to obtain compact, high-level image representations. These features are then processed by a novel Dense-Dropout sequence, a core contribution of this work, which enhances non-linear feature learning, reduces overfitting, and improves robustness through diverse feature transformations. Another major contribution is the creation of the Mymensingh Medical College Brain Tumor (MMCBT) dataset, designed to address the lack of reliable brain tumor MRI resources. The dataset comprises MRI scans from 209 subjects (ages 9 to 65), including 3671 tumor and 13273 non-tumor images, all clinically verified under expert supervision. To overcome class imbalance, the tumor class was augmented, resulting in a balanced dataset well-suited for deep learning research.

Problem

• Automating brain tumor detection from MRI scans
• Enhancing feature learning with Dense-Dropout sequence
• Addressing dataset scarcity with balanced MMCBT dataset

Innovation

• Transfer learning with pre-trained ResNet50 model
• Dense-Dropout sequence for enhanced feature learning
• Class balancing via data augmentation techniques