An Efficient and Cohesive System for Enhanced Accuracy in Malignant Brain Tumor Diagnosis

Brain tumors present a significant health challenge, making an accurate diagnosis crucial for effective treatment and improved patient outcomes. This study aims to develop SCSFE-DiagBT, an advanced deep learning framework that integrates classification and segmentation. It is designed for enhanced brain tumor diagnosis using MRI scans.

We propose a Stacked Classification and Segmentation Model with Feature Extraction (SCSFE-DiagBT). This model unifies the processes of classification and segmentation into a single diagnostic pipeline. It leverages the complementarity of both tasks by employing a deep learning approach to efficiently analyze MRI scans.

Experimental results indicate substantial improvements in performance metrics. In classification, metrics such as accuracy, precision, and recall improved by up to 31% compared to baseline models, with high accuracy. For segmentation, the model achieved a 12% improvement in the Dice coefficient, reaching an accuracy of 99.58% and a Dice score of 86.30%. This demonstrates robust generalization and minimal overfitting.

The findings underscore the effectiveness of integrating classification and segmentation in the diagnosis of brain tumors. The enhanced interpretability through feature extraction further supports the model's utility in clinical settings, potentially reducing diagnostic variability associated with manual interpretations.

SCSFE-DiagBT emerges as a highly accurate and viable tool for brain tumor diagnosis. It offers significant advancements over traditional methods and promising implications for improving patient care and outcomes in neuro-oncology.