Multi-scale based Network and Adaptive EfficientnetB7 with ASPP: Analysis of Novel Brain Tumor Segmentation and...

Area of Science:

• Medical imaging analysis
• Deep learning applications in healthcare
• Neurological disorder diagnosis

Background:

• Deep learning enhances medical image analysis accuracy for disease detection.
• Magnetic Resonance Imaging (MRI) is crucial for diagnosing brain disorders due to its soft tissue contrast.
• Conventional brain tumor classification methods face challenges with computational complexity and accuracy.

Purpose of the Study:

• To develop a robust and efficient deep learning framework for brain tumor segmentation and classification.
• To assist clinicians in precise and early diagnosis of brain tumors for effective treatment planning.
• To improve upon the limitations of conventional classification techniques in terms of accuracy and computational load.

Main Methods:

• Utilized Multiscale Bilateral Awareness Network (MBANet) for segmenting abnormal regions in MRI images.
• Employed a novel Region Vision Transformer-based Adaptive EfficientNetB7 with Atrous Spatial Pyramid Pooling (RVAEB7-ASPP) for classification.
• Optimized model hyperparameters using the Modified Random Parameter-based Hippopotamus Optimization Algorithm (MRP-HOA).

Main Results:

• The proposed MRP-HOA-RVAEB7-ASPP model achieved a classification accuracy of 98.2%.
• Demonstrated significant outperformance compared to existing state-of-the-art methods in brain tumor classification.
• Verified model effectiveness through comprehensive experimental evaluation and performance metrics.

Conclusions:

• MBANet effectively segments brain tumors, and RVAEB7-ASPP reliably classifies them.
• The integrated MRP-HOA-RVAEB7-ASPP model optimizes feature extraction and parameter tuning, enhancing accuracy and robustness.
• The framework offers a reliable solution for early brain tumor detection, improving patient outcomes through timely intervention.