A Multilevel Deep Learning Model for Automated Brain Tumor Segmentation Using Magnetic Resonance Images

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Neuroscience

Background:

• Accurate brain tumor segmentation from MRI is crucial for clinical assessment and treatment planning.
• Challenges include complex anatomy, tumor variability, and imbalanced data, necessitating automated methods.
• Existing methods struggle with precise delineation due to these inherent difficulties.

Purpose of the Study:

• To develop and evaluate a novel multilevel deep learning model for automated brain tumor segmentation using MRI.
• To address the challenge of class imbalance in brain tumor datasets.
• To improve the accuracy and reliability of automated tumor segmentation in clinical practice.

Main Methods:

• A modified Synthetic Minority Oversampling Technique (SMOTE) was used for imbalance-aware preprocessing.
• A Multilevel Architecture-Based Modified U-Net was employed to capture multiscale spatial features.
• The BraTS dataset was utilized for model development and evaluation, with performance assessed using Dice, Jaccard, and MCC metrics.

Main Results:

• The proposed model demonstrated superior performance compared to BWT-based and conventional U-Net methods across various tumor grades.
• Higher Dice, Jaccard, and MCC scores indicated improved agreement with expert annotations.
• The model achieved more accurate and consistent tumor delineation in MRI images.

Conclusions:

• The multilevel deep learning model offers an effective framework for automated brain tumor segmentation from MRI.
• The combination of imbalance-aware preprocessing and a modified U-Net architecture enhances segmentation performance.
• Future research will involve external dataset validation and comparison with state-of-the-art models.