A deep learning framework integrating MRI image preprocessing methods for brain tumor segmentation and...

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Oncology

Background:

• Accurate glioma grading is essential for effective treatment planning and patient prognosis.
• Magnetic Resonance Imaging (MRI) is a key modality for glioma diagnosis.
• Deep learning offers potential for automated and accurate glioma classification from MRI data.

Purpose of the Study:

• To develop and evaluate a deep learning model for accurate glioma diagnosis and grading using MRI.
• To compare the performance of different deep learning architectures (VGG, GoogleNet) combined with UNet segmentation.
• To investigate the impact of preprocessing techniques, including data augmentation and Window Setting Optimization, on model performance.

Main Methods:

• A deep learning pipeline involving MRI image segmentation using UNet architecture.
• Extraction of brain tumor regions post-segmentation.
• Classification of gliomas into high-grade and low-grade categories using VGG and GoogleNet.
• Implementation of data augmentation and Window Setting Optimization for preprocessing.

Main Results:

• The combination of data augmentation and Window Setting Optimization yielded high Dice coefficients (0.82, 0.91, 0.72) for tumor segmentation.
• Most models achieved approximately 93% accuracy on the testing dataset.
• The VGG combined with UNet segmentation pipeline demonstrated the highest accuracy at 97.44%.

Conclusions:

• The presented deep learning architecture provides a realistic model for glioma detection.
• Data augmentation and segmentation are crucial for enhancing the performance of glioma classification models.
• The findings support the use of advanced AI techniques for improved glioma diagnosis and patient care.