Deep Learning for Precise MRI Segmentation of Lower-Grade Gliomas

Abstract

Brain tumors represent a significant public health issue worldwide, affecting individuals across all age groups and leading to severe neurological and cognitive deficits. Lower-grade gliomas (LGGs), classified by the World Health Organization (WHO) as grade II or III, are characterized by more diffuse infiltration into brain tissue compared to high-grade gliomas but exhibit a slower growth rate. Precise evaluation of tumor resection and detection of residual tumor cells are critical, as incomplete resection is associated with an increased risk of disease recurrence. This study reviews an automated, deep learning-based approach for brain tumor segmentation in Magnetic Resonance Imaging (MRI) using the U-Net architecture to improve diagnostic precision. Utilizing the Multimodal Brain Tumor Image Segmentation Benchmark (BRATS) dataset, the study applies preprocessing, data augmentation, and model training conducted on Google Colaboratory. Performance evaluation metrics, including the Dice Similarity Coefficient (DSC), sensitivity, and specificity, indicate the model’s effectiveness, with a DSC of 0.89, sensitivity of 0.87, and specificity of 0.99. The study also highlights the potential of radiogenomics, which correlates imaging features with tumor genomics, to enable personalized treatment strategies for LGG patients and improve survival outcomes. This work underscores the value of deep learning in automated MRI segmentation and its potential to significantly enhance patient outcomes in clinical practice.