A Reflective Study on Deep Learning in Brain Tumor Segmentation

Abstract

Brain tumor segmentation is a critical step in medical image analysis, aiding in diagnosis, treatment planning, and outcome prediction. Accurate segmentation can significantly impact patient care, yet it remains a complex and demanding task due to heterogeneous nature of tumor structures and imaging modalities. In recent years, deep learning has emerged as a powerful tool in medical imaging, offering state-of-the-art performance in segmentation tasks. While numerous studies have introduced advanced models and achieved impressive results, the specific practical challenges encountered during basic implementations are often underreported. For beginners and researchers new to the field, understanding these foundational obstacles is essential for building robust and effective models. This reflective study addresses that gap by exploring the application of a Convolutional Neural Network (CNN) for brain tumor segmentation using Brain Tumor Segmentation (BraTS) dataset. The CNN model was trained with hyperparameters optimized through Bayesian optimization to improve performance. Rather than focusing solely on metrics, this work emphasizes the difficulties faced throughout the process, including data variability, class imbalance, and preprocessing sensitivity. The aim is to shed light on common yet underdiscussed issues and provide practical insights to help guide new researchers in the field of deep learning for medical image analysis.