Context-aware subregion-wise patching for effective brain tumor segmentation in 3D MRI scans

Early diagnosis of brain tumors is imperative to save lives. However, automated diagnosis faces several challenges, such as the use of 3D magnetic resonance imaging (MRI) scans, which demand substantial computational resources especially since medical images are highly prone to class imbalance, making models biased toward the majority class. Various patching techniques, such as fixed, random, and random biased, have been employed to address these issues. However, they often fail to effectively preserve the entire tumor due to cropped tumor regions effectively. Conversely, overlapping patches that preserve tumor regions exacerbate class imbalance because of redundant background pixels. Furthermore, the number of patches increases as the overlap decreases, resulting in a nonlinear relationship that introduces redundancy and affects overall segmentation performance. Moreover, these techniques do not enable the model to focus on regions of interest corresponding to specific tumor subregions. In this study, we propose a novel subregion based patching technique that selects patches from both the whole and tumor core regions.The proposed patching technique is independent of the reference standard labels for patch localization. These labels are used only after patch centers are fixed, to extract labeled patches for supervised training. This patching technique centers the tumor within each patch, effectively preserving its overall morphology and preventing cropped tumor boundaries. This technique enhances model training by enabling independent learning from distinct subregions. We evaluate our technique against state-of-the-art patching techniques and conduct extensive experiments using the BraTS 2020 dataset. The proposed patching technique consistently improves patch centering and class balance, leading to measurable gains in segmentation performance under the evaluated settings, achieving Dice scores of 96% for the whole tumor, 90% for the tumor core, and 92% for the enhancing tumor region. Code is available at: https://github.com/Mutyybaa/-TDA-and-CCA-Sub-Region-based-Patching- .