Effective Segmentation of Post-Treatment Gliomas Using Simple Approaches: Artificial Sequence Generation and Ensemble Models

• DOI: 10.48550/arxiv.2409.08143
• Publication Date: 2024-09-12
• Authors: Heejong Kim; Léo Milecki; Mina C. Moghadam; Fengbei Liu; Minh‐Hoang Nguyen; Eric Qiu; Abhishek Thanki; Mert R. Sabuncu
• Keywords: Sequence (biology)

Segmentation is a crucial task in the medical imaging field and often an important primary step or even prerequisite to analysis of volumes. Yet treatments such as surgery complicate accurate delineation regions interest. The BraTS Post-Treatment 2024 Challenge published first public dataset for post-surgery glioma segmentation addresses aforementioned issue by fostering development automated tools MRI data. In this effort, we propose two straightforward approaches enhance performances deep learning-based methodologies. First, incorporate additional input based on simple linear combination available sequences input, which highlights enhancing tumors. Second, employ various ensembling methods weigh contribution battery models. Our results demonstrate that these significantly improve performance compared baseline models, underscoring effectiveness improving image tasks.