ScholarWorks@성균관대학교

초록

Artificial intelligence (AI)-driven analysis of medical images is a critical component of modern consumer healthcare systems, where early and accurate detection of abnormalities, such as tumors, is paramount. However, most existing medical AI systems aim at single-organ segmentation, that restricts their applicability to realistic multi-organ diagnostic workflow. Different from existing works, this paper proposes a unified model that can segment tumors in both the liver and lung. To achieve this, we propose a simple yet effective approach that integrates organ-specific context into the model, enabling it to handle heterogeneous tumor properties without complex architectural modifications. Our experiments with the Medical Segmentation Decathlon (MSD) liver and lung benchmark datasets demonstrate that the proposed approach achieves robust tumor segmentation performance across both organs, outperforming single-organ tumor segmentation models and highlighting the potential of tumor segmentation in different organs with a single model.

키워드