Kidney Stones in Unenhanced Abdominal CT Images

This study applies deep semantic segmentation models to improve automated kidney and kidney stone segmentation on unenhanced abdominal CT images, addressing limitations in prior renal imaging AI work. The authors propose a specialized training scheme and evaluate five segmentation network variants, demonstrating improved 2D and 3D performance when using their approach. Additionally, they release an open-source CT dataset to support further research, positioning the work as a foundational step toward AI-driven diagnosis and personalized management of kidney disease.