Multi-Granularity 3D Kidney Lesion Characterization from CT Volumes
Researchers have developed LesionDETR, a novel deep learning model designed to characterize kidney lesions from CT scans with greater detail. Unlike previous methods that focused on organ-level predictions, LesionDETR predicts individual lesions and their attributes, such as type, size, and enhancement. The model achieved strong performance on two datasets, demonstrating its potential for generating structured radiology reports. AI
IMPACT Enables more detailed and structured radiology reports for kidney lesions.