Improving the radiographic image analysis of the classic metaphyseal lesion via conditional diffusion models

Shaoju Wu, Sila Kurugol, Andy Tsai

Abstract: Highlights•We propose a new generative diffusion model for generating realistic radiographic images.•Classifiers trained using our generated data outperformed other data augmentation methods.•Our data augmentation method provides extra segmentation masks, enabling the training of an automated algorithm for segmentation.•This is the first application of generative diffusion models for infant abuse diagnosis.

External IDs:dblp:journals/mia/WuKT24