Pouya Afshin, David Helminiak, Tianling Niu, Julie M. Jorns, Tina Yen, Bing Yu, Dong Hye Ye
A self-supervised learning approach using a latent diffusion model significantly improves breast cancer classification accuracy in deep ultraviolet images by generating high-quality synthetic training data.
Breast cancer surgery requires accurate assessment of tissue margins to ensure all cancerous cells are removed while preserving healthy tissue. A new imaging technique using deep ultraviolet light provides detailed images for this purpose, but there is a lack of labeled data to train effective AI models. Researchers have developed a method using a self-supervised learning model to create synthetic images that mimic real ones, enhancing the training data available. This approach significantly improves the accuracy of breast cancer classification, achieving over 96% accuracy, which could lead to better surgical outcomes.
