Latent Diffusion Models with Self-Distillation from Separated Conditions for Accurate Prostate Cancer Grading
Latent Diffusion Models (LDMs) with Self-Distillation from Separated Conditions (DISC) can generate high-quality synthetic histopathology tiles that accurately capture multiple Gleason Grades, leading to significant improvements in both pixel-level and slide-level prostate cancer grading performance.