Class-N-Diff: Classification-Induced Diffusion Model Can Make Fair Skin Cancer Diagnosis

Generative models have shown great performance improvement in generating synthetic data, especially in medical imaging. Traditional class-conditioned generative models often struggle to generate images that accurately represent specific medical classes, which limits their effectiveness in tasks such as skin cancer diagnosis. To address this problem, we propose a classification-induced diffusion model, Class-N-Diff, that can simultaneously generate and classify dermoscopic images. Our Class-N-Diff model integrates a convolutional classifier within a diffusion model to guide the generative model based on its class conditions. We integrate a classification objective within the generative process. Thus, the model has better control over class-conditioned image synthesis, resulting in more realistic and diverse images. Additionally, the classifier achieves enhanced performance, showing its value for downstream diagnostic tasks. This unique integration in our Class-N-Diff makes it a robust tool for enhancing the quality and utility of synthetic dermoscopic image generation. Our code is available at https://github.com/Munia03/Class-N-Diff.