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

2025-10-19

AI Summary

Traditional class-conditional generative models struggle to simultaneously achieve realism, class fidelity, and fairness in medical image synthesis—particularly for dermatoscopic images—often underrepresenting rare disease classes. To address this, we propose the Classification-Induced Diffusion Model (CIDM), the first framework to bidirectionally integrate a deep classifier into the diffusion process: forward classification guidance steers reverse sampling for enhanced class consistency, while backward classification gradients refine both generation quality and discriminative capability. CIDM jointly trains the diffusion model and classifier, incorporating a fine-grained class-conditioning mechanism. This yields substantial improvements in synthetic image realism, diversity, and rare-class recognition. Evaluated on multiple dermatoscopic datasets, CIDM boosts classification accuracy by 3.2–5.8% over baselines and effectively mitigates data imbalance. The approach establishes a new paradigm for fair and reliable AI-assisted skin cancer diagnosis.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Humans and AI: Crowd Sourcing and Human Computation

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Economics, Online Markets and Human Computation: Fairness and ethical considerations in crowd work and in human-in-the-loop AI systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Generative models, especially Diffusion Models, have demonstrated remarkable capability in generating high-quality synthetic data, including medical images. However, traditional class-conditioned generative models often struggle to generate images that accurately represent specific medical categories, limiting their usefulness for applications such as skin cancer diagnosis. To address this problem, we propose a classification-induced diffusion model, namely, Class-N-Diff, to simultaneously generate and classify dermoscopic images. Our Class-N-Diff model integrates a classifier within a diffusion model to guide image generation based on its class conditions. Thus, the model has better control over class-conditioned image synthesis, resulting in more realistic and diverse images. Additionally, the classifier demonstrates improved performance, highlighting its effectiveness for downstream diagnostic tasks. This unique integration in our Class-N-Diff makes it a robust tool for enhancing the quality and utility of diffusion model-based synthetic dermoscopic image generation. Our code is available at https://github.com/Munia03/Class-N-Diff.

Problem

• Improving class-conditioned medical image generation accuracy
• Enhancing skin cancer diagnosis through synthetic data
• Integrating classification guidance into diffusion model training

Innovation

• Integrates classifier into diffusion model for guidance
• Simultaneously generates and classifies dermoscopic images
• Enhances control over class-conditioned image synthesis