Doctor Approved: Generating Medically Accurate Skin Disease Images through AI-Expert Feedback

2025-06-14

AI Summary

Medical imaging data are scarce and expensive to annotate, resulting in synthetically generated images from diffusion models that often lack clinical credibility—thereby limiting the generalizability of diagnostic models. To address this, we propose MAGIC, an AI-expert collaborative framework that introduces a multimodal large language model (MLLM) as a scalable, automated medical evaluator, translating expert clinical criteria into executable image optimization feedback. MAGIC integrates knowledge-guided feedback distillation with conditional diffusion-based synthesis to generate high-fidelity skin lesion images at low annotation cost. Clinical evaluation by dermatologists confirms high clinical credibility of the synthesized images. In a 20-class skin disease classification task, MAGIC improves overall accuracy by 9.02%; under few-shot settings, the improvement reaches 13.89%. This work bridges the gap between generative modeling and clinical utility through human-in-the-loop, knowledge-infused synthesis.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language

Application Category

• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Large language models for search

Abstract

Paucity of medical data severely limits the generalizability of diagnostic ML models, as the full spectrum of disease variability can not be represented by a small clinical dataset. To address this, diffusion models (DMs) have been considered as a promising avenue for synthetic image generation and augmentation. However, they frequently produce medically inaccurate images, deteriorating the model performance. Expert domain knowledge is critical for synthesizing images that correctly encode clinical information, especially when data is scarce and quality outweighs quantity. Existing approaches for incorporating human feedback, such as reinforcement learning (RL) and Direct Preference Optimization (DPO), rely on robust reward functions or demand labor-intensive expert evaluations. Recent progress in Multimodal Large Language Models (MLLMs) reveals their strong visual reasoning capabilities, making them adept candidates as evaluators. In this work, we propose a novel framework, coined MAGIC (Medically Accurate Generation of Images through AI-Expert Collaboration), that synthesizes clinically accurate skin disease images for data augmentation. Our method creatively translates expert-defined criteria into actionable feedback for image synthesis of DMs, significantly improving clinical accuracy while reducing the direct human workload. Experiments demonstrate that our method greatly improves the clinical quality of synthesized skin disease images, with outputs aligning with dermatologist assessments. Additionally, augmenting training data with these synthesized images improves diagnostic accuracy by +9.02% on a challenging 20-condition skin disease classification task, and by +13.89% in the few-shot setting.

Problem

• Addressing medical data scarcity for ML diagnostic models
• Improving accuracy of AI-generated skin disease images
• Reducing expert workload in synthetic image validation

Innovation

• Uses diffusion models for synthetic image generation
• Incorporates expert feedback via MLLM evaluators
• Improves diagnostic accuracy with AI-augmented data