Diffusion-Based Data Augmentation for Medical Image Segmentation

2025-08-25

AI Summary

Medical image segmentation is hindered by the scarcity of annotated data for rare pathologies. To address this, we propose a text-guided diffusion augmentation framework. Methodologically, we innovatively integrate a latent diffusion model with spatial mask inpainting, conditioning image synthesis on medical text descriptions and lesion location masks to generate high-fidelity abnormal images. Concurrently, we introduce a lightweight latent-space segmentation network to enforce spatial consistency and enable automatic quality filtering of generated samples—supporting single-step inference without manual annotation. Evaluated on three benchmarks—CVC-ClinicDB, Kvasir-SEG, and REFUGE2—our method achieves state-of-the-art performance, improving Dice scores by 8–10% over prior work. Notably, it reduces false-negative rates for challenging cases—including small and flat polyps—by up to 28%, demonstrating significant gains in detecting subtle, low-contrast lesions.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Natural Language Processing: Sentence-level Semantics, Textual Inference, etc.

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Medical image segmentation models struggle with rare abnormalities due to scarce annotated pathological data. We propose DiffAug a novel framework that combines textguided diffusion-based generation with automatic segmentation validation to address this challenge. Our proposed approach uses latent diffusion models conditioned on medical text descriptions and spatial masks to synthesize abnormalities via inpainting on normal images. Generated samples undergo dynamic quality validation through a latentspace segmentation network that ensures accurate localization while enabling single-step inference. The text prompts, derived from medical literature, guide the generation of diverse abnormality types without requiring manual annotation. Our validation mechanism filters synthetic samples based on spatial accuracy, maintaining quality while operating efficiently through direct latent estimation. Evaluated on three medical imaging benchmarks (CVC-ClinicDB, Kvasir-SEG, REFUGE2), our framework achieves state-of-the-art performance with 8-10% Dice improvements over baselines and reduces false negative rates by up to 28% for challenging cases like small polyps and flat lesions critical for early detection in screening applications.

Problem

• Augmenting scarce annotated medical data for rare abnormalities
• Generating diverse pathological samples without manual annotation
• Ensuring spatial accuracy in synthetic medical image segmentation

Innovation

• Diffusion-based abnormality synthesis via inpainting
• Latent-space segmentation validation for quality control
• Text-guided generation without manual annotation