Latent Space Synergy: Text-Guided Data Augmentation for Direct Diffusion Biomedical Segmentation

TL;DR

SynDiff is a novel framework that uses text-guided diffusion models to generate synthetic medical images, significantly improving segmentation accuracy while enabling real-time clinical deployment.

Contribution

It introduces a direct latent estimation method for diffusion, allowing single-step inference and effective synthetic data augmentation for biomedical segmentation.

Findings

Achieves 96.0% Dice on CVC-ClinicDB

Maintains real-time inference speed

Enhances segmentation robustness with synthetic data

Abstract

Medical image segmentation suffers from data scarcity, particularly in polyp detection where annotation requires specialized expertise. We present SynDiff, a framework combining text-guided synthetic data generation with efficient diffusion-based segmentation. Our approach employs latent diffusion models to generate clinically realistic synthetic polyps through text-conditioned inpainting, augmenting limited training data with semantically diverse samples. Unlike traditional diffusion methods requiring iterative denoising, we introduce direct latent estimation enabling single-step inference with T x computational speedup. On CVC-ClinicDB, SynDiff achieves 96.0% Dice and 92.9% IoU while maintaining real-time capability suitable for clinical deployment. The framework demonstrates that controlled synthetic augmentation improves segmentation robustness without distribution shift. SynDiff bridges the gap between data-hungry deep learning models and clinical constraints, offering an efficient solution for deployment in resourcelimited medical settings.