Enhancing Label-efficient Medical Image Segmentation with Text-guided Diffusion Models

TL;DR

TextDiff leverages inexpensive medical text annotations to enhance semantic representations in diffusion models, significantly improving medical image segmentation performance with minimal training data.

Contribution

The paper introduces TextDiff, a novel method that integrates diagnostic text with diffusion models for improved label-efficient medical image segmentation.

Findings

Outperforms state-of-the-art methods on QaTa-COVID19 and MoNuSeg datasets.

Requires only training of cross-attention and pixel classifier, reducing annotation costs.

Achieves superior segmentation accuracy with few training samples.

Abstract

Aside from offering state-of-the-art performance in medical image generation, denoising diffusion probabilistic models (DPM) can also serve as a representation learner to capture semantic information and potentially be used as an image representation for downstream tasks, e.g., segmentation. However, these latent semantic representations rely heavily on labor-intensive pixel-level annotations as supervision, limiting the usability of DPM in medical image segmentation. To address this limitation, we propose an enhanced diffusion segmentation model, called TextDiff, that improves semantic representation through inexpensive medical text annotations, thereby explicitly establishing semantic representation and language correspondence for diffusion models. Concretely, TextDiff extracts intermediate activations of the Markov step of the reverse diffusion process in a pretrained diffusion model on large-scale natural images and learns additional expert knowledge by combining them with complementary and readily available diagnostic text information. TextDiff freezes the dual-branch multi-modal structure and mines the latent alignment of semantic features in diffusion models with diagnostic descriptions by only training the cross-attention mechanism and pixel classifier, making it possible to enhance semantic representation with inexpensive text. Extensive experiments on public QaTa-COVID19 and MoNuSeg datasets show that our TextDiff is significantly superior to the state-of-the-art multi-modal segmentation methods with only a few training samples.