L-Diffusion: Laplace Diffusion for Efficient Pathology Image Segmentation

Weihan Li, Linyun Zhou, YangJian, Shengxuming Zhang, Xiangtong Du, Xiuming Zhang, Jing Zhang, ChaoqingXu, Mingli Song, Zunlei Feng

TL;DR: We propose a contrastive learning framework based on Laplace diffusion, with theoretical proof and quantitative validation. The model achieves accurate cross-scale multi-label segmentation of pathological images under limited supervision.

Abstract: Pathology image segmentation plays a pivotal role in artificial digital pathology diagnosis and treatment. Existing approaches to pathology image segmentation are hindered by labor-intensive annotation processes and limited accuracy in tail-class identification, primarily due to the long-tail distribution inherent in gigapixel pathology images. In this work, we introduce the Laplace Diffusion Model, referred to as L-Diffusion, an innovative framework tailored for efficient pathology image segmentation. L-Diffusion utilizes multiple Laplace distributions, as opposed to Gaussian distributions, to model distinct components—a methodology supported by theoretical analysis that significantly enhances the decomposition of features within the feature space. A sequence of feature maps is initially generated through a series of diffusion steps. Following this, contrastive learning is employed to refine the pixel-wise vectors derived from the feature map sequence. By utilizing these highly discriminative pixel-wise vectors, the segmentation module achieves a harmonious balance of precision and robustness with remarkable efficiency. Extensive experimental evaluations demonstrate that L-Diffusion attains improvements of up to 7.16\%, 26.74\%, 16.52\%, and 3.55\% on tissue segmentation datasets, and 20.09\%, 10.67\%, 14.42\%, and 10.41\% on cell segmentation datasets, as quantified by DICE, MPA, mIoU, and FwIoU metrics. The source are available at https://github.com/Lweihan/LDiffusion.

Lay Summary:

Accurately identifying different parts of tissue or cells in pathology images is a key step in helping doctors diagnose and treat diseases. However, this process usually takes a lot of time and effort because it depends on expert labeling, and it often struggles to recognize rare types of cells or tissue accurately.

Our work presents a new method called L-Diffusion that makes this task much more efficient and reliable. Instead of using traditional approaches, L-Diffusion takes a different mathematical path to better understand the features inside these massive medical images. It gradually improves its understanding of the image in steps and then uses a smart way of comparing small parts of the image to sharpen its predictions.

As a result, our method can more precisely detect and segment both common and rare parts of tissue or cells. In tests on multiple datasets, L-Diffusion showed major improvements in accuracy, confirming that it is both powerful and practical for real-world use in medical image analysis.

You can find the code and more details at: https://github.com/Lweihan/LDiffusion

Application-Driven Machine Learning: This submission is on Application-Driven Machine Learning.

Link To Code: https://github.com/Lweihan/LDiffusion

Primary Area: Applications->Computer Vision

Keywords: Diffusion Model, Contrastive Learning, Multi-label Segmentation, Pathological slides

Submission Number: 4492