• We propose a weakly-supervised segmentation method for IHC staining pathological cell images, which only uses positive cell counting labels. • It transforms image level counting results into superpixel level classification results as coarse segmentation masks via the designed adaptive top-k instance selection strategy. • It further generates refined pixel-wise cell segmentation masks via the proposed superpixel mask refining approach based on IHC prior knowledge. • Experiments on two public datasets show that our method performs excellently in cell counting and segmentation tasks, and keeps strong generalization. Immunohistochemistry (IHC) can specifically stain immune cells, and quantitative analysis of tumor-infiltrating immune cells is crucial for the diagnosis of malignant tumors and the prognosis of immunotherapy. However, for current deep learning-based immuno-quantitative methods, it is difficult to obtain labeled data for fine-grained analysis tasks, such as cell segmentation, because manual labeling is laborious. Moreover, information contained in coarse-grained image-level labels, such as positive cell counting labels, is difficult to leverage. Here, we propose a novel multi-stage automated immuno-quantitative analysis model for cell counting and segmentation, which is solely supervised by positive cell counting labels. In order to excavate the potential information of counting labels for cell segmentation, based on the idea of Multiple Instance Learning (MIL), we transform entire image (bag) level counting results into superpixel (instance) level classification results as coarse segmentation masks, using a specifically designed adaptive top-k instance selection strategy. Furthermore, a mask refinement approach is proposed, which can refine coarse masks generated by superpixel instance classification into pixel-level pseudo masks as supervision of the segmentation network. Our method enables all-in-one immuno-quantitative analysis of cell counting and segmentation with only positive cell counting labels, alleviating the dependence on fine-grained labels and economizing labeling cost. Our work is the first to apply MIL to small-scale IHC-stained cell image segmentation task. We trained and tested our model on the public LYSTO dataset and conducted generalization tests on the NuClick IHC dataset. Our model achieved segmentation dice scores of 0.809 and 0.711, respectively, demonstrating approximately 7.44% and 3.95% performance improvements over the state-of-the-art (SOTA) methods. The experiments show that our model performs excellently in cell counting and cell segmentation tasks while maintaining strong generalization capabilities.

Few-shot learning presents a critical solution for cancer diagnosis in computational pathology (CPath), addressing fundamental limitations in data availability, particularly the scarcity of expert annotations and patient privacy constraints. A key challenge in this paradigm stems from the inherent disparity between the limited training set of whole slide images (WSIs) and the enormous number of contained patches, where a significant portion of these patches lacks diagnostically relevant information, potentially diluting the model’s ability to learn and focus on critical diagnostic features. While recent works attempt to address this by incorporating additional knowledge, several crucial gaps hinder further progress: (1) despite the emergence of powerful pathology foundation models (FMs), their potential remains largely untapped, with most approaches limiting their use to basic feature extraction; (2) current language guidance mechanisms attempt to align text prompts with vast numbers of WSI patches all at once, struggling to leverage rich pathological semantic information. To this end, we introduce the knowledge-enhanced adaptive visual compression framework, dubbed FOCUS, which uniquely combines pathology FMs with language prior knowledge to enable a focused analysis of diagnostically relevant regions by prioritizing discriminative WSI patches. Our approach implements a progressive three-stage compression strategy: we first leverage FMs for global visual redundancy elimination, and integrate compressed features with language prompts for semantic relevance assessment, then perform neighbor-aware visual token filtering while preserving spatial coherence. Extensive experiments on pathological datasets spanning breast, lung, and ovarian cancers demonstrate its superior performance in few-shot pathology diagnosis. Codes are available at https://github.com/dddavid4real/FOCUS.

Pathological image has been recognized as the gold standard for cancer diagnosis for more than a century. However, some internal regions of pathological images may inevitably exhibit various degradation issues, including low resolution, image blurring, and image noising, which will affect disease diagnosis, staging, and risk stratification.Existing pathological image restoration methods were mainly based on generative adversarial networks (GANs) to improve image quality, which are limited by the inherent instability and loss of structural details, often resulting in artifacts in the restored images.Large scale of whole slide images (WSIs) also makes it hard for efficient processing and restoration. To address these limitations, we propose a conditional visual autoregressive model (CVARPath) for next-scale token prediction, guided by the degraded tokens from the current scale. We introduce a novel framework that employs quantified encoders specifically designed for pathological image generation, which learns consistent sparse vocabulary tokens through self-supervised contrastive learning. Furthermore, our method efficiently compresses image patches into compact degraded sparse tokens at smaller scales and reconstructs high-quality large-scale whole slide images (WSIs). This is achieved using only an 8x8 vocabulary index for 256x256 images while maintaining minimal reconstruction loss. Experimental results demonstrate that our approach significantly enhances image quality, achieving an approximately 30% improvement in mean Frechet inception distance (FID) compared to popular conditional GANs and diffusion models across various degradation scenarios in pathological images.

Whole slide image (WSI) analysis presents significant computational challenges due to the massive number of patches in gigapixel images. While transformer architectures excel at modeling long-range correlations through self-attention, their quadratic computational complexity makes them impractical for computational pathology applications. Existing solutions like local-global or linear self-attention reduce computational costs but compromise the strong modeling capabilities of full self-attention. In this work, we propose **Querent**, *i.e.*, the **quer**y-awar**e** long co**nt**extual dynamic modeling framework, which achieves a theoretically bounded approximation of full self-attention while delivering practical efficiency. Our method adaptively predicts which surrounding regions are most relevant for each patch, enabling focused yet unrestricted attention computation only with potentially important contexts. By using efficient region-wise metadata computation and importance estimation, our approach dramatically reduces computational overhead while preserving global perception to model fine-grained patch correlations. Through comprehensive experiments on biomarker prediction, gene mutation prediction, cancer subtyping, and survival analysis across over 10 WSI datasets, our method demonstrates superior performance compared to the state-of-the-art approaches. Codes are available at https://github.com/dddavid4real/Querent.

We introduce LYSTO, the Lymphocyte Assessment Hackathon, which was held in conjunction with the MICCAI 2019 Conference in Shenzhen (China). The competition required participants to automatically assess the number of lymphocytes, in particular T-cells, in images of colon, breast, and prostate cancer stained with CD3 and CD8 immunohistochemistry. Differently from other challenges setup in medical image analysis, LYSTO participants were solely given a few hours to address this problem. In this paper, we describe the goal and the multi-phase organization of the hackathon; we describe the proposed methods and the on-site results. Additionally, we present post-competition results where we show how the presented methods perform on an independent set of lung cancer slides, which was not part of the initial competition, as well as a comparison on lymphocyte assessment between presented methods and a panel of pathologists. We show that some of the participants were capable to achieve pathologist-level performance at lymphocyte assessment. After the hackathon, LYSTO was left as a lightweight plug-and-play benchmark dataset on grand-challenge website, together with an automatic evaluation platform.

Fluorescence staining is an important technique in life science for labeling cellular constituents. However, it also suffers from being time-consuming, having difficulty in simultaneous labeling, etc. Thus, virtual staining, which does not rely on chemical labeling, has been introduced. Recently, deep learning models such as transformers have been applied to virtual staining tasks. However, their performance relies on large-scale pretraining, hindering their development in the field. To reduce the reliance on large amounts of computation and data, we construct a Swin-transformer model and propose an efficient supervised pretraining method based on the masked autoencoder (MAE). Specifically, we adopt downsampling and grid sampling to mask 75% of pixels and reduce the number of tokens. The pretraining time of our method is only 1/16 compared with the original MAE. We also design a supervised proxy task to predict stained images with multiple styles instead of masked pixels. Additionally, most virtual staining approaches are based on private datasets and evaluated by different metrics, making a fair comparison difficult. Therefore, we develop a standard benchmark based on three public datasets and build a baseline for the convenience of future researchers. We conduct extensive experiments on three benchmark datasets, and the experimental results show the proposed method achieves the best performance both quantitatively and qualitatively. In addition, ablation studies are conducted, and experimental results illustrate the effectiveness of the proposed pretraining method. The benchmark and code are available at https://github.com/birkhoffkiki/CAS-Transformer.

Biomedical microscopy images with high-resolution (HR) and axial information can help analysis and diagnosis. However, obtaining such images usually takes more time and economic costs, which makes it impractical in most scenarios. In this paper, we first propose a novel Self-texture Transfer Super-resolution and Refocusing Network (STSRNet) to reconstruct HR multi-focal plane (MFP) images from a single 2D low-resolution (LR) wide field image without relying on scanning or any special devices. The proposed STSRNet consists of three parts: the backbone module for extracting features, the self-texture transfer module for transferring and fusing features, and the flexible reconstruction module for SR and refocusing. Specifically, the self-texture transfer module is designed for images with self-similarity such as cytological images and it searches for similar textures within the image and transfers to help MFP reconstruction. As for reconstruction module, it is composed of multiple pluggable components, each of which is responsible for a specific focal plane, so as to performs SR and refocusing all focal planes at one time to reduce computation. We conduct extensive experiments on cytological images and the experiments show that MFP images reconstructed by STSRNet have richer details in the axial and horizontal directions than input images. At the same time, the reconstructed MFP images also perform better than single 2D wide field images on high-level tasks. The proposed method provides relatively high-quality MFP images when real MFP images cannot be obtained, which greatly expands the application potential of LR wide-field images. To further promote the development of this field, we released our cytology dataset named RSDC for more researchers to use.

In the cytopathology screening of cervical cancer, high-resolution digital cytopathological slides are critical for the interpretation of lesion cells. However, the acquisition of high-resolution digital slides requires high-end imaging equipment and long scanning time. In the study, we propose a GAN-based progressive multi-supervised super-resolution model called PathSRGAN (pathology super-resolution GAN) to learn the mapping of real low-resolution and high-resolution cytopathological images. With respect to the characteristics of cytopathological images, we design a new two-stage generator architecture with two supervision terms. The generator of the first stage corresponds to a densely-connected U-Net and achieves 4× to 10× super resolution. The generator of the second stage corresponds to a residual-in-residual DenseBlock and achieves 10× to 20× super resolution. The designed generator alleviates the difficulty in learning the mapping from 4× images to 20× images caused by the great numerical aperture difference and generates high quality high-resolution images. We conduct a series of comparison experiments and demonstrate the superiority of PathSRGAN to mainstream CNN-based and GAN-based super-resolution methods in cytopathological images. Simultaneously, the reconstructed high-resolution images by PathSRGAN improve the accuracy of computer-aided diagnosis tasks effectively. It is anticipated that the study will help increase the penetration rate of cytopathology screening in remote and impoverished areas that lack high-end imaging equipment.