Recognition and quantitative analytics of histopathological cells are the golden standard for diagnosing multiple cancers. Despite recent advances in deep learning techniques that have been widely investigated for the automated segmentation of various types of histopathological cells, the heavy dependency on specific histopathological image types with sufficient supervised annotations, as well as the limited access to clinical data in hospitals, still pose significant challenges in the application of computer-aided diagnosis in pathology. In this paper, we focus on the model generalization of cell segmentation towards cross-tissue histopathological images. Remarkably, a novel target-specific finetuning-based self-supervised domain adaptation framework is proposed to transfer the cell segmentation model to unlabeled target datasets, without access to source datasets and annotations. When performed on the target unlabeled histopathological image set, the proposed method only needs to tune very few parameters of the pre-trained model in a self-supervised manner. Considering the morphological properties of pathological cells, we introduce two constraint terms at both local and global levels into this framework to access more reliable predictions. The proposed cross-domain framework is validated on three different types of histopathological tissues, showing promising performance in self-supervised cell segmentation. Additionally, the whole framework can be further applied to clinical tools in pathology without accessing the original training image data. The code and dataset are released at: https://github.com/NeuronXJTU/SFDA-CellSeg.
7 October 2022 in The Journal of Organic Chemistry
Purpose: To investigate the association between Syndecan-2 gene methylation and protein expression dynamics in colorectal cancer and evaluate its clinicopathological significance. Methods: We analyzed 197 tissues (96 colorectal cancer, 64 adenoma, and 37 normal tissues). Syndecan-2 methylation was quantified using quantitative methylation-specific polymerase chain reaction, and protein expression was evaluated via immunohistochemistry. Results: Syndecan-2 methylation frequency was significantly higher in colorectal cancer (81.25%) and adenoma (64.06%) than in normal (5.41%) tissues. It showed high diagnostic accuracy for colorectal cancer (area under the curve = 0.88). Although methylation status was unrelated to most clinical parameters, higher methylation levels were correlated with poor tumor differentiation ( p = 0.022). Syndecan-2 protein was aberrantly expressed in tumor stroma but lacked clinical correlations. A significant positive correlation between methylation and protein expression was observed across the sample ( p < 0.05); however, this association was not maintained within the colorectal cancer subgroup specifically, showing no intrasample concordance (Kappa statistic = −0.046). Conclusion: Syndecan-2 methylation serves as a highly sensitive epigenetic biomarker for the early detection of colorectal cancer. The findings reveal a complex and heterogeneous relationship between Syndecan-2 methylation and its protein expression during colorectal cancer progression, suggesting stage-specific regulatory mechanisms. Further multiomics studies are warranted to elucidate the underlying functional dynamics.
The revisions of the 6th edition of the WHO Classification of Digestive System Tumors regarding appendiceal and colorectal neoplasms comprehensively integrates contemporary research breakthroughs with evolving clinical necessities. Concerning appendiceal tumors, the new edition pioneers the incorporation of molecular prognostic subtyping while underscoring the imperative of mismatch repair assessment, thereby providing a pathological framework to guide targeted therapeutic interventions and genetic risk management. Regarding colorectal tumors, the text offers a systematic review of the clinicopathological characteristics inherent to serrated lesions and juvenile polyps, concurrently standardizing endoscopic biopsy protocols and histological categorization for dysplasia associated with inflammatory bowel disease. It also introduces rare subtypes, specifically low-grade tubuloglandular adenocarcinoma and lymphoglandular complex-like adenocarcinoma, clarifying their indolent biological nature to prevent excessive therapeutic measures. In evaluating dysplastic changes, the revised criteria now encompass novel entities—such as serrated, crypt cell, and goblet cell-deficient dysplasia—alongside traditional adenomatous dysplasia. This expansion may significantly enhance early neoplastic detection and refines risk stratification accuracy, potentially informing optimal clinical management. Furthermore, the inclusion of consensus molecular subtypes (CMS/PDS) and other prognostic models establishes a pathological foundation for future molecularly driven stratified therapies and clinical trial architectures. Collectively, these revisions may carry profound implications for clinical decision-making across surgical planning, immunotherapy selection, endoscopic surveillance, and genetic counseling, marking a pivotal advancement in the precision management of colorectal neoplasms.
Objectives: To explore the current research progress and trends in the application of artificial intelligence (AI) for preoperative diagnosis of thyroid nodules (TN), and to identify the key research directions through bibliometric analysis. Methods: Articles on the application of AI for preoperative diagnosis of TN were retrieved from the Web of Science core collection and Scopus databases. Tools like VOSviewer, CiteSpace, the bibliometrix, Scimago Graphica, and Charticulator were used for bibliometric analysis. Trends in annual publishing volumes, collaborations between authors and institutions, highly cited articles, keyword co-occurrence, keyword clustering, and keyword burst analysis are all included. Results: A total of 1370 publications were included in this study, with an rapid growth trend observed in the annual publications in this field. China has made significant contributions, with Shanghai Jiao Tong University producing the most publications. Chinese scholar XuDong emerged as the most prolific author (30 publications). Journal analysis revealed that Thyroid is the highest-impact journal in terms of citation frequency. High-frequency keywords in the field include “TN”, “cancer” and “ultrasound”, while recent emerging keywords include “semantic segmentation”, and “AI”. Conclusions: Since 2019, the application of AI in the preoperative diagnosis of TN has attracted increasing research attention, with China emerging as a major contributor to this research field. The number of journals publishing in this domain has significantly increased, and the publications reflect a trend toward interdisciplinary research. The research focus has gradually shifted from conventional diagnostic approaches toward pixel-level lesion quantification, fine-grained feature extraction and the integration of AI-enabled diagnostic approaches, highlighting the potential of AI to improve TN assessment.
Recently, there is growing attention on one-stage panoptic segmentation methods which aim to segment instances and stuff jointly within a fully convolutional pipeline efficiently. However, most of the existing works directly feed the backbone features to various segmentation heads ignoring the demands for semantic and instance segmentation are different: The former needs semantic-level discriminative features, while the latter requires features to be distinguishable across instances. To alleviate this, we propose to first predict semantic-level and instance-level correlations among different locations that are utilized to enhance the backbone features, and then feed the improved discriminative features into the corresponding segmentation heads, respectively. Specifically, we organize the correlations between a given location and all locations as a continuous sequence and predict it as a whole. Considering that such a sequence can be extremely complicated, we adopt Discrete Fourier Transform (DFT), a tool that can approximate an arbitrary sequence parameterized by amplitudes and phrases. For different tasks, we generate these parameters from the backbone features in a fully convolutional way which is optimized implicitly by corresponding tasks. As a result, these accurate and consistent correlations contribute to producing plausible discriminative features which meet the requirements of the complicated panoptic segmentation task. To verify the effectiveness of our methods, we conduct experiments on several challenging panoptic segmentation datasets and achieve state-of-the-art performance on MS COCO with $45.1$\% PQ and ADE20k with $32.6$\% PQ.