Exploiting Geometric Features via Hierarchical Graph Pyramid Transformer for Cancer Diagnosis Using Histopathological Images.

Cancer is widely recognized as the primary cause of mortality worldwide, and pathology analysis plays a pivotal role in achieving accurate cancer diagnosis. The intricate representation of features in histopathological images encompasses abundant information crucial for disease diagnosis, regarding cell appearance, tumor microenvironment, and geometric characteristics. However, recent deep learning methods have not adequately exploited geometric features for pathological image classification due to the absence of effective descriptors that can capture both cell distribution and gathering patterns, which often serve as potent indicators. In this paper, inspired by clinical practice, a Hierarchical Graph Pyramid Transformer (HGPT) is proposed to guide pathological image classification by effectively exploiting a geometric representation of tissue distribution which was ignored by existing state-of-the-art methods. First, a graph representation is constructed according to morphological feature of input pathological image and learn geometric representation through the proposed multi-head graph aggregator. Then, the image and its graph representation are feed into the transformer encoder layer to model long-range dependency. Finally, a locality feature enhancement block is designed to enhance the 2D local representation of feature embedding, which is not well explored in the existing vision transformers. An extensive experimental study is conducted on Kather-5K, MHIST, NCT-CRC-HE, and GasHisSDB for binary or multi-category classification of multiple cancer types. Results demonstrated that our method is capable of consistently reaching superior classification outcomes for histopathological images, which provide an effective diagnostic tool for malignant tumors in clinical practice.

Derivation and Comprehensive Analysis of Aging Patterns in Patients with Bladder Cancer.

BACKGROUND: Aging is an essential risk factor for cancer. However, aging-related genes (ARGs) have not been comprehensively analyzed in bladder cancer (BC). Therefore, the study is aimed at derivating a risk stratification system for BC patients based on ARGs. METHODS: Public databases were used to acquire ARGs sets, transcriptome files, and clinical data. The "limma" package was then used to screen for differential ARGs while also using univariate Cox regression analysis to explore for prognostic ARGs. The "ConsensusClusterPlus" package was used to perform aging patterns in BC patients based on the above prognostic ARGs. Subsequently, aging patterns were investigated in survival prediction, mutation landscape, immunotherapy, immunological checkpoints, and immune microenvironment. We likewise utilized gene enrichment analysis to explore the biological functions that were behind the findings. To construct a risk signature and nonogram for prognostic prediction, we used LASSO and Cox regression analysis based on differential genes in aging patterns. In addition, we plotted a nomogram and validate the accuracy of the risk signature in GEO and TCGA cohorts. We explored the possible biological mechanism using GSEA analysis and preliminarily identified a hub gene using PPI network. Finally, we validated the expression of hub gene in BC cell lines. RESULTS: We screened 84 downregulated ARGs, 74 upregulated ARGs, and 32 prognostic ARGs in the human aging genome resource. The aging patterns based on prognostic genes had excellent survival prediction (p < 0.001) and discriminatory ability in 405 BC patients. In addition, we found no significant differences in aging patterns in mutation analysis, which were all characterized by TP53, TTN, and KMT2D mutations. It is worth noting that cluster B in the aging patterns has a better response to immunotherapy and a more active immune microenvironment (p < 0.05). In addition, gene enrichment analysis showed that aging patterns may be related to biological processes such as Staphylococcus aureus infection, phagosome, and cytokine-cytokine receptor interaction. Subsequently, we constructed a risk signature based on 16 differential genes from different aging patterns and had good survival prediction ability in both GEO and TCGA cohort. Specifically, survival analysis revealed a significantly shorter survival time in the high-risk group than in the low-risk group (TCGA and GEO, p < 0.001). In addition, AUC values in the ROC analysis predicted 1, 3, and 5 years in TCGA cohort that are 0.713, 0.714, and 0.738, respectively. AUC values predicted 1, 3, and 5 years in GEO cohort that are 0.606, 0.663, and 0.718, respectively. There is no doubt that risk score was an independent prognostic factor from results of multivariate Cox regression analysis in BC patients (p < 0.001). There were also significant differences in immune cell infiltration, immune checkpoint, and immune score between the two groups (p < 0.05), but it should not be ignored that the correlation with the HLA expression was weak. Finally, we identified and validated CLIC3 as a hub gene that may be involved in the Wnt signaling pathway, etc. CONCLUSION: We provided robust evidences that aging patterns based on ARGs can guide targeted therapy and survival prediction in BC patients.