Boxiang Zhang

Few-shot learning is widely applied in the current stage for remote sensing image classification to use prior knowledge to identify new classes faster. However, since existing few-shot remote sensing image classification methods only process the feature vectors extracted from the complete image, ignoring the localized knowledge in the input incorporated into the target can better use the contextual information to capture more local details. To address these problems, we propose a metric learning framework based on maximizing mutual information between different views (MMI-ML). Specifically, we introduce a self-supervised model to train an embedding network with enhanced feature representation by maximizing mutual information of global features and local features at different scales. In addition, we design a new embedding network to make it more appropriate for the self-supervised model. Finally, we devise a new loss function in the training stage, which can effectively speed up the convergence of the model. We conduct comparative experiments on three public remote sensing datasets, and the experimental results show that the classification accuracy of the MMI-ML framework is improved by up to 3.22%.

Existing methods of cross-modal domain adaptation for 3D semantic segmentation predict results only via 2D-3D complementarity that is obtained by cross-modal feature matching. However, as lacking supervision in the target domain, the complementarity is not always reliable. The results are not ideal when the domain gap is large. To solve the problem of lacking supervision, we introduce masked modeling into this task and propose a method Mx2M, which utilizes masked cross-modality modeling to reduce the large domain gap. Our Mx2M contains two components. One is the core solution, cross-modal removal and prediction (xMRP), which makes the Mx2M adapt to various scenarios and provides cross-modal self-supervision. The other is a new way of cross-modal feature matching, the dynamic cross-modal filter (DxMF) that ensures the whole method dynamically uses more suitable 2D-3D complementarity. Evaluation of the Mx2M on three DA scenarios, including Day/Night, USA/Singapore, and A2D2/SemanticKITTI, brings large improvements over previous methods on many metrics.

Existing methods of cross-modal domain adaptation for 3D semantic segmentation predict results only via 2D-3D complementarity that is obtained by cross-modal feature matching. However, as lacking supervision in the target domain, the complementarity is not always reliable. The results are not ideal when the domain gap is large. To solve the problem of lacking supervision, we introduce masked modeling into this task and propose a method Mx2M, which utilizes masked cross-modality modeling to reduce the large domain gap. Our Mx2M contains two components. One is the core solution, cross-modal removal and prediction (xMRP), which makes the Mx2M adapt to various scenarios and provides cross-modal self-supervision. The other is a new way of cross-modal feature matching, the dynamic cross-modal filter (DxMF) that ensures the whole method dynamically uses more suitable 2D-3D complementarity. Evaluation of the Mx2M on three DA scenarios, including Day/Night, USA/Singapore, and A2D2/SemanticKITTI, brings large improvements over previous methods on many metrics.

Background: Fibroblast growth factor receptors (FGFRs) modulate numerous cellular processes in tumor cells and tumor microenvironment. However, the effect of FGFRs on tumor prognosis and tumor-infiltrating lymphocytes in gastric cancer (GC) remains controversial.Methods: The expression of four different types of FGFRs was analyzed via GEPIA, TCGA-STAD, and GTEX databases and our 27 pairs of GC tumor samples and the adjacent normal tissue. Furthermore, the Kaplan–Meier plot and the TCGA database were utilized to assess the association of FGFRs with clinical prognosis. The R software was used to evaluate FGFRs co-expression genes with GO/KEGG Pathway Enrichment Analysis. In vitro and in vivo functional analyses and immunoblotting were performed to verify FGFR4 overexpression consequence. Moreover, the correlation between FGFRs and cancer immune infiltrates was analyzed by TIMER and TCGA databases. And the efficacy of anti-PD-1 mAb treatment was examined in NOG mouse models with overexpressed FGFR1 or FGFR4.Results: The expression of FGFRs was considerably elevated in STAD than in the normal gastric tissues and was significantly correlated with poor OS and PFS. ROC curve showed the accuracy of the FGFRs in tumor diagnosis, among which FGFR4 had the highest ROC value. Besides, univariate and multivariate analysis revealed that FGFR4 was an independent prognostic factor for GC patients. According to a GO/KEGG analysis, the FGFRs were implicated in the ERK/MAPK, PI3K-AKT and extracellular matrix (ECM) receptor signaling pathways. In vivo and in vitro studies revealed that overexpression of FGFR4 stimulated GC cell proliferation, invasion, and migration. In addition, FGFR1 expression was positively correlated with infiltrating levels of CD8+ T-cells, CD4+ T-cells, macrophages, and dendritic cells in STAD. In contrast, FGFR4 expression was negatively correlated with tumor-infiltrating lymphocytes. Interestingly, overexpression of FGFR1 in the NOG mouse model improved the immunotherapeutic impact of GC, while overexpression of FGFR4 impaired the effect. When combined with an FGFR4 inhibitor, the anti-tumor effect of anti-PD-1 treatment increased significantly in a GC xenograft mouse model with overexpressed FGFR4.Conclusions: FGFRs has critical function in GC and associated with immune cell infiltration, which might be a potential prognosis biomarker and predictor of response to immunotherapy in GC.

The characterization of the full-sized pore structure is important for the evaluation and prediction of the reservoir of shale gas with strong heterogeneity. It is of great scientific significance to explore the pore structure characteristics of overmature coal-bearing shale. Core descriptions, X-ray diffraction (XRD), vitrinite reflectance (Ro), field emission scanning electron microscopy (FE-SEM), high-pressure mercury intrusion porosimetry (MIP), and low-pressure N₂/CO₂ gas adsorption (N₂-/CO₂-GA) experiments were performed on overmature coal-bearing shale samples from the Wuxiang block, south-central Qinshui Basin, China. The results show that the total organic carbon (TOC) ranged from 0.29 to 8.36%, with an average of 3.84%, and the organic matter (OM) is dominated by type III kerogen. The minerals in the shale primarily consist of clay (43–85.5%, averaging 52.1%) and quartz (12.6–61.2%, averaging 43.5%). The major clay minerals are illite-smectite (I/S) and illite, ranging from 22.5 to 55.6% (mean 41.4%) and 8.7–52.7% (mean 32%), respectively. FE-SEM images reveal that intraparticle pores (IntraP pores) and interparticle pores (InterP pores) are widely developed in clay minerals, and organic pores are occasionally present. Mesopores make the greatest contribution to the total pore volume (PV), and micropores are the major contributors to the specific surface area (SSA). Clays are the main controllers of micropore development. Mesopores developed in the clay mineral layers are promoted by I/S but inhibited by illite. Macropores and microfractures are mainly developed in clays and quartz and do not correlate significantly with the TOC, or mineral composition, due to the influence of compaction and cementation. The TOC and minerals affect pore structure characteristics mainly by influencing micropores.