The evolutionarily conserved Hippo signaling pathway is a key regulator of stem cell self-renewal, differentiation, and organ size. While alterations in Hippo signaling are causally linked to uncontrolled cell growth and a broad range of malignancies, genetic mutations in the Hippo pathway are uncommon and it is unclear how the tumor suppressor function of the Hippo pathway is disrupted in human cancers. Here, we report a novel epigenetic mechanism of Hippo inactivation in the context of hepatocellular carcinoma (HCC). We identify a member of the microrchidia (MORC) protein family, MORC2, as an inhibitor of the Hippo pathway by controlling upstream Hippo regulators, neurofibromatosis 2 (NF2) and kidney and brain protein (KIBRA). Mechanistically, MORC2 forms a complex with DNA methyltransferase 3A (DNMT3A) at the promoters of NF2 and KIBRA, leading to their DNA hyper-methylation and transcriptional repression. As a result, NF2 and KIBRA are crucial targets of MORC2 to regulate confluence-induced activation of Hippo signaling and contact inhibition of cell growth under both physiological and pathological conditions. The MORC2-NF2/KIBRA axis is critical for maintaining self-renewal, sorafenib resistance, and oncogenicity of HCC cells in vitro and in nude mice. Furthermore, MORC2 expression is elevated in HCC tissues, associated with stem-like properties of cancer cells, and disease progression in patients. Collectively, MORC2 promotes cancer stemness and tumorigenesis by facilitating DNA methylation-dependent silencing of Hippo signaling and could be a potential molecular target for cancer therapeutics.

Pathological image classification is critical for early cancer diagnosis and precise subtyping. However, pathological images exhibit significant heterogeneity and complex textures. Existing methods often fail to fully exploit local details, while frequency-domain approaches lack effective inter-subband interaction, hindering the fusion of global context and local fine-grained information. To address this, this study propose a Multi-scale Frequency-domain Hybrid Attention mechanism (MFHA). MFHA uses wavelet transform to decompose images into low-frequency subbands (global structure) and high-frequency subbands (microscopic details). By integrating multi-scale convolutions with subband fusion, it enhances high-frequency feature representation, enabling effective joint modeling of global structure and local texture. This study further introduce a spatial attention module with cosine similarity and multi-dimensional statistics to improve feature robustness and discriminability. Experiments show that our method outperforms baselines on multiple pathological datasets, with accuracy gains of 1.97% and 2.71%. This work provides a novel perspective for pathological image feature co-modeling, boosting classification accuracy and robustness. The code is available at https://github.com/AnaStartz/WSI-processing-framework.

Background:Lung adenocarcinoma (LUAD) is one of the most prevalent types of lung cancer globally; it is characterized by high incidence and mortality rates and contributes to over 1.8 million deaths annually. PSMB7, a crucial component of the 20S proteasome involved in protein degradation and antigen presentation, has been implicated in various cancers; however, its specific function in LUAD remains inadequately explored. Methods:This research aimed to investigate the expression of PSMB7 in LUAD and its clinical significance using real-time quantitative PCR, immunohistochemistry, differential expression analysis, pathway enrichment analysis, immune cell infiltration, and DNA methylation. Results:PSMB7 expression levels in LUAD tissues were considerably higher than those in the surrounding normal lung tissues and were associated with advanced pathological stages and poorer clinical outcomes. High PSMB7 expression was correlated with reduced overall and disease-specific survival. Functional enrichment analysis indicated that the differentially expressed genes associated with PSMB7 were mainly involved in protein-DNA complex assembly and chromatin remodeling. Moreover, LUAD tissues showed lower DNA methylation in PSMB7 promoters than that in normal lung tissues, which was correlated with reduced survival rates. A negative correlation was observed between PSMB7 levels and immune cell infiltration, particularly for effector memory T, B, follicular helper T, and mast cells. Conclusions:We identified PSMB7 as a promising biomarker for LUAD prognosis because of its strong association with tumor progression and immune microenvironment modulation. Future studies should explore therapeutic strategies targeting PSMB7 to improve patient outcomes for LUAD.

The development of lung squamous cell carcinoma (LUSC) is associated with the intratumoral microbiota. To facilitate faster clinical decisions and predict patient prognosis, we constructed an intratumoral microbial abundance prognostic scoring (MAPS) model for LUSC and analyzed its prognostic performance. Data on the LUSC tumor microbiome, patient survival, and clinical information were downloaded from The Cancer Microbiome Atlas and The Cancer Genome Atlas databases. Differentially abundant microbial genera in LUSC tumors were analyzed, and their prognostic value was evaluated. The differential abundance of key genera in the MAPS model was validated using lung adenocarcinoma (LUAD) tumors and normal tissues. Of 52 microbial genera with increased abundance and 437 with decreased abundance in LUSC tumors, 462 were highly related to the disease. Seven of 13 genera that were significantly related to prognosis were selected to construct the MAPS model. The MAPS risk grouping was identified as a prognostic risk factor for LUSC. Among the seven genera in the MAPS model, Indibacter, Oceanospirillum, Thalassomonas, and Thermopetrobacter differed in abundance between LUAD tumors and normal tissues and may be the key intratumoral microorganisms involved in LUSC and LUAD development. In conclusion, our MAPS model may be a powerful prognostic biomarker for LUSC.

Hemangioblastoma is an uncommon tumor of uncertain histogenesis, primarily found in the central nervous system. However, extraneural cases have been reported in visceral organs such as the kidneys, pancreas, peritoneum, and liver. Hemangioblastoma occurring in the gastrointestinal tract is extremely rare, with only 4 cases can be retrieved. Here, we presented a case of rectal hemangioblastoma. A 59-year-old woman sought medical attention for altered stool consistency and the colonoscopy revealed a submucosal lesion in the rectum. Subsequently, she underwent endoscopic submucosal dissection. Grossly, the excised lesion was a broad-based polypoid mass measuring 1.2 × 0.8 × 0.8 cm, with a grayish-white appearance and no signs of hemorrhage or necrosis. Histologically, the tumor consisted of a rich network of thin-walled blood vessels interspersed with vesicular neoplastic cells, with minimal mitotic activity. Immunohistochemistry found that neoplastic cells were positive for D2-40, and β-catenin showed normal membranous staining pattern. No pathogenic mutations in the VHL gene were detected by NGS in this sample. However, we found genetic alterations in other genes potentially associated with the pathogenesis of hemangioblastoma, such as TSC, SDH and PTEN. Based on the above findings, the diagnosis of hemangioblastoma was made. The differential diagnosis included hemangioma, lymphangioma, neuroendocrine tumor, metastatic renal cell carcinoma, perivascular epithelioid cell tumor, gastrointestinal stromal tumor, and well-differentiated liposarcoma. This case underscored the importance of considering hemangioblastoma in the differential diagnosis of rectal submucosal lesions. It highlighted the need for a thorough diagnostic approach that integrates colonoscopic evaluation with histopathological examination, particularly in patients presenting with altered stool consistency or melena. Furthermore, the NGS results implicated that TSC, SDH and PTEN may contribute to hemangioblastoma development. These alterations could constitute a novel diagnostic signature, a premise that warrants definitive investigation through larger, multi-institutional studies.

Background Our objective was to elucidate the pivotal roles of BCAR1 in unfolded protein response (UPR), autophagy and vasculogenic mimicry (VM) formation, processes that essential for the metastasis of lung adenocarcinoma (LUAD) cells. Methods The morphological assessment of endoplasmic reticulum (ER) status and autolysosomes in H1975 and H1299 LUAD cells following BCAR1 knockout (KO) was conducted using transmission electron microscope. The expression of markers and cellular functions related to the UPR, autophagy, and VM formation were examined in LUAD cells tissues. Additionally, proteomic analysis of LUAD cells was performed via mass spectrometry, and the pertinent signaling pathways were analyzed using bioinformatics tools. Results BCAR1-KO inhibited autophagy and UPR induced triggered starvation in LUAD cells. Cleaved-ATF6a-mediated UPR and subsequent autophagy, enhanced by BCAR1, were confirmed using the UPR stimulator and blocker. High BCAR1 expression, along with elevated UPR and autophagy, predicts poor prognosis in LUAD patients. BCAR1-KO reduced tube formation and VM markers expressions in LUAD cells. Additionally, BCAR1 expression positively correlated with VM formation in BALB/c-nu mice xenografts and LUAD patient tissues. Conclusion BCAR1 promotes LUAD metastasis by enhancing cancer cell survival in nutrient-poor environments through ATF6-mediated UPR activation and autophagy. As BCAR1 induces VM formation, metastatic lesions eventually colonize. Thus, BCAR1 is a promising anti-metastasis target.

Pathologic grading of laryngeal squamous cell carcinoma (LSCC) plays a crucial role in diagnosis, prognosis, and migration. However, the grading performance and interpretability of the intelligent grading model based on LSCC low magnification images are poor. This is because it lacks the delicate nuclear information and information more relevant to grading contained in the high magnification images labeled by pathologists. Yet, low magnification images have information such as tissue texture and contours. Thus, we proposed an end-to-end transformer network with manifold adversarial multi-modal learning (MamlFormer). It effectively fuses and learns LSCC high and low magnification pathology image modalities. Firstly, we demonstrate the feasibility and sufficient conditions for modal fusion of LSCC high and low magnification images from Hoeffding's inequality and multimodal co-regularization. Secondly, we design a new manifold block. It constructs the manifold subspace by some principles. Those principles are divisibility, recoverability, and local distance closest of the feature matrix before and after the mapping of the LSCC each magnification image modalities. Meanwhile it can well solve the problems of redundant feature matrix information and weak modal semantic consistency after multimodal learning. Thirdly, we utilize the encoder and the adversarial loss function to implement adversarial block. It can adaptively learn the latent metrics of the modal distributions of LSCC high and low magnification images. Therefore, it also enhances the complementarity of LSCC high and low magnification image modalities. Then, numerous experiments show that MamlFormer outperforms other SOTA models in both grading performance and interpretability. Finally, we also performed generalization experiments on highly prevalent cervix squamous cell carcinoma. The MamlFormer over is superior to other SOTA models in terms of grading performance and interpretability. This indicates its excellent generalization performance and clinical practicability.

Background: Detecting programmed death ligand 1 (PD-L1) expression based on immunohistochemical (IHC) staining is an important guide for the treatment of lung cancer with immune checkpoint inhibitors. However, this method has problems such as high staining costs, tumor heterogeneity, and subjective differences among pathologists. Therefore, the application of deep learning models to segment and quantitatively predict PD-L1 expression in digital sections of Hematoxylin and eosin (H&E) stained lung squamous cell carcinoma is of great significance. Methods: We constructed a dataset comprising H&E-stained digital sections of lung squamous cell carcinoma and used a Transformer Unet (TransUnet) deep learning network with an encoder-decoder design to segment PD-L1 negative and positive regions and quantitatively predict the tumor cell positive score (TPS). Results: The results showed that the dice similarity coefficient (DSC) and intersection overunion (IoU) of deep learning for PD-L1 expression segmentation of H&E-stained digital slides of lung squamous cell carcinoma were 80 and 72%, respectively, which were better than the other seven cutting-edge segmentation models. The root mean square error (RMSE) of quantitative prediction TPS was 26.8, and the intra-group correlation coefficients with the gold standard was 0.92 (95% CI: 0.90-0.93), which was better than the consistency between the results of five pathologists and the gold standard. Conclusion: The deep learning model is capable of segmenting and quantitatively predicting PD-L1 expression in H&E-stained digital sections of lung squamous cell carcinoma, which has significant implications for the application and guidance of immune checkpoint inhibitor treatments. And the link to the code is https://github.com/Baron-Huang/PD-L1-prediction-via-HE-image.

PURPOSE:Upper tract urothelial carcinoma (UTUC) and urothelial carcinoma of the bladder (UCB) share histomorphological and therapeutic features but distinct epidemiologic and clinicopathologic characteristics. We examined alterations of chromatin regulatory genes in molecular subtypes, clonal relatedness, and T-cell receptor (TCR) diversity in UTUC and UCB. MATERIALS AND METHODS:Targeted next-generation sequencing or whole-exome DNA sequencing and TCR sequencing were conducted with 34 UTUC and 49 UCB specimens from 63 patients. Tumors were subtyped based on the expression of CK5 and GATA3. Results of tissue microarray of 78 muscle-invasive bladder cancer (MIBC) samples were used as prognostic factors of different subtypes of MIBC. RESULTS:Chromatin regulatory genes were frequently mutated in both UTUC and UCB. Rapid relapse and progression of non-MIBC are correlated with alterations of KMT2C and EP300. Frequency of alterations in chromatin regulatory genes is higher in UTUC patients with SBS22 and SBS2 signatures and lower in UCB patients with SBS2 and SBS6 signatures. GATA3 and CK5 double-positive patients with higher frequencies of SMARCA4, ARID1A, and EP300 mutations have better prognoses than patients with basal subtypes. Although UTUC and UCB in the same patient can be either clonally related or developed independently, mutated genes in chromatin pathway were enriched in the related clones. Compared to UTUC, UCB had more deleterious mutations in DNA damage repair (DDR) genes, higher levels of tumor mutation burden (TMB) and copy number variations (CNVs), as well as higher TCR clonality and lower TCR diversity. CONCLUSIONS:Since genetic alterations of the chromatin pathway genes are important in both UTUC and UCB, they could serve as potential biomarkers for predicting disease progression and therapeutic targets. Differences in mutation frequencies of DDR pathway, TMB, CNV, and TCR might be the contributing factors for the distinct responses to immune checkpoint inhibitor (ICI) between UTUC and UCB.

The quantitative analysis of Programmed death-ligand 1 (PD-L1) via Immunohistochemical (IHC) plays a crucial role in guiding immunotherapy. However, IHC faces challenges, including high costs, time consumption and result variability. Conversely, Hematoxylin-Eosin (H&E) staining offers cost-effectiveness, speed, and stable results. Nonetheless, H&E staining, which solely visualizes cellular morphological features, lacks clinical applicability in detecting biomarker expressions like PD-L1. Substituting H&E staining for IHC in determining PD-L1 status is a clinically significant and challenging task. Motivated by above observations, we propose a Multi-Task supervised learning (MTSL)-based connectivity region attention network (MCRANet) for PD-L1 status segmentation in H&E stained images. To reduce interference from non-tumor areas, the MTSL-based region attention is proposed to enhances the network's capability to distinguish between tumor and non-tumor regions. Consequently, this augmentation further improves the network's segmentation efficacy for PD-L1 positive and negative regions. Furthermore, the PD-L1 expression regions demonstrate interconnection throughout the tissue section. Leveraging this topological prior knowledge, we integrate a connectivity modeling module (CM module) within the MTSL-based region attention module (MRA module) to enhance the precision of MTSL-based region attention localization. This integration further improves the structural similarity between the segmentation results and the ground truth. Extensive visual and quantitative results demonstrate that our supervised-learning-guided MRA module produces more interpretable attention and the introduced CM module provides accurate positional attention to the MRA module. Compared to other state-of-the-art networks, MCRANet exhibits superior segmentation performance with a dice similarity coefficient (DSC) of 79.6 % on the lung squamous cell carcinoma (LUSC) PD-L1 status dataset.

Grading laryngeal squamous cell carcinoma (LSCC) based on histopathological images is a clinically significant yet challenging task. However, more low-effect background semantic information appeared in the feature maps, feature channels, and class activation maps, which caused a serious impact on the accuracy and interpretability of LSCC grading. While the traditional transformer block makes extensive use of parameter attention, the model overlearns the low-effect background semantic information, resulting in ineffectively reducing the proportion of background semantics. Therefore, we propose an end-to-end network with transformers constrained by learned-parameter-free attention (LA-ViT), which improve the ability to learn high-effect target semantic information and reduce the proportion of background semantics. Firstly, according to generalized linear model and probabilistic, we demonstrate that learned-parameter-free attention (LA) has a stronger ability to learn highly effective target semantic information than parameter attention. Secondly, the first-type LA transformer block of LA-ViT utilizes the feature map position subspace to realize the query. Then, it uses the feature channel subspace to realize the key, and adopts the average convergence to obtain a value. And those construct the LA mechanism. Thus, it reduces the proportion of background semantics in the feature maps and feature channels. Thirdly, the second-type LA transformer block of LA-ViT uses the model probability matrix information and decision level weight information to realize key and query, respectively. And those realize the LA mechanism. So, it reduces the proportion of background semantics in class activation maps. Finally, we build a new complex semantic LSCC pathology image dataset to address the problem, which is less research on LSCC grading models because of lacking clinically meaningful datasets. After extensive experiments, the whole metrics of LA-ViT outperform those of other state-of-the-art methods, and the visualization maps match better with the regions of interest in the pathologists' decision-making. Moreover, the experimental results conducted on a public LSCC pathology image dataset show that LA-ViT has superior generalization performance to that of other state-of-the-art methods.

Supplementary tables from S1-S6.

The clinical outcome of resectable non-small-cell lung cancer (NSCLC) patients receiving neoadjuvant chemoimmunotherapy is good but varies greatly. In addition, the pathological response after neoadjuvant chemoimmunotherapy is significantly associated with survival outcomes. The aim of this retrospective study was to identify which population of patients with locally advanced and oligometastatic NSCLC has a favorable pathological response after neoadjuvant chemoimmunotherapy. NSCLC patients treated with neoadjuvant chemoimmunotherapy were enrolled between February 2018 and April 2022. Data on clinicopathological features were collected and evaluated. Multiplex immunofluorescence was performed on pre-treatment puncture specimens and surgically resected specimens. In total, 29 patients with stages III and IV locally advanced or oligometastatic NSCLC who received neoadjuvant chemoimmunotherapy and R0 resection were enrolled. The results showed that 55% (16/29) of patients had a major pathological response (MPR) and 41% (12/29) of patients had a complete pathological response (pCR). In the stroma area of the pre-treatment specimen, the higher infiltration of CD3+ PD-L1+ tumor-infiltrating lymphocytes (TILs) and the lower infiltration of CD4+ and CD4+ FOXP3+ TILs were more likely to appear in patients with pCR. However, in the tumor area, the higher infiltration of CD8+ TILs was more likely to appear in patients with non-MPR. In the post-treatment specimen, we found increased infiltration of CD3+ CD8+ , CD8+ GZMB+ , and CD8+ CD69+ TILs and decreased infiltration of PD-1+ TILs both in the stroma and tumor areas. Neoadjuvant chemoimmunotherapy achieved an MPR rate of 55% and induced greater immune infiltration. In addition, we observed that the baseline TILs and their spatial distribution correlate to the pathological response.

Rhabdomyosarcoma (RMS) with TFCP2 rearrangement has been identified recently. This entity has a distinctive clinicopathologic features: a rapidly aggressive clinical course, a preference for the craniofacial bones, a spindle and epithelioid histomorphology, and positive immunohistochemistry for epithelial markers, ALK, and myogenic markers. RMS with TFCP2 rearrangement is rare and may be misdiagnosed as other spindle cell tumors. Here, we report a case of this entity arising in the mandible, which was initially diagnosed as ossifying fibroma in primary tumor in another hospital. A 26-year-old man presented with a recurred mass in the mandible for 1 month after the operation of mandibular tumor. The first excisional specimen was initially diagnosed as ossifying fibroma in another hospital. Histopathologic examination revealed the tumor with a hybrid spindle cell and epithelioid cytomorphology, spindle cells and spindle-to-epithelioid cells with eosinophilic and rich cytoplasm, with high-grade features, prominent nucleoli and some atypical mitosis. Immunohistochemical analysis revealed positivity for desmin, MYOD1, pan-keratin, ALK (5A4), ALK (D5F3). Based on the morphology and immunophenotype, molecular studies were performed, which revealed a FUS::TFCP2 fusion transcript, confirming the diagnosis of Rhabdomyosarcoma with FUS::TFCP2 fusion. Making a correct diagnosis is primarily dependent on awareness by the pathologist of this rare subtype of RMS and careful histopathological evaluation, supported by immunohistochemical and molecular analysis, to avoid potential diagnostic pitfalls.

Anaplastic lymphoma kinase-tyrosine kinase inhibitors (ALK-TKI; ALKi) have shown potent antitumor activity in metastatic non-small-cell lung cancer (NSCLC) with ALK rearrangement (ALK+); however, their efficacy in neoadjuvant settings has been poorly explored. This retrospective study aimed to examine the clinical activity and tumor immune microenvironment (TIME) changes of neoadjuvant ALKi therapy. ALK+ NSCLC patients treated with neoadjuvant ALKi at three hospitals in China between February 2018 and January 2023 were assessed. Data on clinical features and radiographic and pathological responses were collected and evaluated. Multiplex immunofluorescence was performed on pretreatment biopsy specimens and surgically resected specimens to investigate the impact of ALKi on TIME. A total of 12 patients with stage IIA–IIIB NSCLC who received neoadjuvant ALKi therapy were analyzed. The objective response rate was 91.7

Background and objective:Preoperative transcatheter rectal arterial chemoembolization (TRACE) can enhance the pathological response rate in some patients with locally advanced rectal cancer (LARC). However, how to accurately identify patients who can benefit from this neoadjuvant modality therapy remains to be further studied. Deficient mismatch repair (dMMR) protein plays a crucial role in maintaining genome stability. A proportion of patients with rectal cancer are caused by the loss of mismatch repair (MMR) protein. Given the role of MMR in guiding the efficacy in patients with colorectal carcinoma (CRC), this study is designed to evaluate the effect of dMMR status on the response to neoadjuvant therapy through a retrospective analysis.Methods:We launched a retrospective study. First, we selected patients with LARC from the database, and these patients had received preoperative TRACE combined with concurrent chemoradiotherapy. Then, the tumor tissue biopsied by colonoscopy before intervention was taken for immunohistochemistry. According to the expression of MLH-1, MSH-2, MSH-6 and PMS-2, these patients were divided into dMMR protein group and proficient MMR (pMMR) protein group. All patients underwent pathological examination at the end of neoadjuvant therapy, either surgically excised tissue or colonoscopically biopsied tissue. The end point was the pathologic complete response (pCR) after TRACE combined with concurrent chemoradiotherapy.Results:From January 2013 to January 2021, a total of 82 patients with LARC received preoperative TRACE combined with concurrent chemoradiotherapy, and the treatment was well tolerated. Among 82 patients, there were 42 patients in the pMMR group and 40 patients in the dMMR group. 69 patients returned to the hospital for radical resection. In 8 patients, the colonoscopy showed good tumor regression grade after 4 weeks of interventional therapy and refused surgery. The remaining five patients were neither surgically treated nor reexamined by colonoscopy. 77 patients were eventually enrolled in the study. Individually, the pCR rates of these two groups (10%, 4/40 vs. 43%, 16/37) showed significant difference (P < 0.05). Biomarker analysis indicated that patients with dMMR protein had a better propensity for pCR.Conclusion:In patients with LARC, preoperative TRACE combined with concurrent chemoradiotherapy showed good pCR rates, especially in patients with dMMR. Patients with MMR protein defects have a better propensity for pCR.

基于显微成像技术的肿瘤分级对于乳腺癌诊断和预后有着重要的意义,且诊断结果需具备高精度和可解释性.目前,集成Attention的CNN模块深度网络归纳偏差能力较强,但可解释性较差;而基于ViT块的深度网络其可解释性较好,但归纳偏差能力较弱.本文通过融合ViT块和集成Attention的CNN块,提出了一种端到端的自适应模型融合的深度网络.由于现有模型融合方法存在负融合现象,无法保证ViT块和集成Attention的CNN块同时具有良好的特征表示能力;另外,两种特征表示之间相似度高且冗余信息多,导致模型融合能力较差.为此,本文提出一种包含多目标优化、自适应特征表示度量和自适应特征融合的自适应模型融合方法,有效地提高了模型的融合能力.实验表明本文模型的准确率达到95.14％,相比ViT-B/16提升了9.73％,比FABNet提升了7.6％;模型的可视化图更加关注细胞核异型的区域(例如巨型核、多形核、多核和深色核),与病理专家所关注的区域更加吻合.整体而言,本文所提出的模型在精度和可解释性上均优于当前最先进的(state of the art)模型.

Supplementary Figure S2. Doxycycline treatment induces miR-663 expression in GBM cells transfected with miR-663 inducible lentiviral vectors.

The tumor grading of laryngeal cancer pathological images needs to be accurate and interpretable. The deep learning model based on the attention mechanism-integrated convolution (AMC) block has good inductive bias capability but poor interpretability, whereas the deep learning model based on the vision transformer (ViT) block has good interpretability but weak inductive bias ability. Therefore, we propose an end-to-end ViT-AMC network (ViT-AMCNet) with adaptive model fusion and multiobjective optimization that integrates and fuses the ViT and AMC blocks. However, existing model fusion methods often have negative fusion: 1). There is no guarantee that the ViT and AMC blocks will simultaneously have good feature representation capability. 2). The difference in feature representations learning between the ViT and AMC blocks is not obvious, so there is much redundant information in the two feature representations. Accordingly, we first prove the feasibility of fusing the ViT and AMC blocks based on Hoeffding's inequality. Then, we propose a multiobjective optimization method to solve the problem that ViT and AMC blocks cannot simultaneously have good feature representation. Finally, an adaptive model fusion method integrating the metrics block and the fusion block is proposed to increase the differences between feature representations and improve the deredundancy capability. Our methods improve the fusion ability of ViT-AMCNet, and experimental results demonstrate that ViT-AMCNet significantly outperforms state-of-the-art methods. Importantly, the visualized interpretive maps are closer to the region of interest of concern by pathologists, and the generalization ability is also excellent. Our code is publicly available at https://github.com/Baron-Huang/ViT-AMCNet.