Deciphering tertiary lymphoid structure heterogeneity reveals prognostic signature and therapeutic potentials for colorectal cancer: a multicenter retrospective cohort study.

BACKGROUND: Tertiary lymphoid structures (TLSs) exert a crucial role in the tumor microenvironment (TME), impacting tumor development, immune escape, and drug resistance. Nonetheless, the heterogeneity of TLSs in colorectal cancer (CRC) and their impact on prognosis and treatment response remain unclear. METHODS: We collected genome, transcriptome, clinicopathological information, and digital pathology images from multiple sources. An unsupervised clustering algorithm was implemented to determine diverse TLS patterns in CRC based on the expression levels of 39 TLS signature genes (TSGs). Comprehensive explorations of heterogeneity encompassing mutation landscape, TME, biological characteristics, response to immunotherapy, and drug resistance were conducted using multi-omics data. TLSscore was then developed to quantitatively assess TLS patterns of individuals for further clinical applicability. RESULTS: Three distinct TLS patterns were identified in CRC. Cluster 1 exhibited upregulation of proliferation-related pathways, high metabolic activity, and intermediate prognosis, while Cluster 2 displayed activation of stromal and carcinogenic pathways and a worse prognosis. Both Cluster 1 and Cluster 2 may potentially benefit from adjuvant chemotherapy. Cluster 3, characterized by the activation of immune regulation and activation pathways, demonstrated a favorable prognosis and enhanced responsiveness to immunotherapy. We subsequently employed a regularization algorithm to construct the TLSscore based on 9 core genes. Patients with lower TLSscore trended to prolonged prognosis and a more prominent presence of TLSs, which may benefit from immunotherapy. Conversely, those with higher TLSscore exhibited increased benefits from adjuvant chemotherapy. CONCLUSIONS: We identified distinct TLS patterns in CRC and characterized their heterogeneity through multi-omics analyses. The TLSscore held promise for guiding clinical decision-making and further advancing the field of personalized medicine in CRC.

Clinical, endoscopic, pathological characteristics and management of cap polyposis: experience from a Tertiary Hospital in China.

Background and aims: Cap polyposis (CP) is a rare kind of benign disease, and the majority of previously published relevant articles involve a small number of patients. Hence, we summarized our experience to contribute additional data, hoping to raise awareness of this disease. Methods: From 1 January 2017 to 1 November 2021, consecutive patients diagnosed with CP were retrospectively reviewed. Their medical histories, and laboratory, imaging, endoscopic, and pathology results were analyzed. We made telephone calls to the patients and searched for the information in our electronic medical records to obtain the follow-up results. Results: Forty-one patients were chosen for analysis. The median age of the patients was 20 years old, and 90.24% (37 patients) of the patients were male. The majority of the patients presented with hematochezia. The rectum was the most commonly affected site, and the Helicobacter pylori infection rate was high. There were multiple and combined treatments for these patients. These treatments can be divided into 3 main categories: medical therapy, endotherapy and surgery. Medical therapy helped to diminish the size of but the polyps were difficult to resolve; however, the patients' symptoms could be diminished. Twenty-three patients underwent surgical resection, and 12 patients received endotherapy. We further compared the two methods of polyp resection. Both endotherapy and surgery were safe, and the recurrence risk was not significantly different between the two kinds of therapy (p = 0.321). Conclusion: The clinical improvement of medical treatments was not satisfactory, and endotherapy or surgical resection could remove the polyposis and provide temporary relief, but the recurrence rates were high.

Structure Embedded Nucleus Classification for Histopathology Images.

Nuclei classification provides valuable information for histopathology image analysis. However, the large variations in the appearance of different nuclei types cause difficulties in identifying nuclei. Most neural network based methods are affected by the local receptive field of convolutions, and pay less attention to the spatial distribution of nuclei or the irregular contour shape of a nucleus. In this paper, we first propose a novel polygon-structure feature learning mechanism that transforms a nucleus contour into a sequence of points sampled in order, and employ a recurrent neural network that aggregates the sequential change in distance between key points to obtain learnable shape features. Next, we convert a histopathology image into a graph structure with nuclei as nodes, and build a graph neural network to embed the spatial distribution of nuclei into their representations. To capture the correlations between the categories of nuclei and their surrounding tissue patterns, we further introduce edge features that are defined as the background textures between adjacent nuclei. Lastly, we integrate both polygon and graph structure learning mechanisms into a whole framework that can extract intra and inter-nucleus structural characteristics for nuclei classification. Experimental results show that the proposed framework achieves significant improvements compared to the previous methods. Code and data are made available via https://github.com/lhaof/SENC.

Ulcerative colitis complicated with hypermucinous dysplasia.

Patients with ulcerative colitis are at increased risk for colorectal neoplasia compared to the general population. The risk factors include family history of colorectal cancer, wide extent of colitis, disease duration, cumulative inflammatory burden, and primary sclerosing cholangitis. Here, we report a case of colorectal neoplasia developed in a patient with ulcerative colitis.

Inhibition of FABP5 attenuates inflammatory bowel disease by modulating macrophage alternative activation.

Fatty acid binding protein 5 (FABP5) is an intracellular chaperone of fatty acid molecules that regulates lipid metabolism and cell growth. However, its role in intestinal inflammation remains enigmatic. Through examination of human tissue samples and single-cell data, we observed a significant upregulation of FABP5 within the mucosa of patients afflicted with ulcerative colitis (UC) and Crohn's disease (CD), predominantly localized in intestinal macrophages. Herein, we investigate the regulation of FABP5-IN-1, a FABP5 inhibitor, on various cells of the gut in an inflammatory environment. Our investigations confirmed that FABP5 ameliorates DSS-induced colitis in mice by impeding the differentiation of macrophages into M1 macrophages in vitro and in vivo. Furthermore, following FABP5-IN-1 intervention, we observed a notable restoration of intestinal goblet cells and tuft cells, even under inflammatory conditions. Additionally, FABP5-IN-1 exhibits a protective effect against DSS-induced colitis by promoting the polarization of macrophages towards the M2 phenotype in vivo. In summary, FABP5-IN-1 confers protection against DSS-induced acute colitis through a multifaceted approach, encompassing the reduction of inflammatory macrophage infiltration, macrophage polarization, regulating Th17/Treg cells to play an anti-inflammatory role in IBD. The implications for IBD are underscored by the comprehensive in vivo and in vitro experiments presented in this article, thereby positioning FABP5 as a promising and novel therapeutic target for the treatment of IBD.

A female with recurrent gastrointestinal bleeding: an obscure diminutive lesion requiring a massive blood transfusion.

Obscure gastrointestinal bleeding (OGIB) is bleeding of unknown origin after a negative initial or primary colonoscopy and upper endoscopy result. Small bowel bleeding accounts for 5% of GI bleeding but it is the most prominent cause of OGIB. We present a case with an obscure diminutive polypoid vascular anomaly of small intestine. In this case, intraoperative enteroscopy seems to be the last trump card for OGIB, especially for large amount loss of blood. It not only helped to find the obscure cause for bleeding, but also preserved the small intestine.

Prognostic factors and role of postoperative radiotherapy in surgically resected thymomas.

Objective: To investigate the prognostic factors in and role of postoperative radiotherapy (PORT) for surgically resected thymomas. Methods: A total of 1540 patients with pathologically confirmed thymomas undergoing resection between 2000 and 2018 were identified retrospectively from the SEER (Surveillance, Epidemiology, and End Results) database. Tumors were restaged as local (limited to thymus), regional (invasion to mediastinal fat and other neighboring structures), or distant stage. Disease-specific survival (DSS) and overall survival (OS) were estimated by the Kaplan-Meier method and the log-rank test. Adjusted hazard ratios (HRs) with 95% CIs were calculated by Cox proportional hazards modeling. Results: Tumor stage and histology were independent predictors of both DSS (regional: HR, 3.711; 95% CI, 2.006-6.864; distant: HR, 7.920; 95% CI, 4.061-15.446; type B2/B3: HR, 1.435; 95% CI, 1.008-2.044) and OS (regional: HR, 1.461; 95% CI, 1.139-1.875; distant: HR, 2.551; 95% CI, 1.855-3.509; type B2/B3: HR, 1.409; 95% CI, 1.153-1.723). For patients with regional stage and type B2/B3 thymomas, PORT was associated with better DSS after thymectomy/thymomectomy (HR, 0.268; 95% CI, 0.099-0.727), but the association was not significant after extended thymectomy (HR, 1.514; 95% CI, 0.516-4.44). Among patients with lymph node metastases, those who received PORT (HR, 0.372; 95% CI, 0.146-0.949), chemotherapy (HR, 0.843; 95% CI, 0.303-2.346), or both (HR, 0.296, 95% CI, 0.071-1.236) had a better OS. Conclusions: The extent of invasion and tumor histology were independent predictors of worse survival following surgical resection of thymoma. Patients with regional invasion and type B2/B3 thymoma who undergo thymectomy/thymomectomy may benefit from PORT, while patients with nodal metastases may benefit from multimodal therapy, including PORT and chemotherapy.

An ICD-Associated DAMP Gene signature predicts survival and immunotherapy response of patients with lung adenocarcinoma.

BACKGROUND: While some lung adenocarcinoma (LUAD) patients benefit long-term from treatment with immune checkpoint inhibitors, the sad reality is that a considerable proportion of patients do not. The classification of the LUAD tumor microenvironment (TME) can be used to conceptually comprehend primary resistance mechanisms. In addition, the most recent research demonstrates that the release of damage-associated molecular pattern (DAMP) in TME by immunogenic cell death (ICD) may contribute to the adaptive immune response. Currently, however, there is no such comprehensive research on this topic in LUAD patients. Therefore, we set out to investigate how to reverse the poor infiltration characteristics of immune cells and boost antitumor immunity by identifying DAMP model. METHODS: In this study, ICD-related DAMP genes were selected to investigate their effects on the prognosis of LUAD. To create a risk signature using the TCGA-LUAD cohort, the univariate COX regression and the least absolute shrinkage and selection operator regression were carried out, and the results were verified in a GEO dataset. Subsequently, the multivariate COX regression was applied to establish a prognostic nomogram. And the ESTIMATE and ssGSEA algorithms were utilized to analyze immune activity and the TIDE algorithm was for responsiveness to immunotherapy. Moreover, clinical tissue samples were used to verify the differential expression of 9 DAMP genes in the signature. RESULTS: We identified two distinct DAMP molecular subtypes, and there are remarkable differences in survival probability between the two subtypes, and patients with higher levels of DAMP-related genes are "hot tumors" with increased immune activity. In addition, 9 DAMP genes were selected as prognostic signature genes, and clinical outcomes and immunotherapy response were better for participants in the low-risk group. Importantly, according to the area under the curve (AUC) value in evaluating the efficacy of immunotherapy, this signature is superior to existing predictors, such as PD-L1 and TIDE. CONCLUSIONS: Our study suggests ICD plays an important part in modeling the TME of LUAD patients. And this signature could be utilized as a reliable predictor to estimate clinical outcomes and predict immunotherapy efficacy among LUAD patients.

[Corrigendum] miR­1299/NOTCH3/TUG1 feedback loop contributes to the malignant proliferation of ovarian cancer.

Following the publication of the above article, a concerned reader drew to the authors' attention that the data shown for the 'CAOV3/NC mimics' experiment in Fig. 2D on p. 443 appeared to be the same as that shown for the 'TUG1­sh+miR­1299 inhibitors' experiment in Fig. 4H on p. 444. The authors have examined their original data, and realize that the same data was inadvertently included in the two figures. Consequently, the corrected version of Fig. 2, featuring the correct data for the 'CAOV3/NC mimics' experiment in Fig. 2D, is shown opposite. The overall conclusions of this study were not affected by this error. All the authors agree to the publication of this corrigendum, and are grateful to the Editor of Oncology Reports for allowing them the opportunity to publish this; furthermore, they apologize to the readership for any inconvenience caused. [Oncology Reports 44: 438-448, 2020; DOI: 10.3892/or.2020.7623].

HNRNPA2B1-Mediated MicroRNA-92a Upregulation and Section Acts as a Promising Noninvasive Diagnostic Biomarker in Colorectal Cancer.

MicroRNA-92a (miR-92a) may serve as a novel promising biomarker in multiple cancers, including colorectal cancer (CRC); however, the diagnostic accuracy and the underlying molecular mechanism of miR-92a in CRC is poorly understood. We first carried out meta-analysis and found that serum/plasma miR-92a yield better diagnostic efficacy when compared to stool samples and CRC tissues, and this finding was validated by our independent study through stool sample. Multiple bioinformatics assay indicated that miR-92a expression was positively correlated with heterogeneous nuclear ribonucleoproteins A2/B1 (HNRNPA2B1) expression and closely related with the clinical characteristics of CRC. Experimental evidence showed that knockdown of HNRNPA2B1 could significantly decrease miR-92a expression and secretion in RKO cells. HNRNPA2B1 mediated miR-92a via m6A RNA modification. These findings indicate that HNRNPA2B1-m6A RNA modification-derived MicroRNA-92a upregulation and section from the local CRC acts a candidate noninvasive serum biomarker in colorectal cancer. Our study provides a novel insight into miR-92a mechanisms in relation to both expression and secretion for CRC diagnosis.

Early Results of Thoracic Endovascular Aortic Repair for the Management of Acute Uncomplicated Type B Aortic Dissection.

Historically, optimal medical therapy (OMT) has been the primary therapy for acute uncomplicated type B aortic dissection (auTBAD). However, recent data suggest that OMT provides poor long-term results, and aortic remodeling induced by thoracic endovascular aortic repair (TEVAR) may improve survival. This study compares adverse events and survival among auTBAD patients receiving either TEVAR or OMT. A retrospective analysis identified 146 consecutive auTBAD patients presenting to a single institution between 1/2012 and 10/2020. Patients were divided into 2 groups based upon whether they received TEVAR (n = 50) or OMT (n = 96) at index hospitalization. Major morbidity and survival were compared between groups. 67.1% of patients presented with a Debakey IIIB dissection with maximum thoracic aortic diameter of 4.3 ± 1.0 cm. Over follow-up, 35% of OMT patients failed medical therapy and underwent intervention (n = 23 TEVAR, n = 11 open). An additional 13 died for an all-cause failure rate of 49%. The composite incidence of renal failure, stroke, spinal cord ischemia, and retrograde type A dissections was similar between groups (TEVAR:6.0% vs OMT:4.2%). In-hospital mortality was 0%. Kaplan-Meier analysis demonstrated a trend towards improved survival among the TEVAR group at 1 and 3 years but no difference in overall survival (HR:0.30, 95% CI:0.08-1.08, P = 0.066). Five-year survival was 91% with TEVAR and 82% with OMT. Complete false lumen thrombosis was achieved in 72.1% with TEVAR and 20.0% with OMT (P < 0.001). In experienced centers, there is equivalent early mortality in the treatment of auTBAD with TEVAR compared to OMT. TEVAR provides superior aortic remodeling to OMT in auTBAD, which may translate into improved long-term survival.

Clinical, endoscopic and pathological characteristics of primary gastrointestinal T-cell and NK/T-cell lymphomas.

BACKGROUND AND AIMS: Gastrointestinal T-cell and NK/T-cell lymphomas are relatively rare and may be difficult to diagnose. Therefore, we performed a retrospective study of the clinical, endoscopic and pathological characteristics of these lymphomas, to provide additional data on this issue. METHODS: From April 2013 to April 2021, consecutive patients diagnosed with primary gastrointestinal T-cell and NK/T-cell lymphomas were retrospectively reviewed. Their medical histories, laboratory, imaging, endoscopic, and pathology results were analyzed. RESULTS: Forty-two patients were finally chosen, among whom, 24 patients had ENKTCL, 9 patients had MEITL, 2 patients had ALCL, ALK-, 1 patient had ALCL, ALK+, and 6 patients had PTCL, NOS. The median age of all the patients was 48 years old, and 73.81% (31 patients) were male. The patients' symptoms were abdominal pain, diarrhea, gastrointestinal bleeding, weight loss, fever, and others. The endoscopic results of 26 patients could be traced, and 69.23% of the patients showed multiple lesions. Ulcerative and ulceroinfiltrative lesions were common. Among the pathologic findings, necrosis, ulceration, and crypt atrophy were commonly found while epitheliotropism was relatively less common. Twelve patients (28.57%) had a history of misdiagnosis. After a median follow-up time of 26.9 months, 26 patients (66.70%) died of the disease. The median overall survival time was 8 months. CONCLUSIONS: These lymphomas had nonspecific clinical manifestations, various endoscopic features, and were likely to be misdiagnosed as other diseases. The prognosis is still poor, and more in-depth research is needed to develop more precise treatments.

Is simply washing hands before dialysis procedures sufficient for reducing peritoneal dialysis peritonitis?-A single center study from 2015 to 2020 in Yiwu, China.

INTRODUCTION: The study aimed to improve the qualified rate of hand hygiene and reduce the incidence of peritonitis in peritoneal dialysis (PD) patients. METHODS: A hand hygiene questionnaire was distributed to patients during home visits and outpatient visits in 2015 and 2020. Hand-washing practices were evaluated by collecting cultures from the hands of patients after hand washing, evaluating their household environment, and recording the antimicrobial resistance of pathogenic bacteria. RESULTS: Compared to patients in 2015, patients in 2020 had fewer errors in hand washing (p < 0.05), but the rate of qualification after hand washing was lower (p < 0.01). Furthermore, patients who used hand disinfectants after washing had a higher qualified rate. Coagulase-negative staphylococcus (CNS) was the most common isolated bacteria. From 2015 to 2020, the annual incidence of CNS PD peritonitis did not decrease, while the proportion of methicillin-resistant CNS decreased. CONCLUSION: The use of hand disinfectants after standard hand washing may help reduce the incidence of peritonitis in PD patients.

Identification of a prognostic risk-scoring model and risk signatures based on glycosylation-associated cluster in breast cancer.

Breast cancer is a heterogeneous disease whose subtypes represent different histological origins, prognoses, and therapeutic sensitivity. But there remains a strong need for more specific biomarkers and broader alternatives for personalized treatment. Our study classified breast cancer samples from The Cancer Genome Atlas (TCGA) into three groups based on glycosylation-associated genes and then identified differentially expressed genes under different glycosylation patterns to construct a prognostic model. The final prognostic model containing 23 key molecules achieved exciting performance both in the TCGA training set and testing set GSE42568 and GSE58812. The risk score also showed a significant difference in predicting overall clinical survival and immune infiltration analysis. This work helped us to understand the heterogeneity of breast cancer from another perspective and indicated that the identification of risk scores based on glycosylation patterns has potential clinical implications and immune-related value for breast cancer.

Developing and validating a prognostic nomogram for ovarian clear cell carcinoma patients: A retrospective comparison of lymph node staging schemes with competing risk analysis.

Purpose: Lymph node (LN) involvement is a key factor in ovarian clear cell carcinoma (OCCC) although, there several indicators can be used to define prognosis. This study examines the prognostic performances of each indicator for OCCC patients by comparing the number of lymph nodes examined (TNLE), the number of positive lymph nodes (PLN), lymph node ratio (LNR), and log odds of metastatic lymph nodes (LODDS). Methods: 1,300 OCCC patients who underwent lymphadenectomy between 2004 and 2015 were extracted from the Surveillance Epidemiology and End Results (SEER) database. Primary outcomes were Overall Survival (OS) and the cumulative incidence of Cancer-Specific Survival (CSS). Kaplan-Meier's and Fine-Gray's tests were implemented to assess OS and CSS rates. After conducting multivariate analysis, nomograms using OS and CSS were constructed based upon an improved LN system. Each nomograms' performance was assessed using Receiver Operating Characteristics (ROC) curves, calibration curves, and the C-index which were compared to traditional cancer staging systems. Results: Multivariate Cox's regression analysis was used to assess prognostic factors for OS, including age, T stage, M stage, SEER stage, and LODDS. To account for the CSS endpoint, a proportional subdistribution hazard model was implemented which suggested that the T stage, M stage, SEER stage, and LNR are all significant. This enabled us to develop a LODDS-based nomogram for OS and a LNR-based nomogram for CSS. C-indexes for both the OS and CSS nomograms were higher than the traditional American Joint Committee on Cancer (AJCC), 8th edition, staging system. Area Under the Curve (AUC) values for predicting 3- and 5-year OS and CSS between nomograms also highlighted an improvement upon the AJCC staging system. Calibration curves also performed with consistency, which was verified using a validation cohort. Conclusions: LODDS and LNR may be better predictors than N stage, TNLE, and PLNs. For OCCC patients, both the LODDS-based and LNR-based nomograms performed better than the AJCC staging system at predicting OS and CSS. However, further large sample, real-world studies are necessary to validate the assertion.

Plasma Cell-Free DNA as a Novel Biomarker for the Diagnosis and Monitoring of Atherosclerosis.

Atherosclerosis (AS) is the leading cause of cardiovascular diseases (CVDs) with a high rate of mortality worldwide. Plasma cell-free DNA (cfDNA), mainly originating from apoptosis, necrosis, and active secretion, has been recognized as a promising biomarker for the diagnosis and prognosis of multiple cancers, whereas there are no reports about cfDNA in CVDs. Here, we found an increased quantity and decreased integrity of cfDNA (cfDI) in the serum from AS patients compared with normal controls. Moreover, the reduced cfDI is inversely correlated with serum LDL levels, carotid plaque size, and carotid plaque thickness in the progression of AS. Consistently, in vivo experiments confirmed that the release and cleavage of cfDNA were increased concomitantly with the development and progression of AS in ApoE-/- mice. Our study sheds light on the potential of cfDNA and cfDI as molecular biomarkers for detecting and monitoring AS.

Pyroptosis correlates with tumor immunity and prognosis.

Pyroptosis, as a proinflammatory form of regulated cell death, plays an important role in multiple cancers. However, the diagnostic and prognostic values of pyroptosis and its interaction with tumor immunity in pan-cancer are still unclear. Here, we show an elevated general expression of 17 pyroptosis-associated genes of tumor patients with high-immune-activity and a reduced pyroptosis in low-immune-activity tumors. Moreover, pyroptosis is positively correlated with immune infiltration and immune-related signatures across 30 types of cancer. Furthermore, our experimental data suggest that pyroptosis directly modulate the expression of immune checkpoint molecules and cytokines. We generate a pyroptosis score model as a potential independent prognostic indicator in melanoma patients. Interestingly, 3 of pyroptosis-associated genes including CASP1, CASP4 and PYCARD, can predict the effectiveness of anti-PD-1 immunotherapy for patients with melanoma. Our study demonstrates that pyroptosis correlates with tumor immunity and prognosis, might be used as a potential target for immune therapy.

Ferroptosis Associates With Diagnosis and Prognosis by Promoting Antitumor Immune Response in Melanoma.

Immunotherapy has greatly improved the clinical benefits of cancer treatment, especially in melanoma. Ferroptosis is a novel mechanism of cell death which relates to immunity. This study aimed at understanding the potential link between ferroptosis and cancer immunocompetent in melanoma using multiple bioinformatics analyses. By the WGCNA assay, we first constructed a key module-gene of ferroptosis, which was strongly correlated with the diagnosis, prognosis, and infiltration of immune cells in melanoma. The elevated module-gene could effectively distinguish melanoma from normal tissues and acted as a good prognostic marker. The module-gene of ferroptosis was positively correlated with the infiltration of immune cells. In particular, the module was positively correlated with the expression of PD-L1 and sensitively increased after effective anti-PD-1 treatment. Furthermore, the differential expression of the module-gene between normal and tumor tissues was observed in pan-cancer. The similarity correlations of the module-gene with infiltration of immune cells and the expressions of PD-L1 were confirmed in the pan-cancer level. Our study demonstrated that the key module-gene of ferroptosis was closely related with diagnosis, prognosis, and anti-immune response in melanoma, as well as in pan-cancer.

Deep Learning Model for Predicting the Pathological Complete Response to Neoadjuvant Chemoradiotherapy of Locally Advanced Rectal Cancer.

Objective: This study aimed to develop an artificial intelligence model for predicting the pathological complete response (pCR) to neoadjuvant chemoradiotherapy (nCRT) of locally advanced rectal cancer (LARC) using digital pathological images. Background: nCRT followed by total mesorectal excision (TME) is a standard treatment strategy for patients with LARC. Predicting the PCR to nCRT of LARC remine difficulty. Methods: 842 LARC patients treated with standard nCRT from three medical centers were retrospectively recruited and subgrouped into the training, testing and external validation sets. Treatment response was classified as pCR and non-pCR based on the pathological diagnosis after surgery as the ground truth. The hematoxylin & eosin (H&E)-stained biopsy slides were manually annotated and used to develop a deep pathological complete response (DeepPCR) prediction model by deep learning. Results: The proposed DeepPCR model achieved an AUC-ROC of 0.710 (95% CI: 0.595, 0.808) in the testing cohort. Similarly, in the external validation cohort, the DeepPCR model achieved an AUC-ROC of 0.723 (95% CI: 0.591, 0.844). The sensitivity and specificity of the DeepPCR model were 72.6% and 46.9% in the testing set and 72.5% and 62.7% in the external validation cohort, respectively. Multivariate logistic regression analysis showed that the DeepPCR model was an independent predictive factor of nCRT (P=0.008 and P=0.004 for the testing set and external validation set, respectively). Conclusions: The DeepPCR model showed high accuracy in predicting pCR and served as an independent predictive factor for pCR. The model can be used to assist in clinical treatment decision making before surgery.

Preoperative Prediction of Lymph Node Metastasis in Colorectal Cancer with Deep Learning.

Objective. To develop an artificial intelligence method predicting lymph node metastasis (LNM) for patients with colorectal cancer (CRC). Impact Statement. A novel interpretable multimodal AI-based method to predict LNM for CRC patients by integrating information of pathological images and serum tumor-specific biomarkers. Introduction. Preoperative diagnosis of LNM is essential in treatment planning for CRC patients. Existing radiology imaging and genomic tests approaches are either unreliable or too costly. Methods. A total of 1338 patients were recruited, where 1128 patients from one centre were included as the discovery cohort and 210 patients from other two centres were involved as the external validation cohort. We developed a Multimodal Multiple Instance Learning (MMIL) model to learn latent features from pathological images and then jointly integrated the clinical biomarker features for predicting LNM status. The heatmaps of the obtained MMIL model were generated for model interpretation. Results. The MMIL model outperformed preoperative radiology-imaging diagnosis and yielded high area under the curve (AUCs) of 0.926, 0.878, 0.809, and 0.857 for patients with stage T1, T2, T3, and T4 CRC, on the discovery cohort. On the external cohort, it obtained AUCs of 0.855, 0.832, 0.691, and 0.792, respectively (T1-T4), which indicates its prediction accuracy and potential adaptability among multiple centres. Conclusion. The MMIL model showed the potential in the early diagnosis of LNM by referring to pathological images and tumor-specific biomarkers, which is easily accessed in different institutes. We revealed the histomorphologic features determining the LNM prediction indicating the model ability to learn informative latent features.