A CT-based Machine Learning Radiomics Model for the Prediction of Gastric Cancer Differentiation and Mechanism Exploration

A CT-based Machine Learning Radiomics Model for the Prediction of Gastric Cancer Differentiation and Mechanism Exploration | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article A CT-based Machine Learning Radiomics Model for the Prediction of Gastric Cancer Differentiation and Mechanism Exploration Rixin Su, Yu Zhang, Jie Cao, Fangfang Chen, Xuemeng Li, Ping Li, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7782471/v1 This work is licensed under a CC BY 4.0 License Status: Under Revision Version 1 posted 11 You are reading this latest preprint version Abstract Objectives To develop a CT-based radiomics model for predicting tumor differentiation in patients with gastric cancer. Exploring Rad-score correlation with gene expression and related mechanisms. Materials and Methods Clinical data and imaging of 162 gastric cancer patients were retrospectively analyzed. Patients were randomly allocated to training and validation cohorts. The least absolute shrinkage and selection operator (LASSO) methods were utilized to identify characteristics and develop the Rad-score. Clinical-radiomics models were developed and evaluated for predictive efficacy and clinical incremental value. Screening hub genes and exploring the pathways of hub genes through machine learning, bioinformatics analysis and experimental validation. Results Clinical-radiomics models based on N stage, M stage and Rad-score were developed. The receiver operating characteristic (ROC) curves indicated that the model had good predictive accuracy in the training (AUC = 0.872) and validation groups (AUC = 0.935). The calibration curves indicated a strong correlation between the observed values and the predicted outcomes. The decision curve analysis demonstrated a substantial net benefit associated with the clinical-radiomics model. The clinical impact curve (CIC) illustrated the effective clinical applicability of this model. Analysis of the sequencing data revealed that the key gene IGHG1 was significantly associated with Rad-score. The possible mechanisms are related to the TGF-β signaling, epithelial-mesenchymal transition and KRAS signaling pathway. Conclusions The predictive model based on N stage, M stage and Rad-score can effectively predict the differentiation in gastric cancer patients. Radiomics enables noninvasive prediction of tumor differentiation status while elucidating the expression levels of the IGHG1 and the underlying pathway. gastric cancer radiomics machine learning differentiation mechanism Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 INTRODUCTION Gastric cancer ranks as the fifth most prevalent malignancy worldwide, with approximately 970,000 new cases diagnosed annually, accounting for 4.8% of all cancer cases. It represents the fourth leading cause of cancer-related mortality ( 1 ). The differentiation status of gastric cancer significantly influences therapeutic decision-making and prognostic outcomes, with poorly differentiated subtypes typically associated with an unfavorable prognosis ( 2 ). Early identification of tumor differentiation grade is crucial for implementing timely clinical interventions and improving patient outcomes. Currently, preoperative assessment of gastric cancer differentiation primarily relies on endoscopic biopsy. However, this technique is limited by its superficial sampling of mucosal layers ( 3 ). In cases of relatively large tumors, biopsy specimens may demonstrate moderately or well-differentiated histology, while deeper tumor portions might harbor poorly differentiated components. Therefore, there exists an urgent clinical need for reliable methods to accurately evaluate tumor differentiation status prior to treatment initiation. CT provides a comprehensive assessment of tumor morphology and has become a widely utilized imaging modality for gastric cancer evaluation ( 4 ). Gastric carcinomas with varying differentiation grades exhibit distinct CT morphological manifestations due to their differing biological aggressiveness. However, challenges such as poor interobserver consistency and suboptimal diagnostic accuracy may arise in clinical practice, attributable to the irregular tumor contours and the presence of non-quantifiable lesions. In contrast to conventional CT imaging characteristics, radiomics represents an advanced methodological framework that enables high-throughput feature extraction, quantitative analysis, and decision support from medical imaging data. This technique facilitates the quantification of subvisual high-dimensional features from large-scale medical images, thereby characterizing tumor heterogeneity and microenvironment more precisely than traditional visual assessment ( 5 , 6 ). The radiomics approach has been extensively investigated for oncological applications, including diagnosis, staging, and prognostic prediction. Particularly in determining gastric cancer differentiation status, radiomics has demonstrated substantial potential to improve histological grading accuracy ( 7 , 8 ). The molecular biological characteristics of tumors, serving as critical determinants of their malignant phenotypes and therapeutic responsiveness, have traditionally been assessed through invasive techniques such as tissue biopsy or genomic sequencing. However, these conventional approaches are inherently limited by their invasiveness, sampling bias, and challenges posed by spatiotemporal heterogeneity ( 9 ). Evidence demonstrated that radiomics could noninvasively decode molecular profiles of tumors. Through the utilization of high-throughput techniques to extract quantitative characteristics from medical imaging, radiomics enables comprehensive characterization of spatial heterogeneity patterns, which show significant correlations with specific molecular alterations, including gene mutations and cellular proliferation markers ( 10 – 12 ). This non-invasive approach for molecular profiling establishes a novel paradigm for implementing precision oncology strategies. Nonetheless, the relationship between radiomic features of gastric cancer and gene expression is still not well defined. This study developed a machine learning model based on CT radiomics features to noninvasively predict the histological differentiation grade in patients with gastric cancer. The model's predictive efficacy was thoroughly assessed. The relationship between radiomics and gene expression was investigated through the integration of radiomic features and genomic datasets. The roles of significant genes were confirmed using bioinformatics analytical techniques alongside experimental approaches. MATERIALS AND METHODS Clinical Data This study retrospectively analyzed a total of 488 gastric cancer patients who were treated at our hospital from December 1, 2021, to June 1, 2025. Inclusion criteria: ( 1 ) Pathologically confirmed gastric adenocarcinoma; ( 2 ) Underwent standard CT examination at diagnosis; ( 3 ) Complete clinical data available. Exclusion criteria:( 1 ) Poor-quality CT images with artifacts (n = 30); ( 2 ) Concurrent other primary malignant tumors (n = 168); ( 3 ) Incomplete clinical data (n = 128). According to the inclusion and exclusion criteria, 162 patients were ultimately enrolled and randomly divided into training (n = 113) and validation cohorts (n = 49) at a 7:3 ratio. Clinical data were retrieved from the medical record system, including age, sex, primary tumor location, tumor diameter, T stage, N stage, M stage, ECOG score, CEA, CA199 and differentiation grade. The study design is illustrated in Fig. 1 . CT Image Scan To ensure image quality, patients were instructed in advance regarding the precautions during the examination: fasting for 6 ~ 8 hours and drinking 800 ~ 1,000 mL of water 20 minutes before the scan. In the supine position, patients were asked to take a deep breath and hold it during the CT examination to avoid respiratory motion artifacts. The imaging was conducted utilizing a multidetector-row spiral computed tomography scanner (DiscoveryHD750, GE Healthcare). Scanning parameters are shown in Table S1 (Supplemental Material). Lesion Segmentation, Radiomics Feature Extraction and Selection The procedure for the extraction and selection of CT and radiomic parameters is illustrated in Fig. 2 . The portal venous-phase CT images were exported from the PACS system. The region of interest (ROI) was manually delineated by a radiologist, who possesses eight years of diagnostic experience, utilizing the ITK-SNAP software (Version 3.8.0). This delineation was performed on the slice exhibiting the most extensive tumor area, specifically along the tumor margin. After a period of one week, 30 cases were randomly chosen and independently re-evaluated by the same radiologist, along with another radiologist who has a decade of experience, to examine both intra-observer and inter-observer reliability. All segmentation procedures were performed blindly without knowledge of the patients' clinical or pathological diagnoses. The delineated ROI images were imported into PyRadiomics software (version 3.0.1) for automated radiomics feature extraction. In total, 1656 radiomic features were derived from the designated region of interest within the CT images. These features were classified into five distinct categories: first-order features (396 features), shape-based features (28 features), features derived from the gray-level co-occurrence matrix (GLCM, 528 features), features from the gray-level run length matrix (GLRLM, 352 features), and features from the gray-level size zone matrix (GLSZM, 352 features). The radiomics features were selected through a four-step method, and the radiomics score (Rad-score) was calculated. Step 1: The stability of radiomics features was evaluated using the intraclass correlation coefficient (ICC), and only features with excellent consistency were retained. Step 2: The Mann-Whitney U test was applied to further screen for features with statistically significant differences ( P 0.75 for exclusion. Step 4: The least absolute shrinkage and selection operator (LASSO) technique was employed to identify the pertinent features among the remaining variables. For each patient, the Rad-score was calculated as a linear aggregation of the selected features, with each feature being weighted according to its respective coefficient. Model Construction and Performance Validation The diagnostic model was developed utilizing the findings from multivariate analysis. To evaluate the predictive efficacy of the model, the area under the receiver operating characteristic curve (ROC) was employed. Calibration curves were utilized to examine the model's calibration, whereas decision curve analysis (DCA) was implemented to ascertain the net benefits across various decision thresholds. Additionally, clinical impact curves (CIC) were constructed to evaluate the clinical relevance of the nomogram. The validation cohort was subsequently utilized for the validation process. Analysis of Differentially Expressed Genes (DEGs) and Gene Enrichment Analysis Transcriptome sequencing was performed on tumor tissues from 7 patients, and RNA-seq data were successfully obtained. The DEGs between the well/moderate differentiation and poor differentiation were analyzed using the "limma" R package in R software. The visualization of differentially expressed genes (DEGs) data was accomplished using the "pheatmap" and "ggplot2" packages, which facilitated the creation of heatmaps and volcano plots. To enrich the biological processes (BPs) associated with the Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways of DEGs, the "clusterProfiler" package was utilized. For conducting Gene Set Enrichment Analysis (GSEA), the "GSEABase" package was employed, and the resultant findings were illustrated through the "enrichplot" package in R. Machine Learning Screening Hub Gene The identification of hub genes was conducted utilizing LASSO logistic regression and Random Forest (RF) machine learning methodologies. The LASSO logistic regression was executed through the "glmnet" package, with the minimal lambda value deemed as the most suitable. For the random forest analysis, the "randomForestSRC" package was employed. Genes that were consistently identified by both algorithms underwent additional investigation. Hub Gene Expression Differential and Functional Enrichment Analysis An analysis of differential gene expression for the hub gene was conducted. Differentially expressed genes (DEGs) were identified by categorizing hub genes into high and low expression groups. The DEGs were selected using the thresholds of P 2. Furthermore, enrichment analysis of the DEGs was carried out utilizing Gene Set Enrichment Analysis (GSEA). Functional and Pathway Validation of the Hub Gene Experiments were conducted using the gastric cancer cell line MKN45. Gene knockdown and overexpression were achieved using shRNA or cDNA. The effects on cellular proliferation were evaluated by CCK-8 assay. Wound healing and Transwell assays were utilized to examine the roles of target genes in cell migration and invasion capabilities. Statistical Analysis Statistical analyses were performed using Graphpad Prism 8.0, SPSS 27 and R 4.4.2 software. Continuous variables with normal distribution were presented as mean ± standard deviation and compared using the independent samples t-test. Categorical data were expressed as frequencies and analyzed by chi-square test or Fisher's exact test, as appropriate. The intra- and inter-observer agreement regarding feature extraction was evaluated using the ICC. An ICC value exceeding 0.75 indicates a strong level of agreement. Logistic regression analysis was employed for both univariate and multivariate assessments. A two-tailed P < 0.05 was deemed statistically significant. RESULTS Baseline Characteristics A total of 162 gastric cancer patients were divided into a training cohort (n = 113) and a validation cohort (n = 49). In the training cohort, there were 40 cases in the well/moderate differentiation group and 73 cases in the poor differentiation group. The validation cohort consisted of 19 cases in the well/moderate differentiation group and 30 cases in the poor differentiation group. No notable disparities were found between the training and validation cohorts regarding patient age, sex, primary tumor location, tumor diameter, T stage, N stage, M stage, ECOG score, CEA, CA199, and Rad-score ( P > 0.05) (Table 1 ). Table 1 Characteristics between training and validation cohorts Variables Training cohort (n = 113) Validation cohort (n = 49) P value Well/Moderate Differentiation Poor Differentiation P value Well/Moderate Differentiation Poor Differentiation P value Age 63.08 ± 8.98 61.51 ± 11.77 0.465 63.95 ± 6.86 61.07 ± 11.77 0.285 0.947 Sex Male Female 32 8 49 24 0.146 17 2 20 10 0.142 0.615 Primary tumor location Gastric Gastric-esophageal junction 28 12 55 18 0.539 16 3 23 7 0.784 0.405 Tumor diameter (cm) 4.25 ± 2.39 4.35 ± 3.06 0.859 3.61 ± 1.74 3.58 ± 2.54 0.960 0.113 T stage T1 ~ 3 T4 28 12 24 49 0.017 14 5 12 18 0.021 0.410 N stage N0 ~ 1 N2 ~ 3 26 14 18 55 < 0.001 11 8 5 25 0.003 0.447 M stage M0 M1 29 11 23 50 < 0.001 17 2 10 20 < 0.001 0.288 ECOG score 0 1 ~ 2 25 15 37 36 0.227 9 10 12 18 0.612 0.160 CEA <5 ≥5 27 13 49 24 0.967 13 6 18 12 0.551 0.622 CA199 <37 ≥37 31 9 58 15 0.808 15 4 24 6 1 0.905 Rad-score 2.35 ± 1.14 3.75 ± 1.11 < 0.001 2.26 ± 1.08 4.01 ± 1.42 < 0.001 0.743 Selection of Radiomics Features and Calculation of Rad-score Figure 2 illustrates the comprehensive procedure involved in the extraction and selection of radiomics features. A total of 1,656 radiomics features were extracted from the ROI on portal venous phase CT images. After screening for intra- and inter-observer ICC > 0.75, 1,108 features were retained. The Mann-Whitney U test identified 115 significant features, and Spearman correlation analysis was applied to eliminate redundancy, resulting in 42 key features. A total of 42 features underwent filtering through LASSO regression analysis, resulting in the identification of four significant features. The chosen features, along with their corresponding regression coefficients, were employed to compute radiomics scores. The radiomics score for each patient was determined using the formula outlined below: Rad-score = (0.0116 × original_firstorder_Median + 0.1696 × original_glrlm_ShortRunEmphasis + 0.0058 × logarithm_gldm_LowGrayLevelEmphasis) + 0.254 × exponential_glszm_GrayLevelNonUniformity The Univariate and Multivariate Logistic Analysis of Parameters The univariate analysis showed that T stage, N stage, M stage and Rad-score were significantly correlated with differentiation ( P < 0.05). Multivariate logistic analysis showed that N stage (OR: 3.533, 95%CI: 1.223–10.208, P = 0.020), M stage (OR:3.321, 95%CI: 1.191–9.257, P = 0.022) and Rads-score (OR: 2.516, 95%CI: 1.561–4.057, P < 0.001) were independent factors influencing differentiation (Table 2 ). Table 2 Univariate and multivariate analysis of characteristics in gastric adenocarcinoma Variables Univariate analysis ( P ) Multivariate logistic regression analysis OR 95%CI P Age 0.462 Sex 0.150 Primary tumor location 0.539 Tumor diameter (cm) 0.857 T stage < 0.001 * 2.092 0.721–6.067 0.174 N stage < 0.001 * 3.533 1.223–10.208 0.020 M stage < 0.001 * 3.321 1.191–9.257 0.022 ECOG score 0.229 CEA 0.967 CA199 0.808 Rad-score < 0.001 * 2.516 1.561–4.057 < 0.001* Construction, Evaluation and Validation of Nomogram Based on a combination of the N stage, M stage and Rad-score, a nomogram was created to predict tumor differentiation in patients with gastric cancer (Fig. 3 A). Each variable was modeled both independently and jointly, with subsequent evaluation of their predictive efficacy. The AUCs for predicting tumor differentiation grade reached 0.702, 0.705, 0.830, 0.787 and 0.872 in the training cohort (Fig. 3 B), and 0.706, 0.781, 0.835, 0.842 and 0.935 in the validation cohort (Fig. 3 C), respectively. The calibration curve of the nomogram indicated a strong concordance between the observed values and the predicted values in both the training and validation cohorts (Fig. 3 D, E). Decision curve analysis (DCA) showed that the model developed had the greatest net benefit in training and validation cohorts (Fig. 3 F, G). CIC showed good clinical applicability of the model in training and validation cohorts (Fig. 3 H, I). Identification and Enrichment Analysis of DEGs A total of 61 DEGs were identified, comprising 49 up-regulated and 12 down-regulated genes (Fig. 4 A). The heatmap displayed the expression profiles of these DEGs (Fig. 4 B). GO enrichment analysis highlighted significant biological processes, including immunoglobulin-mediated immune response, B cell-mediated immunity, lymphocyte-mediated immunity, adaptive immune response and B cell receptor signaling pathway (Fig. 4 C). Additionally, KEGG pathway analysis revealed enrichment in neuroactive ligand-receptor interaction, neuroactive ligand signaling and hormone signaling (Fig. 4 D). Machine Learning Screening Hub Gene Utilizing the LASSO logistic regression algorithm, six hub genes were identified from an initial pool of twenty hub genes (Fig. 4 E, F). The twenty hub genes were subsequently prioritized based on their variable importance scores, determined through the RF algorithm, and those with a variable importance greater than 0.01 were selected as significant hub genes (Fig. 4 G, H). Notably, both machine learning methodologies converged on the common hub gene IGHG1 (Fig. 4 I). IGHG1 Expression Differential and Functional Enrichment Analysis A comparative analysis of IGHG1 expression and Rad-scores between the poor differentiation and well/moderate differentiation groups revealed significantly higher levels of both IGHG1 gene expression ( P < 0.05) and Rad-scores ( P < 0.05) in the poor differentiation group compared to the well/moderate differentiation group (Fig. 5 A, B). The correlation analysis indicated a significant association between the expression levels of IGHG1 and the Rad-score (r = 0.946, P = 0.001) (Fig. 5 C). The GSEA results showed an increase in TGF-β signaling, epithelial-mesenchymal transition (EMT) and KRAS signaling and a decrease in MYC targets V1 in patients in the IGHG1 high expression group compared to the low expression group (Fig D, E). Functional Validation of the Hub Gene IGHG1 knockdown and overexpression were achieved using shRNA. Quantitative polymerase chain reaction (qPCR) and WB results confirmed the successful gene knockdown and overexpression operations (Fig. 6 A, B, F, G). CCK-8 assays demonstrated that knockdown of the IGHG1 gene attenuated the proliferation level of MKN45 cells (Fig. 6 C), while overexpression of the IGHG1 gene enhanced it (Fig. 6 H). Wound healing and Transwell assays proved that knockdown of the IGHG1 gene reduced the migration and invasion abilities of MKN45 cells (Fig. 6 D, E), whereas overexpression of the IGHG1 gene improved these capabilities (Fig. 6 I, J). DISCUSSION The histological differentiation grade of gastric cancer is a crucial indicator for evaluating tumor biological behavior, formulating individualized treatment plans, and predicting prognosis ( 14 ). Traditionally, the determination of differentiation grade relies on postoperative pathological examination, which cannot provide a reference for clinical decision-making before surgery. This study constructed a radiomics model based on CT images, achieving noninvasive prediction of gastric cancer differentiation grade, thereby offering a novel approach for precise preoperative diagnosis and treatment. This study showed that N stage, M stage and Rads-score were independent factors influencing differentiation. The predictive model based on N stage, M stage and Rads-score was developed. The outcomes derived from the ROC indicated that the clinical-radiomics model possesses a strong predictive accuracy. Furthermore, calibration curves demonstrated a favorable alignment between the actual observations and the predicted values. The DCA illustrated a significant net benefit attributed to the model. Additionally, the CIC confirmed the robust clinical applicability of the model. Analyzing the sequencing data identified IGHG1 as a pivotal gene within the radiomics model. The GSEA results showed an increase in TGF-β signaling, EMT and KRAS signaling pathway. Radiomics and its underlying mechanisms were revealed. The TNM staging system, as a cornerstone tool for prognostic assessment in gastric cancer, has been continuously refined in recent years, driven by advancements in molecular biology and imaging ( 15 , 16 ). A significant correlation exists between tumor differentiation degree and TNM stage. Data from this study revealed that the proportion of N2-3 stage patients was significantly higher in the poor differentiation group compared to the well/moderate differentiation group. Research indicated that tumor differentiation was an independent prognostic factor, exhibiting a significant correlation with both the overall TNM stage and its individual components. Well-differentiated tumors were generally associated with lower TNM stages, whereas poorly differentiated tumors tend to correlate with more advanced TNM stages. This association was particularly evident across different T stages, with less differentiated tumors showing a higher propensity for lymph node metastasis ( 17 ). The relationship between differentiation degree and lymph node metastasis risk holds substantial clinical significance, especially in postoperative risk assessment and treatment strategy formulation. Furthermore, other studies had highlighted significant correlations between tumor differentiation and individual components of TNM staging (such as T stage, N stage, and M stage), emphasizing the importance of considering tumor differentiation in the diagnosis and treatment of gastric cancer ( 16 , 18 ). In recent years, CT radiomics has emerged as a cutting-edge interdisciplinary field, demonstrating great potential in quantifying tumor phenotypes through high-throughput extraction and analysis of quantitative imaging features ( 19 , 20 ). Research on radiomics for evaluating gastric cancer differentiation is becoming a crucial breakthrough in improving the precision of preoperative diagnosis and treatment. Multiple retrospective studies confirmed that radiomic features extracted from preoperative contrast-enhanced CT (particularly in the portal venous phase) could be used to construct highly discriminative models for predicting poorly differentiated gastric cancer or distinguishing well/moderately differentiated gastric cancer ( 21 , 22 ). The imaging characteristics of poorly differentiated adenocarcinoma have been extensively analyzed to better understand their manifestations on CT and MRI. Studies found that poorly differentiated adenocarcinomas typically appear as isodense or hypodense on non-contrast CT scans, while exhibiting high signal intensity on T2-weighted imaging. By quantifying these differences through high-throughput feature analysis, radiomics overcomes the subjectivity limitations inherent in traditional visual assessments ( 23 ). In this study, features were screened, including first-order_Median, glrlm_ShortRunEmphasis, gldm_LowGrayLevelEmphasis and glszm_GrayLevelNonUniformity. First-order_Median represents the median gray level intensity within the ROI. Glrlm_ShortRunEmphasis is a measure of the distribution of short run lengths, with a greater value indicative of shorter run lengths and finer textural textures. Gldm_LowGrayLevelEmphasis measures the similarity of gray-level intensity values in the image, where a lower GLN value correlates with a greater similarity in intensity values.Glszm_GrayLevelNonUniformity measures the variability of gray-level intensity values in the image, with a lower value indicating more homogeneity in intensity values. These reflect the heterogeneity within the tumour. In recent years, radiomics has emerged as a pivotal bridge linking macroscopic imaging characteristics with microscopic molecular features, demonstrating tremendous potential in the field of precision diagnosis and treatment for gastric cancer. The findings of this study revealed that IGHG1 expression was significantly correlated with Rad-score (r = 0.946, P = 0.001). The GSEA results showed an increase in TGF-β signaling pathway, EMT and KRAS signaling and a decrease in MYC targets V1 in patients in the IGHG1 high expression group compared to the low expression group. A study revealed that IGHG1 promotes the migration and invasion of gastric cancer cells and is closely associated with EMT ( 24 ). Li et al. demonstrated that the IGHG1/AKT/GSK-3β/β-catenin signaling pathway plays a regulatory role in the malignant biological behavior of gastric cancer ( 25 ). This study revealed the crucial role of IGHG1 in gastric cancer cells, particularly in regulating cell proliferation, migration, and invasion. Our research demonstrates that IGHG1 not only promotes the proliferation of MKN45 cells but also modulates their migratory and invasive capabilities by regulating the activity of the TGF-β signaling pathway. This discovery holds significant importance for understanding the pathogenesis and progression of gastric cancer, especially in elucidating its underlying molecular mechanisms. The TGF-β signaling pathway is known to play a dual role in various cancers, capable of both suppressing tumor formation and promoting tumor progression ( 26 , 27 ). Therefore, by influencing the activation of the TGF-β signaling pathway, IGHG1 may play a pivotal role in tumor growth and metastasis, providing a potential molecular target for the development of novel therapeutic strategies in the future. The limitations of this study still warrant attention. Firstly, the retrospective design may introduce selection bias. Secondly, variations in CT scanner parameters across institutions could affect feature reproducibility, necessitating future multicenter validation to optimize standardization protocols. CONCLUSION The predictive model based on N stage, M stage and Rad-score can effectively predict the differentiation in gastric cancer patients. Radiomics enables noninvasive prediction of tumor differentiation status while elucidating the expression levels of the IGHG1 and the underlying pathway. ETHICAL APPROVAL AND CONSENT TO PARTICIPATE This study was performed in line with the principles of the Declaration of Helsinki. Approval was granted by the Ethics Committee of Anhui Provincial Hospital (Approval No. 2025-RE-262). Informed consent was obtained from all individual participants included in the study. Written informed consent for publication of their clinical details and/or clinical images was obtained from the patient/participant. FUNDING DECLARATION This study was supported by the Anhui Provincial Natural Science Foundation (Grant No. 2208085QH228). DATA AVAILABILITY STATEMENT The raw RNA sequencing data generated in this study have been deposited in the NCBI Sequence Read Archive (SRA) and are publicly available under the BioProject accession number PRJNA1390632. Declarations ETHICAL APPROVAL AND CONSENT TO PARTICIPATE This study was performed in line with the principles of the Declaration of Helsinki. Approval was granted by the Ethics Committee of Anhui Provincial Hospital (Approval No. 2025-RE-262). Informed consent was obtained from all individual participants included in the study. Written informed consent for publication of their clinical details and/or clinical images was obtained from the patient/participant. FUNDING DECLARATION This study was supported by the Anhui Provincial Natural Science Foundation (Grant No. 2208085QH228). DATA AVAILABILITY STATEMENT The raw RNA sequencing data generated in this study have been deposited in the NCBI Sequence Read Archive (SRA) and are publicly available under the BioProject accession number PRJNA1390632. References Bray F, Laversanne M, Sung H, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2024; 74(3):229-63. Feng F, Liu J, Wang F, et al. Prognostic value of differentiation status in gastric cancer. BMC Cancer. 2018; 18(1):865. Lee IS, Park YS, Lee JH, et al. Pathologic discordance of differentiation between endoscopic biopsy and postoperative specimen in mucosal gastric adenocarcinomas. Ann Surg Oncol. 2013; 20(13):4231-7. Kubota K, Suzuki A, Shiozaki H, Wada T, Kyosaka T, Kishida A. Accuracy of Multidetector-Row Computed Tomography in the Preoperative Diagnosis of Lymph Node Metastasis in Patients with Gastric Cancer. Gastrointest Tumors. 2017; 3(3-4):163-70. Lambin P, Leijenaar RTH, Deist TM, et al. Radiomics: the bridge between medical imaging and personalized medicine. Nat Rev Clin Oncol. 2017; 14(12):749-62. Lee S, Choe EK, Kim SY, Kim HS, Park KJ, Kim D. Liver imaging features by convolutional neural network to predict the metachronous liver metastasis in stage I-III colorectal cancer patients based on preoperative abdominal CT scan. BMC Bioinformatics. 2020; 21(Suppl 13):382. Gao X, Ma T, Cui J, et al. A CT-based Radiomics Model for Prediction of Lymph Node Metastasis in Early Stage Gastric Cancer. Acad Radiol. 2021; 28(6):e155-e64. Feng B, Huang L, Li C, et al. A Heterogeneity Radiomic Nomogram for Preoperative Differentiation of Primary Gastric Lymphoma From Borrmann Type IV Gastric Cancer. J Comput Assist Tomogr. 2021; 45(2):191-202. Guo Y, Li H, Chen H, et al. Metagenomic next-generation sequencing to identify pathogens and cancer in lung biopsy tissue. EBioMedicine. 2021; 73:103639. Le NQK, Kha QH, Nguyen VH, Chen YC, Cheng SJ, Chen CY. Machine Learning-Based Radiomics Signatures for EGFR and KRAS Mutations Prediction in Non-Small-Cell Lung Cancer. Int J Mol Sci. 2021; 22(17). Tian R, Li Y, Jia C, et al. Radiomics Model for Predicting TP53 Status Using CT and Machine Learning Approach in Laryngeal Squamous Cell Carcinoma. Front Oncol. 2022; 12:823428. Wang M, Mei T, Gong Y. The quality and clinical translation of radiomics studies based on MRI for predicting Ki-67 levels in patients with breast cancer. Br J Radiol. 2023; 96(1150):20230172. Halder SK, Beauchamp RD, Datta PK. A specific inhibitor of TGF-beta receptor kinase, SB-431542, as a potent antitumor agent for human cancers. Neoplasia. 2005; 7(5):509-21. Japanese Gastric Cancer A. Japanese Gastric Cancer Treatment Guidelines 2021 (6th edition). Gastric Cancer. 2023; 26(1):1-25. Lopez Sala P, Leturia Etxeberria M, Inchausti Iguiniz E, Astiazaran Rodriguez A, Aguirre Oteiza MI, Zubizarreta Etxaniz M. Gastric adenocarcinoma: A review of the TNM classification system and ways of spreading. Radiologia (Engl Ed). 2023; 65(1):66-80. Liu S, Shi H, Ji C, et al. Preoperative CT texture analysis of gastric cancer: correlations with postoperative TNM staging. Clin Radiol. 2018; 73(8):756 e1- e9. Derwinger K, Kodeda K, Bexe-Lindskog E, Taflin H. Tumour differentiation grade is associated with TNM staging and the risk of node metastasis in colorectal cancer. Acta Oncol. 2010; 49(1):57-62. Liu S, Wang H, Guan W, et al. Preoperative apparent diffusion coefficient value of gastric cancer by diffusion-weighted imaging: Correlations with postoperative TNM staging. J Magn Reson Imaging. 2015; 42(3):837-43. Lambin P, Rios-Velazquez E, Leijenaar R, et al. Radiomics: extracting more information from medical images using advanced feature analysis. Eur J Cancer. 2012; 48(4):441-6. Gillies RJ, Kinahan PE, Hricak H. Radiomics: Images Are More than Pictures, They Are Data. Radiology. 2016; 278(2):563-77. Huang J, Yao H, Li Y, et al. Development and validation of a CT-based radiomics nomogram for preoperative prediction of tumor histologic grade in gastric adenocarcinoma. Chin J Cancer Res. 2021; 33(1):69-78. Li Q, Qi L, Feng QX, et al. Machine Learning-Based Computational Models Derived From Large-Scale Radiographic-Radiomic Images Can Help Predict Adverse Histopathological Status of Gastric Cancer. Clin Transl Gastroenterol. 2019; 10(10):e00079. Baba A, Matsushima S, Kessoku H, et al. Radiological features of thyroid-like low-grade nasopharyngeal papillary adenocarcinoma: case series and systematic review. Neuroradiology. 2024; 66(2):249-59. Li Y, Wang P, Ye D, et al. IGHG1 induces EMT in gastric cancer cells by regulating TGF-beta/SMAD3 signaling pathway. J Cancer. 2021; 12(12):3458-67. Li X, Chen W, Yang C, et al. IGHG1 upregulation promoted gastric cancer malignancy via AKT/GSK-3beta/beta-Catenin pathway. Cancer Cell Int. 2021; 21(1):397. Fabregat I, Fernando J, Mainez J, Sancho P. TGF-beta signaling in cancer treatment. Curr Pharm Des. 2014; 20(17):2934-47. Zi Z. Molecular Engineering of the TGF-beta Signaling Pathway. J Mol Biol. 2019; 431(15):2644-54. Additional Declarations No competing interests reported. Supplementary Files SupplementaryMaterial.docx Cite Share Download PDF Status: Under Revision Version 1 posted Editorial decision: Revision requested 10 Feb, 2026 Reviews received at journal 07 Feb, 2026 Reviewers agreed at journal 26 Jan, 2026 Reviews received at journal 17 Jan, 2026 Reviewers agreed at journal 24 Dec, 2025 Reviewers agreed at journal 22 Dec, 2025 Reviewers invited by journal 22 Dec, 2025 Editor assigned by journal 22 Dec, 2025 Editor invited by journal 19 Dec, 2025 Submission checks completed at journal 18 Dec, 2025 First submitted to journal 18 Dec, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7782471","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":565138898,"identity":"28aa75af-d75c-450b-bdf9-c528196149b2","order_by":0,"name":"Rixin Su","email":"","orcid":"","institution":"Department of Chinese Integrative Medicine Oncology, The First Affiliated Hospital of Anhui Medical University","correspondingAuthor":false,"prefix":"","firstName":"Rixin","middleName":"","lastName":"Su","suffix":""},{"id":565138900,"identity":"12c61d69-afea-43e3-9df4-ccbe27a66880","order_by":1,"name":"Yu Zhang","email":"","orcid":"","institution":"Department of Chinese Integrative Medicine Oncology, The First Affiliated Hospital of Anhui Medical University","correspondingAuthor":false,"prefix":"","firstName":"Yu","middleName":"","lastName":"Zhang","suffix":""},{"id":565138901,"identity":"4d20af4e-94d4-467f-910d-a579d9077a18","order_by":2,"name":"Jie Cao","email":"","orcid":"","institution":"The First Affiliated Hospital of USTC, Division of Life Sciences and Medicine, University of Science and Technology of China","correspondingAuthor":false,"prefix":"","firstName":"Jie","middleName":"","lastName":"Cao","suffix":""},{"id":565138902,"identity":"be293e37-06f9-4b63-95d5-7ede5c443291","order_by":3,"name":"Fangfang Chen","email":"","orcid":"","institution":"The First Affiliated Hospital of USTC, Division of Life Sciences and Medicine, University of Science and Technology of China","correspondingAuthor":false,"prefix":"","firstName":"Fangfang","middleName":"","lastName":"Chen","suffix":""},{"id":565138903,"identity":"1cbe48c0-9b5d-44a6-96c4-c5ba8b6d0972","order_by":4,"name":"Xuemeng Li","email":"","orcid":"","institution":"The First Affiliated Hospital of USTC, Division of Life Sciences and Medicine, University of Science and Technology of China","correspondingAuthor":false,"prefix":"","firstName":"Xuemeng","middleName":"","lastName":"Li","suffix":""},{"id":565138905,"identity":"c72fa167-a147-48db-963c-a057c8d0619b","order_by":5,"name":"Ping Li","email":"","orcid":"","institution":"Department of Chinese Integrative Medicine Oncology, The First Affiliated Hospital of Anhui Medical University","correspondingAuthor":false,"prefix":"","firstName":"Ping","middleName":"","lastName":"Li","suffix":""},{"id":565138911,"identity":"5863ebcf-3dbc-467c-a9c0-00b64d57c753","order_by":6,"name":"Geng Bian","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA7klEQVRIie3PMYvCMBTA8VceJEu4rIbehwgIrUKpX6VHoJODkzg4VApOgqsfw8lZ7nG6HNx6g8stnRVEOojYwrk2HQXzH8ILvB8kAC7XE6bxf2CI+/Koo7g9kZwlajVKjZ08BrUU2hfHTy+zkZBj8XeeHOI1ge5GeovA6WvdRPo5C7vv34WpSGKG+vAGIk1/Gx9GEPhqTqYatjTUBUJHBBbCL7661cSb5b3qzOxEBOqUUaxyRIQ2pJ+LsQ87SiQy5i10apjtL6Hcb1Q5pQGTPxcor1EsOe0aSR0KgI/scWO29TqvBBi0WXS5XK4X7Q7bvkh1v56tUgAAAABJRU5ErkJggg==","orcid":"","institution":"The First Affiliated Hospital of USTC, Division of Life Sciences and Medicine, University of Science and Technology of China","correspondingAuthor":true,"prefix":"","firstName":"Geng","middleName":"","lastName":"Bian","suffix":""}],"badges":[],"createdAt":"2025-10-05 03:08:13","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7782471/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7782471/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":99317171,"identity":"1956d7ec-da5d-4836-847f-41c6c5d3b9e3","added_by":"auto","created_at":"2025-12-31 16:29:43","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1875444,"visible":true,"origin":"","legend":"","description":"","filename":"Manuscript.docx","url":"https://assets-eu.researchsquare.com/files/rs-7782471/v1/9670af6cc0098d77c183b4c4.docx"},{"id":99316425,"identity":"a8627f24-720c-4a9e-b06f-e73ad0ecd7d1","added_by":"auto","created_at":"2025-12-31 16:28:27","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":8691,"visible":true,"origin":"","legend":"","description":"","filename":"5a6090ccd1634f62b4286edbf71153a3.json","url":"https://assets-eu.researchsquare.com/files/rs-7782471/v1/1a1f99e6d98ed89d9031b38d.json"},{"id":99192451,"identity":"68e8b897-86f2-4dc3-9fa9-9ad6f3a1cb53","added_by":"auto","created_at":"2025-12-30 01:04:15","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":12650,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterial.docx","url":"https://assets-eu.researchsquare.com/files/rs-7782471/v1/a2986b12b104c74740b8f6f5.docx"},{"id":99316813,"identity":"7b9db72e-040a-4953-a5f1-354e67a459f8","added_by":"auto","created_at":"2025-12-31 16:29:14","extension":"xml","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":97387,"visible":true,"origin":"","legend":"","description":"","filename":"5a6090ccd1634f62b4286edbf71153a31enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7782471/v1/7cd2927945cf7e5aea162331.xml"},{"id":99192466,"identity":"20066918-95cd-456d-b3e5-d5f8d992172c","added_by":"auto","created_at":"2025-12-30 01:04:16","extension":"jpeg","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":5795060,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7782471/v1/6e40264478e7d4ca12a95584.jpeg"},{"id":99317673,"identity":"5941cebf-27f2-4f3c-ad00-c9a351a59206","added_by":"auto","created_at":"2025-12-31 16:30:34","extension":"jpeg","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":4230470,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7782471/v1/2b35be798c4e7bed6005755e.jpeg"},{"id":99192465,"identity":"e8d89795-4f68-40db-b31d-e97c1f7a460a","added_by":"auto","created_at":"2025-12-30 01:04:16","extension":"jpeg","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":5213512,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7782471/v1/35620aa425549286e56bf4fe.jpeg"},{"id":99192457,"identity":"50d9dd32-27d3-468d-80ee-7c95bf2e2d85","added_by":"auto","created_at":"2025-12-30 01:04:15","extension":"jpeg","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":5007436,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7782471/v1/e607b1fe28cda043ed6e23d7.jpeg"},{"id":99192458,"identity":"49d53b2a-247b-4d4e-a1be-dfe3822cb3d2","added_by":"auto","created_at":"2025-12-30 01:04:15","extension":"jpeg","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":4535410,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7782471/v1/baf227ac22392ebfcb991aab.jpeg"},{"id":99192456,"identity":"d333341c-82ba-444a-a2df-d74a8cb40fdd","added_by":"auto","created_at":"2025-12-30 01:04:15","extension":"jpeg","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":5790598,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage6.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7782471/v1/30a4b3992ed9d06049842a9f.jpeg"},{"id":99192459,"identity":"9db68aea-265a-4630-b83c-37b6e102e7d9","added_by":"auto","created_at":"2025-12-30 01:04:15","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":47172,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7782471/v1/5875276aeee37e7b5c1b5499.png"},{"id":99192460,"identity":"616c8dc2-4a73-4952-903b-0607e07d6e59","added_by":"auto","created_at":"2025-12-30 01:04:15","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":64943,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7782471/v1/177f74bcd5d48f849f842e9a.png"},{"id":99316234,"identity":"041668a2-5736-4f3e-bd05-90225e788afe","added_by":"auto","created_at":"2025-12-31 16:27:55","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":64444,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7782471/v1/208b59f578a6d87f88b1c31b.png"},{"id":99192463,"identity":"d626a49e-4f36-45f8-b23c-f4827b9d1bd2","added_by":"auto","created_at":"2025-12-30 01:04:15","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":48444,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7782471/v1/79e74e9f9f484bd6e04009e5.png"},{"id":99318602,"identity":"519e44cb-c37f-4433-8d64-d3116532fe6c","added_by":"auto","created_at":"2025-12-31 16:33:43","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":37188,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7782471/v1/f579b0a6d1f156a990c497a1.png"},{"id":99317528,"identity":"95cdf47a-e97a-4ead-99ed-942e485f48c3","added_by":"auto","created_at":"2025-12-31 16:30:19","extension":"png","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":134668,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7782471/v1/ea638479c13c899c4c299310.png"},{"id":99316445,"identity":"a0508b08-7c94-42d7-9641-b58e2ddef490","added_by":"auto","created_at":"2025-12-31 16:28:28","extension":"xml","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":94378,"visible":true,"origin":"","legend":"","description":"","filename":"5a6090ccd1634f62b4286edbf71153a31structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7782471/v1/54814dc4c82dd802c18a1c05.xml"},{"id":99317503,"identity":"f6c1d2b7-a7bd-4a98-adb2-ac8ad9785c18","added_by":"auto","created_at":"2025-12-31 16:30:18","extension":"html","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":106600,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7782471/v1/0953165f5b6f430f7ffe655d.html"},{"id":99192447,"identity":"12a6027c-3f78-45a3-a934-c7c4004b5741","added_by":"auto","created_at":"2025-12-30 01:04:15","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":1014575,"visible":true,"origin":"","legend":"\u003cp\u003eFlow chart of this study.\u003c/p\u003e","description":"","filename":"image1.png","url":"https://assets-eu.researchsquare.com/files/rs-7782471/v1/a54831f7af9d677ce7b92016.png"},{"id":99317521,"identity":"ee98190b-305e-464e-aa49-defb8dd14862","added_by":"auto","created_at":"2025-12-31 16:30:19","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":1326306,"visible":true,"origin":"","legend":"\u003cp\u003eFlowchart for segmentation, delineation, and model construction.\u003c/p\u003e","description":"","filename":"image2.png","url":"https://assets-eu.researchsquare.com/files/rs-7782471/v1/fec7f117b0200a5e9ddf1bcb.png"},{"id":99318822,"identity":"0d949baa-f35d-46a7-bfa4-d292a95e3849","added_by":"auto","created_at":"2025-12-31 16:35:01","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":1107330,"visible":true,"origin":"","legend":"\u003cp\u003eEvaluation and validation of the nomogram. (A) The nomogram combined radiomics and clinical parameters. (B-C) The ROC of the nomogram in the training and validation cohorts. (D-E) Calibration curves of the nomogram in training and validation cohorts. (F-G) DCA analysis of the nomogram in training and validation cohorts. (H-I) CIC of the nomogram in training and validation cohorts.\u003c/p\u003e","description":"","filename":"image3.png","url":"https://assets-eu.researchsquare.com/files/rs-7782471/v1/d7fcd91b6d67f1b032c3f5d2.png"},{"id":99192453,"identity":"05437009-be27-4a65-a422-48399a5c3e1f","added_by":"auto","created_at":"2025-12-30 01:04:15","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":961200,"visible":true,"origin":"","legend":"\u003cp\u003eDEGs analysis and hub gene screening. (A) The volcano plot illustrates the differentially expressed genes (DEGs) across various differentiation groups. (B) The heatmap visualizes the expression patterns of these DEGs. (C) The Gene Ontology (GO) analysis highlights the biological processes associated with the DEGs. (D) The Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis reveals the enriched terms related to these DEGs. (E) The optimal parameter (lambda) selection within the LASSO model is depicted. (F) The distribution of coefficients for the six hub genes most strongly correlated, as determined by the optimal lambda, is shown. (G-H) The Random Forest (RF) algorithm identifies hub genes with variable importance greater than 0.01. (I) A Venn diagram illustrates the overlap of genes identified by both the LASSO and RF algorithms.\u003c/p\u003e","description":"","filename":"image4.png","url":"https://assets-eu.researchsquare.com/files/rs-7782471/v1/c2908d3e54953edeef652ef3.png"},{"id":99317457,"identity":"77aff118-3852-4e9a-807a-caa0f51b9e31","added_by":"auto","created_at":"2025-12-31 16:30:14","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":683125,"visible":true,"origin":"","legend":"\u003cp\u003eIGHG1 expression differential and functional enrichment analysis. (A) IGHG1 expression between the different differentiation groups. (B) Rad-score value between the different differentiation groups. (C) Correlation analysis between Rad-score and IGHG1 expression. (D) Results from GSEA highlighting the pathways that are significantly upregulated. (E) Results from GSEA illustrating the pathways that are notably downregulated.\u003c/p\u003e","description":"","filename":"image5.png","url":"https://assets-eu.researchsquare.com/files/rs-7782471/v1/bd712cd014c130974f93051a.png"},{"id":99192468,"identity":"8115c48e-e2c9-45d4-b5c0-009cf271bde6","added_by":"auto","created_at":"2025-12-30 01:04:16","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":2871668,"visible":true,"origin":"","legend":"\u003cp\u003eFunctional validation of the hub gene.\u003cstrong\u003e \u003c/strong\u003e(A-B, F-G) The efficiency of IGHG1 knockdown or overexpression in MKN45 cells by shRNAwas assessed by qPCR and WB. (C, H) The CCK-8 assay was conducted to assess the proliferation of MKN45 cells. (D, I) The migratory capacity of the transfected cells was evaluated using a wound healing assay. (E, J) The invasion potential of the transfected cells was measured through transwell assays. *, \u003cem\u003eP\u003c/em\u003e\u0026lt; 0.05, **, \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.01, and ***, \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.001.\u003c/p\u003e","description":"","filename":"image6.png","url":"https://assets-eu.researchsquare.com/files/rs-7782471/v1/c6e81810e382814393259f9f.png"},{"id":99788386,"identity":"00dfd202-7f62-4bdf-a246-d47a13ea745f","added_by":"auto","created_at":"2026-01-08 12:46:33","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":9264134,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7782471/v1/b4377d0a-f870-4653-b647-bc15af6c6187.pdf"},{"id":99192444,"identity":"7732ff5d-cb5b-47c8-a3dd-8cdd3006ed64","added_by":"auto","created_at":"2025-12-30 01:04:15","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":12650,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterial.docx","url":"https://assets-eu.researchsquare.com/files/rs-7782471/v1/8d6545fbbe820ac1faecc8d6.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"A CT-based Machine Learning Radiomics Model for the Prediction of Gastric Cancer Differentiation and Mechanism Exploration","fulltext":[{"header":"INTRODUCTION","content":"\u003cp\u003eGastric cancer ranks as the fifth most prevalent malignancy worldwide, with approximately 970,000 new cases diagnosed annually, accounting for 4.8% of all cancer cases. It represents the fourth leading cause of cancer-related mortality (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e). The differentiation status of gastric cancer significantly influences therapeutic decision-making and prognostic outcomes, with poorly differentiated subtypes typically associated with an unfavorable prognosis (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e). Early identification of tumor differentiation grade is crucial for implementing timely clinical interventions and improving patient outcomes. Currently, preoperative assessment of gastric cancer differentiation primarily relies on endoscopic biopsy. However, this technique is limited by its superficial sampling of mucosal layers (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e). In cases of relatively large tumors, biopsy specimens may demonstrate moderately or well-differentiated histology, while deeper tumor portions might harbor poorly differentiated components. Therefore, there exists an urgent clinical need for reliable methods to accurately evaluate tumor differentiation status prior to treatment initiation.\u003c/p\u003e \u003cp\u003eCT provides a comprehensive assessment of tumor morphology and has become a widely utilized imaging modality for gastric cancer evaluation (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e). Gastric carcinomas with varying differentiation grades exhibit distinct CT morphological manifestations due to their differing biological aggressiveness. However, challenges such as poor interobserver consistency and suboptimal diagnostic accuracy may arise in clinical practice, attributable to the irregular tumor contours and the presence of non-quantifiable lesions. In contrast to conventional CT imaging characteristics, radiomics represents an advanced methodological framework that enables high-throughput feature extraction, quantitative analysis, and decision support from medical imaging data. This technique facilitates the quantification of subvisual high-dimensional features from large-scale medical images, thereby characterizing tumor heterogeneity and microenvironment more precisely than traditional visual assessment (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e). The radiomics approach has been extensively investigated for oncological applications, including diagnosis, staging, and prognostic prediction. Particularly in determining gastric cancer differentiation status, radiomics has demonstrated substantial potential to improve histological grading accuracy (\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe molecular biological characteristics of tumors, serving as critical determinants of their malignant phenotypes and therapeutic responsiveness, have traditionally been assessed through invasive techniques such as tissue biopsy or genomic sequencing. However, these conventional approaches are inherently limited by their invasiveness, sampling bias, and challenges posed by spatiotemporal heterogeneity (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e). Evidence demonstrated that radiomics could noninvasively decode molecular profiles of tumors. Through the utilization of high-throughput techniques to extract quantitative characteristics from medical imaging, radiomics enables comprehensive characterization of spatial heterogeneity patterns, which show significant correlations with specific molecular alterations, including gene mutations and cellular proliferation markers (\u003cspan additionalcitationids=\"CR11\" citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e). This non-invasive approach for molecular profiling establishes a novel paradigm for implementing precision oncology strategies. Nonetheless, the relationship between radiomic features of gastric cancer and gene expression is still not well defined.\u003c/p\u003e \u003cp\u003eThis study developed a machine learning model based on CT radiomics features to noninvasively predict the histological differentiation grade in patients with gastric cancer. The model's predictive efficacy was thoroughly assessed. The relationship between radiomics and gene expression was investigated through the integration of radiomic features and genomic datasets. The roles of significant genes were confirmed using bioinformatics analytical techniques alongside experimental approaches.\u003c/p\u003e"},{"header":"MATERIALS AND METHODS","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eClinical Data\u003c/h2\u003e \u003cp\u003eThis study retrospectively analyzed a total of 488 gastric cancer patients who were treated at our hospital from December 1, 2021, to June 1, 2025. Inclusion criteria: (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) Pathologically confirmed gastric adenocarcinoma; (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) Underwent standard CT examination at diagnosis; (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e) Complete clinical data available. Exclusion criteria:(\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) Poor-quality CT images with artifacts (n\u0026thinsp;=\u0026thinsp;30); (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) Concurrent other primary malignant tumors (n\u0026thinsp;=\u0026thinsp;168); (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e) Incomplete clinical data (n\u0026thinsp;=\u0026thinsp;128). According to the inclusion and exclusion criteria, 162 patients were ultimately enrolled and randomly divided into training (n\u0026thinsp;=\u0026thinsp;113) and validation cohorts (n\u0026thinsp;=\u0026thinsp;49) at a 7:3 ratio. Clinical data were retrieved from the medical record system, including age, sex, primary tumor location, tumor diameter, T stage, N stage, M stage, ECOG score, CEA, CA199 and differentiation grade. The study design is illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eCT Image Scan\u003c/h3\u003e\n\u003cp\u003eTo ensure image quality, patients were instructed in advance regarding the precautions during the examination: fasting for 6\u0026thinsp;~\u0026thinsp;8 hours and drinking 800\u0026thinsp;~\u0026thinsp;1,000 mL of water 20 minutes before the scan. In the supine position, patients were asked to take a deep breath and hold it during the CT examination to avoid respiratory motion artifacts. The imaging was conducted utilizing a multidetector-row spiral computed tomography scanner (DiscoveryHD750, GE Healthcare). Scanning parameters are shown in Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e(Supplemental Material).\u003c/p\u003e\n\u003ch3\u003eLesion Segmentation, Radiomics Feature Extraction and Selection\u003c/h3\u003e\n\u003cp\u003eThe procedure for the extraction and selection of CT and radiomic parameters is illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. The portal venous-phase CT images were exported from the PACS system. The region of interest (ROI) was manually delineated by a radiologist, who possesses eight years of diagnostic experience, utilizing the ITK-SNAP software (Version 3.8.0). This delineation was performed on the slice exhibiting the most extensive tumor area, specifically along the tumor margin. After a period of one week, 30 cases were randomly chosen and independently re-evaluated by the same radiologist, along with another radiologist who has a decade of experience, to examine both intra-observer and inter-observer reliability. All segmentation procedures were performed blindly without knowledge of the patients' clinical or pathological diagnoses. The delineated ROI images were imported into PyRadiomics software (version 3.0.1) for automated radiomics feature extraction.\u003c/p\u003e \u003cp\u003eIn total, 1656 radiomic features were derived from the designated region of interest within the CT images. These features were classified into five distinct categories: first-order features (396 features), shape-based features (28 features), features derived from the gray-level co-occurrence matrix (GLCM, 528 features), features from the gray-level run length matrix (GLRLM, 352 features), and features from the gray-level size zone matrix (GLSZM, 352 features).\u003c/p\u003e \u003cp\u003eThe radiomics features were selected through a four-step method, and the radiomics score (Rad-score) was calculated. Step 1: The stability of radiomics features was evaluated using the intraclass correlation coefficient (ICC), and only features with excellent consistency were retained. Step 2: The Mann-Whitney U test was applied to further screen for features with statistically significant differences (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.01). Step 3: Redundant features were eliminated using Spearman correlation analysis, with a threshold correlation coefficient of \u0026gt;\u0026thinsp;0.75 for exclusion. Step 4: The least absolute shrinkage and selection operator (LASSO) technique was employed to identify the pertinent features among the remaining variables. For each patient, the Rad-score was calculated as a linear aggregation of the selected features, with each feature being weighted according to its respective coefficient.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e\n\u003ch3\u003eModel Construction and Performance Validation\u003c/h3\u003e\n\u003cp\u003eThe diagnostic model was developed utilizing the findings from multivariate analysis. To evaluate the predictive efficacy of the model, the area under the receiver operating characteristic curve (ROC) was employed. Calibration curves were utilized to examine the model's calibration, whereas decision curve analysis (DCA) was implemented to ascertain the net benefits across various decision thresholds. Additionally, clinical impact curves (CIC) were constructed to evaluate the clinical relevance of the nomogram. The validation cohort was subsequently utilized for the validation process.\u003c/p\u003e\n\u003ch3\u003eAnalysis of Differentially Expressed Genes (DEGs) and Gene Enrichment Analysis\u003c/h3\u003e\n\u003cp\u003eTranscriptome sequencing was performed on tumor tissues from 7 patients, and RNA-seq data were successfully obtained. The DEGs between the well/moderate differentiation and poor differentiation were analyzed using the \"limma\" R package in R software. The visualization of differentially expressed genes (DEGs) data was accomplished using the \"pheatmap\" and \"ggplot2\" packages, which facilitated the creation of heatmaps and volcano plots. To enrich the biological processes (BPs) associated with the Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways of DEGs, the \"clusterProfiler\" package was utilized. For conducting Gene Set Enrichment Analysis (GSEA), the \"GSEABase\" package was employed, and the resultant findings were illustrated through the \"enrichplot\" package in R.\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eMachine Learning Screening Hub Gene\u003c/h2\u003e \u003cp\u003eThe identification of hub genes was conducted utilizing LASSO logistic regression and Random Forest (RF) machine learning methodologies. The LASSO logistic regression was executed through the \"glmnet\" package, with the minimal lambda value deemed as the most suitable. For the random forest analysis, the \"randomForestSRC\" package was employed. Genes that were consistently identified by both algorithms underwent additional investigation.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eHub Gene Expression Differential and Functional Enrichment Analysis\u003c/h3\u003e\n\u003cp\u003eAn analysis of differential gene expression for the hub gene was conducted. Differentially expressed genes (DEGs) were identified by categorizing hub genes into high and low expression groups. The DEGs were selected using the thresholds of \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 and |log2 (Fold-change)| \u0026gt; 2. Furthermore, enrichment analysis of the DEGs was carried out utilizing Gene Set Enrichment Analysis (GSEA).\u003c/p\u003e\n\u003ch3\u003eFunctional and Pathway Validation of the Hub Gene\u003c/h3\u003e\n\u003cp\u003eExperiments were conducted using the gastric cancer cell line MKN45. Gene knockdown and overexpression were achieved using shRNA or cDNA. The effects on cellular proliferation were evaluated by CCK-8 assay. Wound healing and Transwell assays were utilized to examine the roles of target genes in cell migration and invasion capabilities.\u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eStatistical Analysis\u003c/h2\u003e \u003cp\u003eStatistical analyses were performed using Graphpad Prism 8.0, SPSS 27 and R 4.4.2 software. Continuous variables with normal distribution were presented as mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation and compared using the independent samples t-test. Categorical data were expressed as frequencies and analyzed by chi-square test or Fisher's exact test, as appropriate. The intra- and inter-observer agreement regarding feature extraction was evaluated using the ICC. An ICC value exceeding 0.75 indicates a strong level of agreement. Logistic regression analysis was employed for both univariate and multivariate assessments. A two-tailed \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was deemed statistically significant.\u003c/p\u003e \u003c/div\u003e"},{"header":"RESULTS","content":"\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eBaseline Characteristics\u003c/h2\u003e \u003cp\u003eA total of 162 gastric cancer patients were divided into a training cohort (n\u0026thinsp;=\u0026thinsp;113) and a validation cohort (n\u0026thinsp;=\u0026thinsp;49). In the training cohort, there were 40 cases in the well/moderate differentiation group and 73 cases in the poor differentiation group. The validation cohort consisted of 19 cases in the well/moderate differentiation group and 30 cases in the poor differentiation group. No notable disparities were found between the training and validation cohorts regarding patient age, sex, primary tumor location, tumor diameter, T stage, N stage, M stage, ECOG score, CEA, CA199, and Rad-score (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026gt;\u0026thinsp;0.05) (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eCharacteristics between training and validation cohorts\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"9\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026plusmn;\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026plusmn;\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026plusmn;\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026plusmn;\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eVariables\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003eTraining cohort (n\u0026thinsp;=\u0026thinsp;113)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colspan=\"3\" nameend=\"c8\" namest=\"c6\"\u003e \u003cp\u003eValidation cohort (n\u0026thinsp;=\u0026thinsp;49)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c9\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003e\u003cem\u003eP\u003c/em\u003e value\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eWell/Moderate\u003c/p\u003e \u003cp\u003eDifferentiation\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePoor\u003c/p\u003e \u003cp\u003eDifferentiation\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cem\u003eP\u003c/em\u003e value\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eWell/Moderate\u003c/p\u003e \u003cp\u003eDifferentiation\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003ePoor\u003c/p\u003e \u003cp\u003eDifferentiation\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003e\u003cem\u003eP\u003c/em\u003e value\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e \u003cp\u003e63.08\u0026thinsp;\u0026plusmn;\u0026thinsp;8.98\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e61.51\u0026thinsp;\u0026plusmn;\u0026thinsp;11.77\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.465\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c6\"\u003e \u003cp\u003e63.95\u0026thinsp;\u0026plusmn;\u0026thinsp;6.86\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c7\"\u003e \u003cp\u003e61.07\u0026thinsp;\u0026plusmn;\u0026thinsp;11.77\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.285\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.947\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSex\u003c/p\u003e \u003cp\u003eMale\u003c/p\u003e \u003cp\u003eFemale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e32\u003c/p\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e49\u003c/p\u003e \u003cp\u003e24\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.146\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e17\u003c/p\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e20\u003c/p\u003e \u003cp\u003e10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.142\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.615\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePrimary tumor location\u003c/p\u003e \u003cp\u003eGastric\u003c/p\u003e \u003cp\u003eGastric-esophageal junction\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e28\u003c/p\u003e \u003cp\u003e12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e55\u003c/p\u003e \u003cp\u003e18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.539\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e16\u003c/p\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e23\u003c/p\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.784\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.405\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTumor diameter (cm)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e \u003cp\u003e4.25\u0026thinsp;\u0026plusmn;\u0026thinsp;2.39\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e4.35\u0026thinsp;\u0026plusmn;\u0026thinsp;3.06\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.859\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c6\"\u003e \u003cp\u003e3.61\u0026thinsp;\u0026plusmn;\u0026thinsp;1.74\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c7\"\u003e \u003cp\u003e3.58\u0026thinsp;\u0026plusmn;\u0026thinsp;2.54\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.960\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.113\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT stage\u003c/p\u003e \u003cp\u003eT1\u0026thinsp;~\u0026thinsp;3\u003c/p\u003e \u003cp\u003eT4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e28\u003c/p\u003e \u003cp\u003e12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e24\u003c/p\u003e \u003cp\u003e49\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.017\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e14\u003c/p\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e12\u003c/p\u003e \u003cp\u003e18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.021\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.410\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eN stage\u003c/p\u003e \u003cp\u003eN0\u0026thinsp;~\u0026thinsp;1\u003c/p\u003e \u003cp\u003eN2\u0026thinsp;~\u0026thinsp;3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e26\u003c/p\u003e \u003cp\u003e14\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e18\u003c/p\u003e \u003cp\u003e55\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e11\u003c/p\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e5\u003c/p\u003e \u003cp\u003e25\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.003\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.447\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eM stage\u003c/p\u003e \u003cp\u003eM0\u003c/p\u003e \u003cp\u003eM1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e29\u003c/p\u003e \u003cp\u003e11\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e23\u003c/p\u003e \u003cp\u003e50\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e17\u003c/p\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e10\u003c/p\u003e \u003cp\u003e20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.288\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eECOG score\u003c/p\u003e \u003cp\u003e0\u003c/p\u003e \u003cp\u003e1\u0026thinsp;~\u0026thinsp;2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e25\u003c/p\u003e \u003cp\u003e15\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e37\u003c/p\u003e \u003cp\u003e36\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.227\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e9\u003c/p\u003e \u003cp\u003e10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e12\u003c/p\u003e \u003cp\u003e18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.612\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.160\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCEA\u003c/p\u003e \u003cp\u003e\u0026lt;5\u003c/p\u003e \u003cp\u003e\u0026ge;5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e27\u003c/p\u003e \u003cp\u003e13\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e49\u003c/p\u003e \u003cp\u003e24\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.967\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e13\u003c/p\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e18\u003c/p\u003e \u003cp\u003e12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.551\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.622\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCA199\u003c/p\u003e \u003cp\u003e\u0026lt;37\u003c/p\u003e \u003cp\u003e\u0026ge;37\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e31\u003c/p\u003e \u003cp\u003e9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e58\u003c/p\u003e \u003cp\u003e15\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.808\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e15\u003c/p\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e24\u003c/p\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.905\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRad-score\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e \u003cp\u003e2.35\u0026thinsp;\u0026plusmn;\u0026thinsp;1.14\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e \u003cp\u003e3.75\u0026thinsp;\u0026plusmn;\u0026thinsp;1.11\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c6\"\u003e \u003cp\u003e2.26\u0026thinsp;\u0026plusmn;\u0026thinsp;1.08\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c7\"\u003e \u003cp\u003e4.01\u0026thinsp;\u0026plusmn;\u0026thinsp;1.42\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.743\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003eSelection of Radiomics Features and Calculation of Rad-score\u003c/h2\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e illustrates the comprehensive procedure involved in the extraction and selection of radiomics features. A total of 1,656 radiomics features were extracted from the ROI on portal venous phase CT images. After screening for intra- and inter-observer ICC\u0026thinsp;\u0026gt;\u0026thinsp;0.75, 1,108 features were retained. The Mann-Whitney U test identified 115 significant features, and Spearman correlation analysis was applied to eliminate redundancy, resulting in 42 key features. A total of 42 features underwent filtering through LASSO regression analysis, resulting in the identification of four significant features. The chosen features, along with their corresponding regression coefficients, were employed to compute radiomics scores. The radiomics score for each patient was determined using the formula outlined below:\u003c/p\u003e \u003cp\u003eRad-score = (0.0116 \u0026times; original_firstorder_Median\u003c/p\u003e \u003cp\u003e+\u0026thinsp;0.1696 \u0026times; original_glrlm_ShortRunEmphasis\u003c/p\u003e \u003cp\u003e+\u0026thinsp;0.0058 \u0026times; logarithm_gldm_LowGrayLevelEmphasis)\u003c/p\u003e \u003cp\u003e+\u0026thinsp;0.254 \u0026times; exponential_glszm_GrayLevelNonUniformity\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eThe Univariate and Multivariate Logistic Analysis of Parameters\u003c/h2\u003e \u003cp\u003eThe univariate analysis showed that T stage, N stage, M stage and Rad-score were significantly correlated with differentiation (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05). Multivariate logistic analysis showed that N stage (OR: 3.533, 95%CI: 1.223\u0026ndash;10.208, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.020), M stage (OR:3.321, 95%CI: 1.191\u0026ndash;9.257, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.022) and Rads-score (OR: 2.516, 95%CI: 1.561\u0026ndash;4.057, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001) were independent factors influencing differentiation (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eUnivariate and multivariate analysis of characteristics in gastric adenocarcinoma\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eVariables\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eUnivariate analysis (\u003cem\u003eP\u003c/em\u003e)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"3\" nameend=\"c5\" namest=\"c3\"\u003e \u003cp\u003eMultivariate logistic regression analysis\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eOR\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e95%CI\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cem\u003eP\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.462\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSex\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.150\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePrimary tumor location\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.539\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTumor diameter (cm)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.857\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT stage\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003csup\u003e*\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2.092\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.721\u0026ndash;6.067\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.174\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eN stage\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003csup\u003e*\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e3.533\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.223\u0026ndash;10.208\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.020\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eM stage\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003csup\u003e*\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e3.321\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.191\u0026ndash;9.257\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.022\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eECOG score\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.229\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCEA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.967\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCA199\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.808\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRad-score\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003csup\u003e*\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2.516\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.561\u0026ndash;4.057\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001*\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eConstruction, Evaluation and Validation of Nomogram\u003c/h2\u003e \u003cp\u003eBased on a combination of the N stage, M stage and Rad-score, a nomogram was created to predict tumor differentiation in patients with gastric cancer (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA). Each variable was modeled both independently and jointly, with subsequent evaluation of their predictive efficacy. The AUCs for predicting tumor differentiation grade reached 0.702, 0.705, 0.830, 0.787 and 0.872 in the training cohort (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eB), and 0.706, 0.781, 0.835, 0.842 and 0.935 in the validation cohort (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eC), respectively. The calibration curve of the nomogram indicated a strong concordance between the observed values and the predicted values in both the training and validation cohorts (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eD, E). Decision curve analysis (DCA) showed that the model developed had the greatest net benefit in training and validation cohorts (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eF, G). CIC showed good clinical applicability of the model in training and validation cohorts (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eH, I).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003eIdentification and Enrichment Analysis of DEGs\u003c/h2\u003e \u003cp\u003eA total of 61 DEGs were identified, comprising 49 up-regulated and 12 down-regulated genes (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eA). The heatmap displayed the expression profiles of these DEGs (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eB). GO enrichment analysis highlighted significant biological processes, including immunoglobulin-mediated immune response, B cell-mediated immunity, lymphocyte-mediated immunity, adaptive immune response and B cell receptor signaling pathway (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eC). Additionally, KEGG pathway analysis revealed enrichment in neuroactive ligand-receptor interaction, neuroactive ligand signaling and hormone signaling (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eD).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003eMachine Learning Screening Hub Gene\u003c/h2\u003e \u003cp\u003eUtilizing the LASSO logistic regression algorithm, six hub genes were identified from an initial pool of twenty hub genes (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eE, F). The twenty hub genes were subsequently prioritized based on their variable importance scores, determined through the RF algorithm, and those with a variable importance greater than 0.01 were selected as significant hub genes (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eG, H). Notably, both machine learning methodologies converged on the common hub gene IGHG1 (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eI).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003eIGHG1 Expression Differential and Functional Enrichment Analysis\u003c/h2\u003e \u003cp\u003eA comparative analysis of IGHG1 expression and Rad-scores between the poor differentiation and well/moderate differentiation groups revealed significantly higher levels of both IGHG1 gene expression (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05) and Rad-scores (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05) in the poor differentiation group compared to the well/moderate differentiation group (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eA, B). The correlation analysis indicated a significant association between the expression levels of IGHG1 and the Rad-score (r\u0026thinsp;=\u0026thinsp;0.946, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.001) (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eC). The GSEA results showed an increase in TGF-β signaling, epithelial-mesenchymal transition (EMT) and KRAS signaling and a decrease in MYC targets V1 in patients in the IGHG1 high expression group compared to the low expression group (Fig D, E).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003eFunctional Validation of the Hub Gene\u003c/h2\u003e \u003cp\u003eIGHG1 knockdown and overexpression were achieved using shRNA. Quantitative polymerase chain reaction (qPCR) and WB results confirmed the successful gene knockdown and overexpression operations (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eA, B, F, G). CCK-8 assays demonstrated that knockdown of the IGHG1 gene attenuated the proliferation level of MKN45 cells (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eC), while overexpression of the IGHG1 gene enhanced it (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eH). Wound healing and Transwell assays proved that knockdown of the IGHG1 gene reduced the migration and invasion abilities of MKN45 cells (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eD, E), whereas overexpression of the IGHG1 gene improved these capabilities (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eI, J).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"DISCUSSION","content":"\u003cp\u003eThe histological differentiation grade of gastric cancer is a crucial indicator for evaluating tumor biological behavior, formulating individualized treatment plans, and predicting prognosis (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e). Traditionally, the determination of differentiation grade relies on postoperative pathological examination, which cannot provide a reference for clinical decision-making before surgery. This study constructed a radiomics model based on CT images, achieving noninvasive prediction of gastric cancer differentiation grade, thereby offering a novel approach for precise preoperative diagnosis and treatment. This study showed that N stage, M stage and Rads-score were independent factors influencing differentiation. The predictive model based on N stage, M stage and Rads-score was developed. The outcomes derived from the ROC indicated that the clinical-radiomics model possesses a strong predictive accuracy. Furthermore, calibration curves demonstrated a favorable alignment between the actual observations and the predicted values. The DCA illustrated a significant net benefit attributed to the model. Additionally, the CIC confirmed the robust clinical applicability of the model. Analyzing the sequencing data identified IGHG1 as a pivotal gene within the radiomics model. The GSEA results showed an increase in TGF-β signaling, EMT and KRAS signaling pathway. Radiomics and its underlying mechanisms were revealed.\u003c/p\u003e \u003cp\u003eThe TNM staging system, as a cornerstone tool for prognostic assessment in gastric cancer, has been continuously refined in recent years, driven by advancements in molecular biology and imaging (\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e). A significant correlation exists between tumor differentiation degree and TNM stage. Data from this study revealed that the proportion of N2-3 stage patients was significantly higher in the poor differentiation group compared to the well/moderate differentiation group. Research indicated that tumor differentiation was an independent prognostic factor, exhibiting a significant correlation with both the overall TNM stage and its individual components. Well-differentiated tumors were generally associated with lower TNM stages, whereas poorly differentiated tumors tend to correlate with more advanced TNM stages. This association was particularly evident across different T stages, with less differentiated tumors showing a higher propensity for lymph node metastasis (\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e). The relationship between differentiation degree and lymph node metastasis risk holds substantial clinical significance, especially in postoperative risk assessment and treatment strategy formulation. Furthermore, other studies had highlighted significant correlations between tumor differentiation and individual components of TNM staging (such as T stage, N stage, and M stage), emphasizing the importance of considering tumor differentiation in the diagnosis and treatment of gastric cancer (\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn recent years, CT radiomics has emerged as a cutting-edge interdisciplinary field, demonstrating great potential in quantifying tumor phenotypes through high-throughput extraction and analysis of quantitative imaging features (\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e). Research on radiomics for evaluating gastric cancer differentiation is becoming a crucial breakthrough in improving the precision of preoperative diagnosis and treatment. Multiple retrospective studies confirmed that radiomic features extracted from preoperative contrast-enhanced CT (particularly in the portal venous phase) could be used to construct highly discriminative models for predicting poorly differentiated gastric cancer or distinguishing well/moderately differentiated gastric cancer (\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e). The imaging characteristics of poorly differentiated adenocarcinoma have been extensively analyzed to better understand their manifestations on CT and MRI. Studies found that poorly differentiated adenocarcinomas typically appear as isodense or hypodense on non-contrast CT scans, while exhibiting high signal intensity on T2-weighted imaging. By quantifying these differences through high-throughput feature analysis, radiomics overcomes the subjectivity limitations inherent in traditional visual assessments (\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e). In this study, features were screened, including first-order_Median, glrlm_ShortRunEmphasis, gldm_LowGrayLevelEmphasis and glszm_GrayLevelNonUniformity. First-order_Median represents the median gray level intensity within the ROI. Glrlm_ShortRunEmphasis is a measure of the distribution of short run lengths, with a greater value indicative of shorter run lengths and finer textural textures. Gldm_LowGrayLevelEmphasis measures the similarity of gray-level intensity values in the image, where a lower GLN value correlates with a greater similarity in intensity values.Glszm_GrayLevelNonUniformity measures the variability of gray-level intensity values in the image, with a lower value indicating more homogeneity in intensity values. These reflect the heterogeneity within the tumour.\u003c/p\u003e \u003cp\u003eIn recent years, radiomics has emerged as a pivotal bridge linking macroscopic imaging characteristics with microscopic molecular features, demonstrating tremendous potential in the field of precision diagnosis and treatment for gastric cancer. The findings of this study revealed that IGHG1 expression was significantly correlated with Rad-score (r\u0026thinsp;=\u0026thinsp;0.946, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.001). The GSEA results showed an increase in TGF-β signaling pathway, EMT and KRAS signaling and a decrease in MYC targets V1 in patients in the IGHG1 high expression group compared to the low expression group. A study revealed that IGHG1 promotes the migration and invasion of gastric cancer cells and is closely associated with EMT (\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e). Li et al. demonstrated that the IGHG1/AKT/GSK-3β/β-catenin signaling pathway plays a regulatory role in the malignant biological behavior of gastric cancer (\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e). This study revealed the crucial role of IGHG1 in gastric cancer cells, particularly in regulating cell proliferation, migration, and invasion. Our research demonstrates that IGHG1 not only promotes the proliferation of MKN45 cells but also modulates their migratory and invasive capabilities by regulating the activity of the TGF-β signaling pathway. This discovery holds significant importance for understanding the pathogenesis and progression of gastric cancer, especially in elucidating its underlying molecular mechanisms. The TGF-β signaling pathway is known to play a dual role in various cancers, capable of both suppressing tumor formation and promoting tumor progression (\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e). Therefore, by influencing the activation of the TGF-β signaling pathway, IGHG1 may play a pivotal role in tumor growth and metastasis, providing a potential molecular target for the development of novel therapeutic strategies in the future.\u003c/p\u003e \u003cp\u003eThe limitations of this study still warrant attention. Firstly, the retrospective design may introduce selection bias. Secondly, variations in CT scanner parameters across institutions could affect feature reproducibility, necessitating future multicenter validation to optimize standardization protocols.\u003c/p\u003e"},{"header":"CONCLUSION","content":"\u003cp\u003eThe predictive model based on N stage, M stage and Rad-score can effectively predict the differentiation in gastric cancer patients. Radiomics enables noninvasive prediction of tumor differentiation status while elucidating the expression levels of the IGHG1 and the underlying pathway.\u003c/p\u003e \u003cp\u003e \u003cstrong\u003eETHICAL APPROVAL AND CONSENT TO PARTICIPATE\u003c/strong\u003e \u003cp\u003e This study was performed in line with the principles of the Declaration of Helsinki. Approval was granted by the Ethics Committee of Anhui Provincial Hospital (Approval No. 2025-RE-262). Informed consent was obtained from all individual participants included in the study. Written informed consent for publication of their clinical details and/or clinical images was obtained from the patient/participant.\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cb\u003eFUNDING DECLARATION\u003c/b\u003e \u003c/p\u003e \u003cp\u003eThis study was supported by the Anhui Provincial Natural Science Foundation (Grant No. 2208085QH228).\u003c/p\u003e \u003cdiv id=\"Sec23\" class=\"Section2\"\u003e \u003ch2\u003eDATA AVAILABILITY STATEMENT\u003c/h2\u003e \u003cp\u003eThe raw RNA sequencing data generated in this study have been deposited in the NCBI Sequence Read Archive (SRA) and are publicly available under the BioProject accession number PRJNA1390632.\u003c/p\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eETHICAL APPROVAL AND CONSENT TO PARTICIPATE\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was performed in line with the principles of the Declaration of Helsinki. Approval was granted by the Ethics Committee of Anhui Provincial Hospital (Approval No. 2025-RE-262). Informed consent was obtained from all individual participants included in the study. Written informed consent for publication of their clinical details and/or clinical images was obtained from the patient/participant.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFUNDING DECLARATION\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was supported by the Anhui Provincial Natural Science Foundation (Grant No. 2208085QH228).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDATA AVAILABILITY STATEMENT\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe raw RNA sequencing data generated in this study have been deposited in the NCBI Sequence Read Archive (SRA) and are publicly available under the BioProject accession number PRJNA1390632.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eBray F, Laversanne M, Sung H, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2024; 74(3):229-63.\u003c/li\u003e\n\u003cli\u003eFeng F, Liu J, Wang F, et al. Prognostic value of differentiation status in gastric cancer. BMC Cancer. 2018; 18(1):865.\u003c/li\u003e\n\u003cli\u003eLee IS, Park YS, Lee JH, et al. Pathologic discordance of differentiation between endoscopic biopsy and postoperative specimen in mucosal gastric adenocarcinomas. Ann Surg Oncol. 2013; 20(13):4231-7.\u003c/li\u003e\n\u003cli\u003eKubota K, Suzuki A, Shiozaki H, Wada T, Kyosaka T, Kishida A. Accuracy of Multidetector-Row Computed Tomography in the Preoperative Diagnosis of Lymph Node Metastasis in Patients with Gastric Cancer. Gastrointest Tumors. 2017; 3(3-4):163-70.\u003c/li\u003e\n\u003cli\u003eLambin P, Leijenaar RTH, Deist TM, et al. Radiomics: the bridge between medical imaging and personalized medicine. Nat Rev Clin Oncol. 2017; 14(12):749-62.\u003c/li\u003e\n\u003cli\u003eLee S, Choe EK, Kim SY, Kim HS, Park KJ, Kim D. Liver imaging features by convolutional neural network to predict the metachronous liver metastasis in stage I-III colorectal cancer patients based on preoperative abdominal CT scan. BMC Bioinformatics. 2020; 21(Suppl 13):382.\u003c/li\u003e\n\u003cli\u003eGao X, Ma T, Cui J, et al. A CT-based Radiomics Model for Prediction of Lymph Node Metastasis in Early Stage Gastric Cancer. Acad Radiol. 2021; 28(6):e155-e64.\u003c/li\u003e\n\u003cli\u003eFeng B, Huang L, Li C, et al. A Heterogeneity Radiomic Nomogram for Preoperative Differentiation of Primary Gastric Lymphoma From Borrmann Type IV Gastric Cancer. J Comput Assist Tomogr. 2021; 45(2):191-202.\u003c/li\u003e\n\u003cli\u003eGuo Y, Li H, Chen H, et al. Metagenomic next-generation sequencing to identify pathogens and cancer in lung biopsy tissue. EBioMedicine. 2021; 73:103639.\u003c/li\u003e\n\u003cli\u003eLe NQK, Kha QH, Nguyen VH, Chen YC, Cheng SJ, Chen CY. Machine Learning-Based Radiomics Signatures for EGFR and KRAS Mutations Prediction in Non-Small-Cell Lung Cancer. Int J Mol Sci. 2021; 22(17).\u003c/li\u003e\n\u003cli\u003eTian R, Li Y, Jia C, et al. Radiomics Model for Predicting TP53 Status Using CT and Machine Learning Approach in Laryngeal Squamous Cell Carcinoma. Front Oncol. 2022; 12:823428.\u003c/li\u003e\n\u003cli\u003eWang M, Mei T, Gong Y. The quality and clinical translation of radiomics studies based on MRI for predicting Ki-67 levels in patients with breast cancer. Br J Radiol. 2023; 96(1150):20230172.\u003c/li\u003e\n\u003cli\u003eHalder SK, Beauchamp RD, Datta PK. A specific inhibitor of TGF-beta receptor kinase, SB-431542, as a potent antitumor agent for human cancers. Neoplasia. 2005; 7(5):509-21.\u003c/li\u003e\n\u003cli\u003eJapanese Gastric Cancer A. Japanese Gastric Cancer Treatment Guidelines 2021 (6th edition). Gastric Cancer. 2023; 26(1):1-25.\u003c/li\u003e\n\u003cli\u003eLopez Sala P, Leturia Etxeberria M, Inchausti Iguiniz E, Astiazaran Rodriguez A, Aguirre Oteiza MI, Zubizarreta Etxaniz M. Gastric adenocarcinoma: A review of the TNM classification system and ways of spreading. Radiologia (Engl Ed). 2023; 65(1):66-80.\u003c/li\u003e\n\u003cli\u003eLiu S, Shi H, Ji C, et al. Preoperative CT texture analysis of gastric cancer: correlations with postoperative TNM staging. Clin Radiol. 2018; 73(8):756 e1- e9.\u003c/li\u003e\n\u003cli\u003eDerwinger K, Kodeda K, Bexe-Lindskog E, Taflin H. Tumour differentiation grade is associated with TNM staging and the risk of node metastasis in colorectal cancer. Acta Oncol. 2010; 49(1):57-62.\u003c/li\u003e\n\u003cli\u003eLiu S, Wang H, Guan W, et al. Preoperative apparent diffusion coefficient value of gastric cancer by diffusion-weighted imaging: Correlations with postoperative TNM staging. J Magn Reson Imaging. 2015; 42(3):837-43.\u003c/li\u003e\n\u003cli\u003eLambin P, Rios-Velazquez E, Leijenaar R, et al. Radiomics: extracting more information from medical images using advanced feature analysis. Eur J Cancer. 2012; 48(4):441-6.\u003c/li\u003e\n\u003cli\u003eGillies RJ, Kinahan PE, Hricak H. Radiomics: Images Are More than Pictures, They Are Data. Radiology. 2016; 278(2):563-77.\u003c/li\u003e\n\u003cli\u003eHuang J, Yao H, Li Y, et al. Development and validation of a CT-based radiomics nomogram for preoperative prediction of tumor histologic grade in gastric adenocarcinoma. Chin J Cancer Res. 2021; 33(1):69-78.\u003c/li\u003e\n\u003cli\u003eLi Q, Qi L, Feng QX, et al. Machine Learning-Based Computational Models Derived From Large-Scale Radiographic-Radiomic Images Can Help Predict Adverse Histopathological Status of Gastric Cancer. Clin Transl Gastroenterol. 2019; 10(10):e00079.\u003c/li\u003e\n\u003cli\u003eBaba A, Matsushima S, Kessoku H, et al. Radiological features of thyroid-like low-grade nasopharyngeal papillary adenocarcinoma: case series and systematic review. Neuroradiology. 2024; 66(2):249-59.\u003c/li\u003e\n\u003cli\u003eLi Y, Wang P, Ye D, et al. IGHG1 induces EMT in gastric cancer cells by regulating TGF-beta/SMAD3 signaling pathway. J Cancer. 2021; 12(12):3458-67.\u003c/li\u003e\n\u003cli\u003eLi X, Chen W, Yang C, et al. IGHG1 upregulation promoted gastric cancer malignancy via AKT/GSK-3beta/beta-Catenin pathway. Cancer Cell Int. 2021; 21(1):397.\u003c/li\u003e\n\u003cli\u003eFabregat I, Fernando J, Mainez J, Sancho P. TGF-beta signaling in cancer treatment. Curr Pharm Des. 2014; 20(17):2934-47.\u003c/li\u003e\n\u003cli\u003eZi Z. Molecular Engineering of the TGF-beta Signaling Pathway. J Mol Biol. 2019; 431(15):2644-54.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-imaging","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bmim","sideBox":"Learn more about [BMC Medical Imaging](http://bmcmedimaging.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bmim/default.aspx","title":"BMC Medical Imaging","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"gastric cancer, radiomics, machine learning, differentiation, mechanism","lastPublishedDoi":"10.21203/rs.3.rs-7782471/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7782471/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eObjectives\u003c/h2\u003e \u003cp\u003eTo develop a CT-based radiomics model for predicting tumor differentiation in patients with gastric cancer. Exploring Rad-score correlation with gene expression and related mechanisms.\u003c/p\u003e\u003ch2\u003eMaterials and Methods\u003c/h2\u003e \u003cp\u003eClinical data and imaging of 162 gastric cancer patients were retrospectively analyzed. Patients were randomly allocated to training and validation cohorts. The least absolute shrinkage and selection operator (LASSO) methods were utilized to identify characteristics and develop the Rad-score. Clinical-radiomics models were developed and evaluated for predictive efficacy and clinical incremental value. Screening hub genes and exploring the pathways of hub genes through machine learning, bioinformatics analysis and experimental validation.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eClinical-radiomics models based on N stage, M stage and Rad-score were developed. The receiver operating characteristic (ROC) curves indicated that the model had good predictive accuracy in the training (AUC\u0026thinsp;=\u0026thinsp;0.872) and validation groups (AUC\u0026thinsp;=\u0026thinsp;0.935). The calibration curves indicated a strong correlation between the observed values and the predicted outcomes. The decision curve analysis demonstrated a substantial net benefit associated with the clinical-radiomics model. The clinical impact curve (CIC) illustrated the effective clinical applicability of this model. Analysis of the sequencing data revealed that the key gene IGHG1 was significantly associated with Rad-score. The possible mechanisms are related to the TGF-β signaling, epithelial-mesenchymal transition and KRAS signaling pathway.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eThe predictive model based on N stage, M stage and Rad-score can effectively predict the differentiation in gastric cancer patients. Radiomics enables noninvasive prediction of tumor differentiation status while elucidating the expression levels of the IGHG1 and the underlying pathway.\u003c/p\u003e","manuscriptTitle":"A CT-based Machine Learning Radiomics Model for the Prediction of Gastric Cancer Differentiation and Mechanism Exploration","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-12-30 01:04:10","doi":"10.21203/rs.3.rs-7782471/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-02-10T06:07:52+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-02-07T08:57:10+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"24113790966303138464520094019787596686","date":"2026-01-26T11:51:43+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-01-17T16:46:41+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"84987615769044987932368976287640367542","date":"2025-12-24T10:04:04+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"42822484111880783952947571488195520438","date":"2025-12-22T13:49:35+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-12-22T08:44:31+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-12-22T06:41:24+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2025-12-19T05:21:44+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-12-18T16:53:17+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Medical Imaging","date":"2025-12-18T16:00:08+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-imaging","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bmim","sideBox":"Learn more about [BMC Medical Imaging](http://bmcmedimaging.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bmim/default.aspx","title":"BMC Medical Imaging","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"8e94d11d-90d7-45c1-a72d-2b8a1c2166e3","owner":[],"postedDate":"December 30th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"in-revision","subjectAreas":[],"tags":[],"updatedAt":"2026-02-10T06:23:13+00:00","versionOfRecord":[],"versionCreatedAt":"2025-12-30 01:04:10","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7782471","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7782471","identity":"rs-7782471","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}