CT semantic features of primary gastric cancer: a preoperative predictive model for peritoneal metastasis with superior efficacy to conventional direct CT assessment.

CT semantic features of primary gastric cancer: a preoperative predictive model for peritoneal metastasis with superior efficacy to conventional direct CT assessment - PMC Skip to main content An official website of the United States government Here's how you know Here's how you know Search Log in Primary site navigation Search Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice BMC Gastroenterol . 2026 May 13;26:333. doi: 10.1186/s12876-026-04891-6 CT semantic features of primary gastric cancer: a preoperative predictive model for peritoneal metastasis with superior efficacy to conventional direct CT assessment Shuxiang Chen 1, ✉ , Huijuan Zhang 1 , Yifan Chen 2 , Shuo Chen 3 Author information Article notes Copyright and License information 1 Department of Radiology, Shengli Clinical Medical College of Fujian Medical University, Fujian Provincial Hospital, Fuzhou University Affiliated Provincial Hospital, Fuzhou, Fujian 350001 China 2 Department of Breast Surgery, Shengli Clinical Medical College of Fujian Medical University, Fujian Provincial Hospital, Fuzhou University Affiliated Provincial Hospital, Fuzhou, Fujian 350001 China 3 Department of Obstetrics and gynecology, Shengli Clinical Medical College of Fujian Medical University, Fujian Provincial Hospital, Fuzhou University Affiliated Provincial Hospital, Fuzhou, Fujian 350001 China ✉ Corresponding author. Received 2025 Oct 20; Accepted 2026 Apr 28; Collection date 2026. © The Author(s) 2026 Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/ . PMC Copyright notice PMCID: PMC13224590  PMID: 42129648 Abstract Objectives This study aims to evaluate the utility of CT semantic features of primary Gastric Cancer(GC) compared to conventional CT assessments for predicting Peritoneal Metastasis(PM) , and to construct a preoperative predictive model. Methods We conducted a retrospective analysis involving 257 pathologically confirmed GC patients (92 with PM and 165 without PM), utilizing preoperative contrast-enhanced abdominal CT alongside clinicopathological data. Univariate and multivariate logistic regression analyses were performed to identify risk factors for PM, leading to the development of three predictive models: one based on primary tumor CT signs, another on peritoneal CT signs, and a combined model integrating both sets of features. Results Independent PM predictors included the primary tumor's maximum size, serosal invasion, thickness, enhancement, and the presence of ascites. The primary tumor model demonstrated superior performance (AUC=0.920) compared to the peritoneal model (AUC=0.822, p<0.001). No significant difference was observed between the primary tumor model and the combined model (AUC=0.936, p=0.178). The combined model exhibited the highest sensitivity at 75.0%, while all models maintained a specificity of 98.2%. Conclusions CT semantic features of primary GC and ascites are effective in predicting PM. The primary tumor-based model surpasses conventional CT in performance, and the combined model further enhances sensitivity. This methodology improves preoperative PM assessment, may help reduce the incidence of non-therapeutic surgeries, and contributes positively to the prognosis of GC patients. Keywords: Gastric cancer, Peritoneal metastasis, Computed tomography, X-ray Introduction Gastric cancer (GC) is a common malignant tumor of the digestive system and the most prevalent among all gastrointestinal malignancies. Peritoneal Metastasis (PM) represents the most frequent pattern of metastasis and recurrence in GC. According to the TNM staging system established by the American Joint Committee on Cancer, the presence of PM is classified as M1, while its absence is classified as M0. Consequently, the presence of PM directly determines a patient’s eligibility for curative surgery. Computed tomography (CT) is the main imaging modality for evaluating GC, with current research primarily focusing on its efficacy for prognosis, TNM staging, and the identification of direct signs of PM, such as peritoneal opacification, thickening, nodularity, and enhancement [ 1 , 2 ]. While advanced PM is relatively straightforward to diagnose, it is associated with poor treatment outcomes and an unfavorable prognosis. In contrast, early-stage PM poses diagnostic challenges due to the absence of obvious clinical symptoms and the limitations of conventional imaging, which result in a low detection rate. This situation presents a significant clinical challenge. Typically, these lesions are small or occult on CT, frequently missed preoperatively, and often only confirmed during surgery [ 3 , 4 ]. Although the interpretation of CT signs provides useful information, some hidden image signs related to clinical findings may be overlooked due to the limited contrast perceived by the naked eye. This limitation hinders a more accurate assessment of occult PM [ 5 , 6 ]. Therefore, developing a non-invasive and preoperative evaluation approach for occult PM in GC is crucial to avoid unnecessary surgeries and to select the most appropriate treatment options in clinical practice. Despite advancements in multi-detector CT technology, the sensitivity for identifying PM remains inadequate, with reported rates ranging from 28.3% to 50.9%. While laparoscopy is regarded as the gold standard for diagnosis, it is an invasive procedure that carries inherent risks and incurs substantial costs [ 7 , 8 ].Furthermore, the field of radiomics has yet to establish standardized, automated, and interpretable predictive models that can be seamlessly integrated into clinical workflows to facilitate truly individualized treatment plans for patients with GC [ 7 , 9 ]. CT currently stands as the predominant imaging modality for the preoperative staging and assessment of metastasis in GC. Traditionally, CT interpretation has been heavily reliant on the direct observation of abnormal peritoneal surface indicators—such as peritoneal thickening, nodular shadows, alterations in the greater omentum, and the presence of ascites—for diagnosing PM [ 10 , 11 ]. However, these direct indicators exhibit limited sensitivity in detecting early or microscopic PM and are associated with a high false-negative rate. Consequently, some patients may miss optimal treatment opportunities or undergo unnecessary surgical procedures. As a result, the development of more sensitive and specific preoperative assessment methods has emerged as a focal point of clinical research. In recent years, researchers have increasingly focused on the CT semantic features of the primary gastric cancer tumor itself, as these features may provide valuable insights for predicting metastatic potential [ 9 , 10 ]. Therefore, this study aims to assess the predictive value of easily accessible imaging features obtained from routine contrast-enhanced whole-abdominal CT scans. We hypothesize that semantic feature analysis of primary GC may assist in the preoperative identification and early detection of high-risk patients. This approach may have the potential to improve the accuracy of preoperative predictions for peritoneal metastasis, inform preoperative treatment planning, avoid unnecessary surgeries, and ultimately lead to better patient outcomes. Materials and methods Ethical approval and patient consent This retrospective study was approved by the Institutional Review Board (IRB) of our hospital. Informed consent was waived due to the retrospective nature of the study and the use of de-identified data. Patient cohort A total of 257 patients were enrolled in this study, comprising 92 patients in PM group and 165 patients in the non-PM group. Peritoneal metastasis in the PM group was confirmed surgically in 69 patients, by abdominal exploration in 21 patients, and via ascites cytology in 2 patients. The cohort included 139 males and 118 females, with a mean age of 63 ± 11 years (range 25–91 years). According to pathological classification, there were 85 cases of tubular adenocarcinoma, 6 cases of papillary adenocarcinoma, 21 cases of moderately differentiated adenocarcinoma, 102 cases of poorly differentiated adenocarcinoma, 32 cases of poorly cohesive or mucinous adenocarcinoma, 5 cases of signet ring cell carcinoma, and 6 cases of mixed or undifferentiated carcinoma. Inclusion criteria Histologically confirmed primary GC via gastroscopy or postoperative pathology; Preoperative whole-abdominal contrast-enhanced multi-detector CT performed within 2 weeks before surgery; Complete clinical, imaging, and pathological data; No prior anticancer treatment (e.g. chemotherapy, radiotherapy). Exclusion criteria Incomplete imaging or clinical data; History of liver cirrhosis, heart failure, peritonitis, abdominal surgery, or other malignancies. CT image protocol Patients underwent whole-abdominal contrast-enhanced CT (diaphragm to pubic symphysis) using Somatom Definition AS 128, Somatom Sensation 64, or LightSpeed 128 scanners (GE Healthcare). Scanning parameters: tube voltage 120 kV, tube current 180–200 mAs, slice thickness/interval 3 mm, field of view 35–40 cm, matrix 256 × 256, pitch 0.6–1.0. Pre-scanning preparation: Fasting for ≥ 6 h, followed by oral ingestion of 600–1000 mL water over 15 min (gastric distension). Patients held their breath in the supine position during scanning. After CT plain scan, iodinated contrast agent (ioversol, 320 mgI/mL, Jiangsu Hengrui) was intravenously injected (1.5 mL/kg, 3 mL/s via dorsal hand vein, Opti Vantage DH injector). Images were acquired in three phases: arterial (25–30 s post-injection), portal-venous (60–70 s), and delayed (180 s). All images were reviewed on a Picture Archiving and Communication System workstation (GE Healthcare). CT image analysis Portal venous phase CT images were evaluated independently by two radiologists with 10 and 20 years of experience in abdominal CT imaging, respectively. Both were blinded to all clinical and pathological data of the patients. The following CT features were assessed by consensus: Location of the gastric lesion (esophagogastric junction, fundus, body, antrum, or diffuse involvement); Maximum tumor size (defined as the longest diameter measured at the largest cross-sectional area); Tumor thickness (measured at the thickest part of the lesion); Degree of lesion enhancement (calculated as the difference in Hounsfield units between portal venous phase and precontrast phase images); Serosal involvement (defined as irregular or rough outer margins of the tumor); Enlarged lymph nodes (short-axis diameter > 10 mm); Presence of ascites; Direct signs of peritoneal metastasis, including omental nodules, omental cake-like thickening, and irregular peritoneal thickening with contrast enhancement. Any discrepancies between the two radiologists were resolved by a third independent radiologist with more than 25 years of experience in abdominal imaging, and a final consensus was reached through joint re-evaluation of the discordant images. Statistical analysis Statistical analyses were conducted using SPSS 22.0 (IBM), MedCalc 15.2.2, and R 3.6.3. Continuous variables were compared using the Student’s t-test or the Mann–Whitney U test; categorical variables were compared using the Chi‑square test or Fisher’s exact test; paired non‑parametric comparisons were performed with the Wilcoxon rank‑sum test. Univariate analysis identified variables that differed between the PM and non‑PM groups; variables with p < 0.05 were included in a multivariate logistic regression model. Model performance was assessed via receiver operating characteristic curve analysis, and areas under the curve were compared using the DeLong test in MedCalc. A two‑tailed p < 0.05 was considered statistically significant. Results Patient demographics and baseline characteristics A comparison of clinical, pathological, and imaging features between the non-PM ( n = 165) and PM ( n = 92) groups is presented in Table 1 . Significant differences were observed in sex, tumor differentiation, maximum tumor size, tumor thickness, lesion enhancement, serosal involvement, direct PM signs, and ascites ( p < 0.05). No significant differences were observed in age or lymph node involvement ( p > 0.05). Table 1. Patient demographics and baseline characteristics Characteristic group non-PM N = 165 PM N = 92 p -value Sex, n (%) < 0.001 2 Female 46 (27.9%) 72 (78.3%) Male 119 (72.1%) 20 (21.7%) Age, Mean ± SD 63 ± 11 63 ± 11 0.901 3 Differentiation, n (%) < 0.001 4 Well 79 (47.9%) 30 (32.6%) Moderately 17 (10.3%) 1 (1.1%) Poorly 69 (41.8%) 61 (66.3%) maxsize, Mean ± SD 1.5 ± 0.5 8.2 ± 7.2 < 0.001 3 thickness, Mean ± SD 1.28 ± 0.45 1.90 ± 0.73 < 0.001 3 enhancement, Mean ± SD 46 ± 24 53 ± 32 0.046 3 Lymphnode, n (%) Negative 76 (46.1%) 44 (47.8%) 0.786 2 Positive 89 (53.9%) 48 (52.2%) CT_serous, n (%) Negative 59 (35.8%) 6 (6.5%) < 0.001 2 Positive 106 (64.2%) 86 (93.5%) CT_lymphnode, n (%) Negative 95 (57.6%) 20 (21.7%) < 0.001 2 Positive 70 (42.4%) 72 (78.3%) CT_peritoneum, n (%) Negative 162 (98.2%) 31 (33.7%) < 0.001 2 Positive 3 (1.8%) 61 (66.3%) ascites, n (%) Negative 162 (98.2%) 45 (48.9%) < 0.001 2 Positive 3 (1.8%) 47 (51.1%) Open in a new tab 1 Pearson’s Chi-squared test; Welch Two Sample t-test; Fisher’s exact test 2 Pearson’s Chi-squared test 3 Welch Two Sample t-test 4 Fisher’s exact test CT semantic feature model and combined model CT semantic feature model (primary tumor + ascites) As shown in Fig. 1 , multivariate logistic regression identified three independent predictors of peritoneal metastasis: Serosal involvement (CT_serous), maximum tumor size (maxsize) and ascites, all of which were statistically significant(p < 0.05). In contrast, tumor thickness and lesion enhancement showed only a statistical trend, with p-values greater than 0.05. Fig. 1. Open in a new tab CT semantic feature model Combined model (primary tumor + ascites + peritoneal signs) The combined model as shown in Fig. 2 , Multivariable logistic regression analysis revealed that tumor maximum size (OR = 4.33, p = 0.002), tumor thickness (OR = 2.74, p = 0.027), ascites (OR = 11.01, p = 0.003), and CT peritoneal signs (OR = 15.58, p = 0.003) were independent predictive factors for peritoneal metastasis in gastric cancer, while CT serosal invasion showed a borderline association ( p = 0.056) and enhancement was not significant ( p = 0.114). Fig. 2. Open in a new tab Combined model (primary tumor + ascites + peritoneal signs) Model parameter and comparison The univariable model based solely on CT_peritoneum showed the poorest fit and discriminative performance, indicating insufficient predictive value of this single CT sign (Table 2 ). In contrast, the multivariable model incorporating CT_serous, maxsize, thickness, enhancement, and ascites demonstrated excellent discriminative ability (AUC = 0.92, 95% CI: 0.878–0.962) while using concise, clinically interpretable variables. The combined model (including all variables) achieved optimal fit indices (lowest AIC and BIC; highest LogLik) and the highest AUC (0.936, 95% CI: 0.902–0.970); however, its discriminative performance did not differ significantly from that of the multivariable model (Table 3 ; Fig. 3 ), suggesting no added benefit from including additional variables. All three models exhibited high specificity (98.2%) for diagnosing PM in primary GC, with no differences in reducing false positives. In terms of sensitivity, the combined model showed the highest sensitivity (75%), followed by the multivariable model without CT_peritoneum (73.9%), while the univariable CT_peritoneum model had the lowest (66.3%). This further confirms the superior diagnostic performance of the combined and multivariable models over the univariable model (Table 4 ). Table 2. Model parameter Model AIC BIC LogLik AUC CI_95 Multivariable (CT_serous+maxsize+thickness+enhancement+ascites) 158.502 179.796 -73.251 0.920 (0.878–0.962) Univariable (CT_peritoneum) 198.329 205.423 -97.164 0.822 (0.773–0.872) Combined (All variables) 150.961 175.805 -68.481 0.936 (0.902–0.970) Open in a new tab Table 3. Model comparison Comparison AUC1 AUC2 Z P_value CT.tumor.signs vs. CT_peritoneum 0.920 0.822 3.590 < 0.001 CT.tumor.signs vs. Combined.model 0.920 0.936 -1.345 0.178 CT_peritoneum vs. Combined.model 0.822 0.936 -5.178 < 0.001 Open in a new tab Fig. 3. Open in a new tab ROC curves of the three predictive models. Red line: primary tumor model; green line: peritoneal model; blue line: combined model Table 4. Quantification of diagnostic performance of predictive models Model True_Positive False_Positive True_Negative False_Negative Sensitivity Specificity Multivariate without CT_peritoneum 68 3 162 24 0.739 0.982 Univariate CT_peritoneum 61 3 162 31 0.663 0.982 Multivariate with CT_peritoneum (combined models) 69 3 162 23 0.750 0.982 Open in a new tab Representative imaging and histopathological findings As shown in Fig. 4 , two representative cases are presented, both of which were predicted by the concise multivariable model to be at high risk of PM. Case A (Fig. 4 A) was a 53‑year‑old female with poorly differentiated gastric cancer. CT revealed diffuse gastric wall thickening without definite peritoneal or omental lesions; however, PM was confirmed by mesenteric biopsy. Case B (Fig. 4 B) was a 73‑year‑old female with moderately differentiated gastric cancer. CT demonstrated diffuse gastric wall thickening, omental and mesenteric nodular thickening, and massive ascites, with PM confirmed intraoperatively. Fig. 4. Open in a new tab Representative case imaging. A 53-year-old female with poorly differentiated gastric adenocarcinoma: CT shows diffuse gastric wall thickening (long arrow) and no overt peritoneal abnormalities (short arrow); biopsy confirms PM. B 73-year-old female with moderately differentiated gastric adenocarcinoma: CT shows gastric wall thickening (long arrow), omental nodular thickening (short arrow), and ascites (asterisk); surgery confirms PM Discussion PM represents a leading driver of treatment failure and mortality in gastric cancer, and its accurate preoperative diagnosis is pivotal for guiding therapeutic strategies—most notably enabling conversion therapy to potentially achieve R0 resection, a critical determinant of improved patient survival [ 12 , 13 ]. Despite CT being the clinical gold standard for GC staging, conventional biomarkers and imaging modalities (including CT) exhibit limited sensitivity and specificity for PM detection, particularly for occult or micrometastatic lesions [ 14 – 17 ]. While CT can identify PM based on classic radiological signs (e.g. peritoneal turbidity, thickening, nodularity), its inability to reliably detect micrometastases underscores an urgent need for refined diagnostic approaches. The present retrospective study addressed this unmet clinical need by confirming that CT semantic features of primary GC—including maximum tumor diameter, gastric wall thickness, venous phase enhancement, and serosal invasion—and ascites are independent predictors of PM. The predictive model developed using these primary tumor CT features demonstrated superior diagnostic performance: it achieved an AUC of 0.92 and a specificity of 98.2%, significantly outperforming the model relying exclusively on direct peritoneal assessment (e.g., the univariate CT-peritoneum model yielded a sensitivity of 0.663, compared to 0.75 for the multivariable model incorporating primary tumor features). Notably, the combined model—integrating both primary tumor and peritoneal signs—failed to demonstrate a statistically significant improvement over the tumor-only model, though it still retained advantages relative to the peritoneum-only model. This finding highlights that semantic features of the primary tumor already encapsulate core information about metastatic potential—information that is more predictive than early, subtle morphological changes in the peritoneum—suggesting that future model optimization should prioritize the exploration and quantification of deep radiomic or deep learning-derived features of the primary tumor. Mechanistically, the link between primary tumor CT features and PM is grounded in established tumor biology. Larger tumor size and serosal invasion directly increase the likelihood of tumor cell detachment and intra-abdominal dissemination [ 18 , 19 ], while venous phase enhancement—selected for our analysis due to its ability to reflect dysfunctional tumor neovasculature —correlates with microvessel density and angiogenic activity, key drivers of tumor invasiveness [ 20 ]. Complementing these observations, recent single-cell RNA sequencing studies have delineated the central role of the tumor microenvironment in PM pathogenesis: tumor-associated macrophages (TAMs) and mast cells mediate pre-metastatic niche formation via the CCL5-CCR1 signaling axis [ 21 ], whereas GC-derived exosomal miR-196a-5p stabilizes Snai1 to drive mesothelial-mesenchymal transition (MMT) in peritoneal mesothelial cells [ 22 ]. Together, these findings provide a molecular rationale for why CT features of the primary tumor can serve as robust surrogates for PM risk. Clinically, this non-invasive CT-based model offers substantial translational value. Its high specificity (98.2%) mitigates the risk of inappropriately excluding non-PM patients from potentially curative radical surgery, while its capacity to stratify patients by PM risk enables targeted laparoscopic confirmation—avoiding the routine use of invasive laparoscopy, which is not universally feasible [ 8 , 23 ]. This aligns with prior literature demonstrating that occult PM is disproportionately prevalent in high-risk subgroups (e.g. patients with positive clinical M staging or signet ring cell histology [ 8 ] and supports the optimal allocation of medical resources. Numerous studies have demonstrated that models extracting hundreds to thousands of quantitative features from CT images can achieve high accuracy in predicting occult PM [ 17 , 24 , 25 ]. For example, multiphase contrast-enhanced CT radiomics models have shown AUCs exceeding 0.87, significantly outperforming single-phase analyses [ 24 ]. Furthermore, nomogram based on the comprehensive CT texture analysis of GC primary tumor and peritoneotome combined with conventional CT signs can predict occult PM [ 17 ]. The integration of radiomics signatures with clinical factors into nomograms or fusion models has consistently yielded superior predictive efficacy compared to either approach alone [ 17 , 25 , 26 ]. Importantly, unlike radiomic models that require dedicated software for feature extraction [ 17 – 19 ].Our semantic feature-based approach, while more interpretable than radiomics, follows the same integrative principle: it synthesizes direct visual assessments of the tumor and peritoneum into a unified risk score. By leveraging visually assessable CT semantic features derived from routine preoperative imaging, the model improves clinical translatability and accessibility. Despite these strengths, the study has several limitations. First, its retrospective design introduces potential selection bias, as it utilized non-prospective, non-randomized cohorts. Second, the single-center data may limit external generalizability; validation in large, multi-center prospective cohorts is therefore critical. Third, the analysis was restricted to the venous phase of CT imaging, which was chosen to optimize tumor and PM visualization; the diagnostic performance of the arterial phase requires further investigation [ 27 , 28 ]. Fourth, variability in scanners and acquisition parameters during the extended data collection period may have introduced bias. Finally, emerging liquid biopsy biomarkers (e.g. circulating tumor cells [CTCs] and circulating tumor DNA [ctDNA]), which demonstrate high prognostic value for metachronous peritoneal metastasis, were not incorporated [ 29 ]. Integrating these biomarkers with radiomic features could improve predictive accuracy. Future research should prioritize two key areas: validating the model in multi-center prospective studies to confirm its generalizability and exploring its integration with novel therapeutic strategies. For instance, studies could investigate the utility of prophylactic hyperthermic intraperitoneal chemotherapy (HIPEC) for model-identified high-risk patients, even in cases of negative diagnostic laparoscopy [ 30 ]. Furthermore, the model could help stratify patients with advanced gastric cancer for precision therapies—such as immunotherapy, regional chimeric antigen receptor T-cell (CAR-T) therapy, or targeted agents—to identify those most likely to benefit [ 30 – 33 ]. Further exploration of associations between CT features and molecular targets may also enable cross-modal integration, ultimately supporting a more personalized approach to patient care. Conclusions Our results confirm that integrating multiple CT semantic features into a predictive model effectively enhances PM diagnostic accuracy in primary gastric cancer. Notably, the concise multivariable model may be more clinically applicable in routine practice, as it balances diagnostic performance with operational convenience, while the combined model could be considered for high-risk patient populations. Acknowledgements The authors would like to thank all participants who were enrolled in this study. Thanks also to the the radiologists and pathologists of Fuzhou University Affiliated Provincial Hospital for their assistance in CT image analysis and pathological confirmation. Abbreviations GC Gastric Cancer PM Peritoneal Metastasis CT Computed Tomography ROC Receiver Operating Characteristic AUC Area Under the Curve PACS Picture Archiving and Communication System HIPEC Hyperthermic Intraperitoneal Chemotherapy CTCs Circulating Tumor Cells ctDNA Circulating Tumor DNA TME Tumor Microenvironment MMT Mesothelial-Mesenchymal Transition AJCC American Joint Committee on Cancer Authors' contributions Conceptualization, SxC; methodology, SxC; software, HjZ, YfC, SC; validation, HjZ, YfC, SC; formal analysis, HjZ, YfC, SC; investigation, HjZ, YfC, SC; resources, SxC; data curation, HjZ, YfC, SC; writing—original draft preparation, SxC; writing—review and editing, SxC; visualization, HjZ, YfC, SC; supervision, SxC, HjZ; project administration, SxC; funding acquisition, SxC. All authors read and approved the final manuscript. Funding This research was funded by the Joint Funds for the Innovation of Science and Technology of Fujian Province, grant number 2024Y9040, and the Startup Fund for Scientific Research of Fujian Medical University, grant number 2018QH1138. Data availability The data used to support the findings of this study are available from the corresponding author upon reasonable request. The data that support the findings of this study are available from the Fuzhou University Affiliated Provincial Hospital, but restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available. Data are however available from the corresponding author upon reasonable request and with permission of the Fuzhou University Affiliated Provincial Hospital. Declarations Ethics approval and consent to participate This study was a retrospective study and was approved by the ethics committee of the Fuzhou University Affiliated Provincial Hospital (No.: 2024-09-080). Written informed consent was waived by the Institutional Review Board. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References 1. Tsurumaru D, Nishimuta Y, Muraki T, et al. Gastric cancer with synchronous and metachronous hepatic metastasis predicted by enhancement pattern on multiphasic contrast-enhanced CT. Eur J Radiol. 2018;108:165–71. [ DOI ] [ PubMed ] [ Google Scholar ] 2. Ji ZH, Peng KW, Li Y. Intraperitoneal free cancer cells in gastric cancer: pathology of peritoneal carcinomatosis and rationale for intraperitoneal chemotherapy/hyperthermic intraperitoneal chemotherapy in gastric cancer. Transl Gastroenterol Hepatol. 2016;1:69. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 3. Chen QY, Liu ZY, Zhong Q, et al. An intraoperative model for predicting survival and deciding therapeutic schedules: a comprehensive analysis of peritoneal metastasis in patients with advanced gastric cancer. Front Oncol. 2020;10:550526. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 4. Parisi A. Porzio G, Ficorella C. Multimodality treatment in metastatic gastric cancer: from past to next future. Cancers (Basel). 2020;12(9):12092598. [ DOI ] [ PMC free article ] [ PubMed ] 5. Lee IS, Lee H, Hur H et al. Transcriptomic profiling identifies a risk stratification signature for predicting peritoneal recurrence and micrometastasis in gastric cancer. Clin Cancer Res. 2021;27(8):2292–300. [ DOI ] [ PMC free article ] [ PubMed ] 6. Liu S, He J, Liu S, et al. Radiomics analysis using contrast-enhanced CT for preoperative prediction of occult peritoneal metastasis in advanced gastric cancer. Eur Radiol. 2020;30(1):239–46. [ DOI ] [ PubMed ] [ Google Scholar ] 7. Ho S, Tay KV. Systematic review of diagnostic tools for peritoneal metastasis in gastric cancer-staging laparoscopy and its alternatives. World J Gastrointest Surg. 2023;15(10):2280–93. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 8. von Bechtolsheim F, Spindler M, Jungmann V, et al. Predictive factors for the detection of occult metastases during staging laparoscopy in patients with gastric carcinoma and adenocarcinoma of the esophagogastric junction. Langenbecks Arch Surg. 2025;410(1):215. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 9. Sun Z, Wang W, Huang W, et al. Noninvasive imaging evaluation of peritoneal recurrence and chemotherapy benefit in gastric cancer after gastrectomy: a multicenter study. Int J Surg. 2023;109(7):2010–24. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 10. Zou T, Chen P, Wang T, et al. A cascaded segmentation-classification deep learning framework for preoperative prediction of occult peritoneal metastasis and early recurrence in advanced gastric cancer. Acad Radiol. 2025;S1076–6332(25):00743–3. [pii]. [ DOI ] [ PubMed ] [ Google Scholar ] 11. Hu W, Lin L, Ma C, et al. CT-based quantification of spatiotemporal heterogeneity for predicting response to neoadjuvant chemotherapy in locally advanced gastric cancer. BMC Med Imaging. 2025;25(1):209. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 12. Hsu JT, Lin YN, Chen YF, et al. A comprehensive overview of gastric cancer management from a surgical point of view. Biomed J. 2025;48(4):100817. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 13. Zheng XH, Zhang W, Yang L, et al. Role of D2 gastrectomy in gastric cancer with clinical para-aortic lymph node metastasis. World J Gastroenterol. 2019;25(19):2338–53. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 14. Kim SJ, Kim HH, Kim YH, et al. Peritoneal metastasis: detection with 16- or 64-detector row CT in patients undergoing surgery for gastric cancer. Radiology. 2009;253(2):407–15. [ DOI ] [ PubMed ] [ Google Scholar ] 15. Li Z, Li Z, Jia S, et al. Depth of tumor invasion and tumor-occupied portions of stomach are predictive factors of intra-abdominal metastasis. Chin J Cancer Res. 2017;29(2):109–17. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 16. Provenzano L, Gwee YX, Conca V, et al. Unveiling the prognostic significance of malignant ascites in advanced gastrointestinal cancers: a marker of peritoneal carcinomatosis burden. Ther Adv Med Oncol. 2024;16:17588359241289517. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 17. Chen S, Zhang H, Wei H, Tong Y, Chen X. Practical nomogram based on comprehensive CT texture analysis to preoperatively predict peritoneal occult metastasis of gastric cancer patients. Front Oncol. 2022;12:882584. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 18. Huang B, Sun Z, Wang Z, et al. Factors associated with peritoneal metastasis in non-serosa-invasive gastric cancer: a retrospective study of a prospectively-collected database. BMC Cancer. 2013;13:57. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 19. Huang J, Luo H, Zhou C, et al. Yield of staging laparoscopy for incurable factors in Chinese patients with advanced gastric cancer. J Laparoendosc Adv Surg Tech A. 2018;28(1):19–24. [ DOI ] [ PubMed ] [ Google Scholar ] 20. Huang J, Chen Y, Zhang Y, et al. Comparison of clinical-computed tomography model with 2D and 3D radiomics models to predict occult peritoneal metastases in advanced gastric cancer. Abdom Radiol (NY). 2022;47(1):66–75. [ DOI ] [ PubMed ] [ Google Scholar ] 21. Sun W, Jiang Z, Deng Z. Comprehensive analysis of the tumor immune microenvironment in gastric cancer and peritoneal metastasis based on single-cell RNA sequencing analysis. Sci Rep. 2025;15(1):32090. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 22. Zhang Z, Wang M, Gao X, Tai G, Li F, Wang Z. Gastric cancer-derived exosomal miR-196a-5p remodels the pre-metastatic niche for peritoneal dissemination. Cell Signal. 2025;135:112039. [ DOI ] [ PubMed ] [ Google Scholar ] 23. Di Gregorio F, Polici M, Pilozzi E, et al. Staging of gastric cancer: CT patterns and correlation with pathologic findings. Radiographics. 2025;45(8):e240186. [ DOI ] [ PubMed ] [ Google Scholar ] 24. Mu XD, Ji DX, Kang DQ. Application value of multiphase contrast-enhanced computed tomography radiomics in preoperative evaluation of peritoneal metastasis in gastric cancer. World J Gastrointest Oncol. 2026;18(2):115404. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 25. Zhu ZN, Feng QX, Li Q, Xu WY, Liu XS. Machine learning-based CT radiomics approach for predicting occult peritoneal metastasis in advanced gastric cancer preoperatively. Clin Radiol. 2025;80:106727. [ DOI ] [ PubMed ] [ Google Scholar ] 26. Chen H, Chen Y, Dong Y, et al. 18F-FDG PET/CT radiomics-based multimodality fusion model for preoperative individualized noninvasive prediction of peritoneal metastasis in advanced gastric cancer. Ann Surg Oncol. 2024;31(9):6017–27. [ DOI ] [ PubMed ] [ Google Scholar ] 27. Komori M, Asayama Y, Fujita N, et al. Extent of arterial tumor enhancement measured with preoperative MDCT gastrography is a prognostic factor in advanced gastric cancer after curative resection. AJR Am J Roentgenol. 2013;201(2):W253–61. [ DOI ] [ PubMed ] [ Google Scholar ] 28. Liu S, Liu S, Ji C, et al. Application of CT texture analysis in predicting histopathological characteristics of gastric cancers. Eur Radiol. 2017;27:4951–9. [ DOI ] [ PubMed ] [ Google Scholar ] 29. Bai L, Ni B, Shen X, et al. Effectiveness of circulating tumor cells and circulating tumor DNA in peritoneal lavage fluid for predicting metachronous peritoneal metastasis of gastric cancer. Transl Res. 2025;283:13–21. [ DOI ] [ PubMed ] [ Google Scholar ] 30. Diamantis A, Samara AA, Lafioniatis A, Janho MB, Floros T, Tepetes K. The role of prophylactic HIPEC in high-risk gastric cancer patients: where do we stand. Cancers (Basel). 2025;17(15):2492. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 31. Guo F, Wang D, Zhang L, et al. Efficacy and safety of anti-PD-1 antibodies plus small molecule anti-angiogenic drugs and chemotherapy in gastric cancer peritoneal metastasis: a multicenter real-world study. BMC Cancer. 2025;25(1):1028. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 32. Huang X, Chen X, Chen J, et al. Immune checkpoint inhibitors combined with chemotherapy in conversion therapy for stage IV gastric cancer: a multicenter retrospective cohort study. Cancer Immunol Immunother. 2025;74(10):302. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 33. Restle D, Amador-Molina A, Misawa K, Banerjee S, Ku G, Adusumilli PS. Intraperitoneal CAR T-cell therapy for peritoneal carcinomatosis from gastroesophageal cancer: preclinical investigations to a phase I clinical trial ( NCT06623396 ). J Immunother Cancer. 2025;13(9):e012292. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Data Availability Statement The data used to support the findings of this study are available from the corresponding author upon reasonable request. The data that support the findings of this study are available from the Fuzhou University Affiliated Provincial Hospital, but restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available. Data are however available from the corresponding author upon reasonable request and with permission of the Fuzhou University Affiliated Provincial Hospital. Articles from BMC Gastroenterology are provided here courtesy of BMC ACTIONS View on publisher site PDF (1.4 MB) Cite Collections Permalink RESOURCES Similar articles Cited by other articles Links to NCBI Databases Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top

Related documents

• Preliminary Study on the Diagnostic Value Comparison of 18F-FAPI-04 PET/CT and PET/MR in Gastric Cancer#842920
• Application of artificial intelligence based on contrast-enhanced CT imaging for predicting peritoneal metastasis in patients with T3/T4 stage gastric cancer.#189943
• The Role of Pre-operative and Post-operative Circulating Tumor Cells in Gastric Cancer.#868941
• GutCore: An Endoscopy Foundation Model for Whole-Case Gastric Cancer Analysis#459880
• Spectral CT-based Nomogram for Noninvasive Assessment and Prognostic Prediction of MSI/dMMR in Esophagogastric Junction Adenocarcinoma.#1147253