Contrast-enhanced CT-based radiomics nomogram for the preoperative prediction of perineural invasion in colorectal cancer: a two-center study

Contrast-enhanced CT-based radiomics nomogram for the preoperative prediction of perineural invasion in colorectal cancer: a two-center study | 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 Article Contrast-enhanced CT-based radiomics nomogram for the preoperative prediction of perineural invasion in colorectal cancer: a two-center study Juan Zhang, Yafei Wang, Fajuan Shen, Song Tian, Lingling Feng, and 4 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6970474/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Objectives/Purpose To develop and validate a radiomics nomogram for preoperative prediction of perineural invasion (PNI) in colorectal cancer (CRC) using multi-center contrast-enhanced CT datasets. Methods This retrospective study enrolled 334 CRC patients from two hospitals, divided into the training, internal validation, and external validation sets. Radiomics features were extracted and selected by the least absolute shrinkage and selection operator (LASSO) logistic regression to build a radiomics model. Clinical independent risk factors were used to construct a clinical model. An ensemble model was established by integrating the Rad-score with clinical predictors. Model performance was evaluated using ROC curves and decision curve analysis (DCA). Results The radiomics model was built from whole-tumor radiomics features, while the clinical model incorporated CEA, CT-reported T stage, and lymph node status. The ensemble model achieved the best performance in the internal validation set, whereas the radiomics model demonstrated greater stability in the external validation set. A nomogram combining the Rad-score and clinical predictors was developed for preoperative PNI prediction. Conclusions The contrast-enhanced CT-based radiomics nomogram is a promising non-invasive tool for preoperative PNI assessment in CRC patients, showing robust performance across multi-center datasets. Biological sciences/Cancer Health sciences/Gastroenterology Computed tomography Perineural invasion Radiomics Colorectal cancer Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Introduction Colorectal cancer (CRC) is the third most widespread cancer worldwide and has an increasing morbidity and mortality, posing a significant threat to public health [ 1 ]. In the case of early- and mid-stage CRC patients, standardized radical surgery is still regarded as the preferred treatment approach. However, postoperative recurrence and metastasis are the major factors contributing to unfavorable patient outcomes. Apart from the commonly-known pathways of direct extension, lymphatic metastasis, and hematogenous metastasis, perineural invasion (PNI) has been increasingly recognized as a possible pathway of tumor spread [ 2 ]. As a pathological variable associated with poor prognosis [ 3 ], perineural invasion (PNI) is defined as the biological process in which cancer cells invade the nerves and spread along the nerve sheaths. Growing evidence suggests that PNI has a substantial impact on the long-term survival and prognosis of CRC patients [ 4 , 5 ]. Several studies have demonstrated that adjuvant therapy could offer potential benefits for PNI-positive patients. Therefore, the PNI status should be considered when assessing the need for adjuvant therapy in rectal cancer patients [ 6 ]. Other studies have shown that neoadjuvant therapy could reduce the PNI-positive rate, and postoperative chemotherapy led to a significant improvement in the overall survival (OS) among PNI-positive patients [ 7 , 8 ]. Nevertheless, the status of PNI can only be identified by pathological evaluation of surgical specimens. Specifically, biopsy and imaging examinations provide limited accuracy in determining the PNI status of CRC patients [ 7 ]. Postoperative pathology is capable of accurately determining the PNI status but it cannot be used to guide preoperative treatment strategies. Notably, accurate prediction of the PNI status for CRC patients before treatment may enable more personalized clinical decision-making. Contrast-enhanced CT is of great significance in localizing colorectal cancer, formulating treatment plans, and monitoring longitudinal responses; in contrast to preoperative biopsy and serum tests, contrast-enhanced CT is a noninvasive examination [ 9 , 10 ]. A growing number of studies have employed radiomics to monitor tumor phenotypes in medical images [ 11 – 13 ]. Several articles have assessed the effectiveness of MRI-based radiomics to predict PNI in RC [ 7 , 14 , 15 ]. However, the sample sizes of the prior studies were rather limited and lacked external validation. Currently, CT-based radiomics studies for PNI prediction in colon cancer are still lacking. Therefore, this study aimed to develop and validate a nomogram integrating CT-based radiomics features with clinical factors, facilitating the preoperative prediction of PNI in a larger group of CRC patients. Materials and methods Patients The ethics committee of the Second Affiliated Hospital of Shandong First Medical University approved our study and all methods were performed in accordance with the relevant guidelines and regulations. Owing to the retrospective design of the study, the need to obtain informed consent was waived by the ethics committee of the Second Affiliated Hospital of Shandong First Medical University. A total of 334 patients who underwent surgical resection for CRC in two hospitals from July 2020 to June 2023 were enrolled in this retrospective study. The following were the inclusion criteria: (1) CRC patients confirmed through pathological examination; (2) contrast-enhanced CT was carried out within 2 weeks before surgery; (3) pathological reports for definite PNI status. The exclusion criteria were: (1) CT images of low quality; (2) patients with malignancies other than CRC; (3) any anti-tumor treatment was given to the patient before surgery. The patient recruitment pathway is shown in Fig. 1 . Finally, 165 CRC patients were recruited from Hospital 1 and were randomly divided into the training set (n = 132) and the internal validation set (n = 33) at a ratio of 8:2 (Fig. 1 ). Following the same inclusion and exclusion criteria, 169 CRC patients were recruited from Hospital 2 and were assigned to the external validation set (n = 169). The medical records were reviewed to acquire the clinical and pathological data of every patient. The baseline data included the age, gender, family history, history of alcoholism, carcinoembryonic antigen (CEA), location, T stage (cT stage) and lymph node status (cLN status) reported by CT, histological grade, pathological T stage (pT) and N stage (pN), lymphovascular invasion (LVI), and tumor deposits, as shown in Table 1 . Table 1 Clinical characteristics of the study population Characteristics PNI+ (n = 131) PNI- (n = 203) P value Cross - validation set (n = 165) testing set (n = 169) Age (Mean ± SD,years) 64.22 ± 11.33 62.38 ± 10.30 0.126 60.66 ± 10.41 65.49 ± 10.54 Gender (Male/Female) 82/49 126/77 0.923 96/69 112/57 Family history(Yes/No) 11/120 18/185 0.882 24/141 5/164 History of alcoholism(Yes/No) 27/104 60/143 0.069 27/138 60/109 CEA(+/-) (positive ≥ 5 ng/ml) 76/55 77/126 < 0.001 64/101 89/80 Location (Rectum/Left Hemicolon/Right Hemicolon) 71/33/27 98/48/57 0.302 78/50/37 91/31/47 cT stage(T1-2/T3/T4) 2/95/34 30/129/44 < 0.001 13/110/42 19/114/36 cLN status(+/-) 88/43 94/109 < 0.001 92/73 90/79 Histological grade (Moderate-high/Poor) 94/37 148/55 0.818 99/66 143/26 pT stage (T1/T2/T3/T4) 0/2/95/34 3/28/141/31 < 0.001 2/11/117/35 1/19/119/30 pN stage (N0/N1/N2) 48/48/35 119/52/32 < 0.001 80/49/36 87/51/31 LVI(+/-) 65/66 33/170 < 0.001 45/120 53/116 Tumor deposits(+/-) 50/81 33/170 < 0.001 39/126 44/125 Abbreviations: CEA, carcinoembryonic antigen; cT stage, CT-reported T stage; cLN status, CT-reported lymph node status; pT stage, pathological T stage; pN stage, pathological N stage; LVI, lymphovascular invasion. Reference standard for pathology Histopathology results of all CRC patients were obtained from electronic medical records. According to the 8th Edition American Joint Committee on Cancer (AJCC) staging system [ 16 ], all the pathological features were assessed by using the hematoxylin and eosin staining of the resected specimens. When cancer cells invaded any layer of the nerve sheath or surrounded more than 33% of the nerve circumference, PNI status was considered positive; otherwise, PNI status was considered negative [ 7 ]. Since a systematic review showed hardly any difference between studies relying on PNI data from pathology reports and those with dedicated review for PNI [ 17 ], re - examination of the pathological slides of all enrolled patients was not conducted. Imaging acquisition and evaluation For all the recruited patients, abdomen-pelvis enhanced CT was performed to find the primary tumor and metastatic lesions. The scanning range extended from the top of the diaphragm down to the symphysis pubis. The arterial and portal phase images were captured at 25 seconds and 60 seconds after the iodinated contrast agent was injected. Considering that the portal phase provided the most optimal visualization of CRC, portal phase images were chosen for target delineation. Supplementary Table S1 comprehensively details the CT scanning parameters and contrast agents employed by the two centers. The picture archiving and communication system provided CT images in the digital imaging and communications in medicine (DICOM) format. Two radiologists were assigned to review the CT images; reader 1 and reader 2 had 5 years and 10 years of experience in abdominal imaging, respectively. Blinded to all clinical and pathological information, the radiologists assessed the location of the primary tumor and TNM staging in accordance with the 8th AJCC staging system. The T stage (cT stage) and lymph node status (cLN status) were determined from the CT images by integrating the results of the two radiologists (Table 1 ). In this study, cT1 and cT2 tumors were grouped as cT1-2 due to the limitations of contrast-enhanced CT in differentiating T1 from T2 in CRC patients. Regional lymph node diameter more than 10 mm or at least 3 clustered lymph nodes were defined as cLN status positive. The weighted kappa statistics test was applied to assess the inter-observer variability regarding cT/cN. During the assessment of CT images, any disagreements between the radiologists were resolved through joint discussion. Imaging segmentation The open-source software ITK-SNAP version 3.8.0 ( http://www.itksnap.org ) was employed to import the portal-phase CT images to delineate the region of interest (ROI). The 3D ROI was defined as the full-volume area covering the entire tumor [26]. The two radiologists manually delineated each 3D ROI along the area of abnormal enhancement on each primary tumor slice. During the delineation process, the radiologists made efforts to exclude intestinal gas, intestinal juice, and feces from the ROIs. Radiomics feature extraction and selection To ensure the consistency in original image thickness between the training and testing sets, all scans were resampled using the Simple ITK BSpline interpolation method. Feature extraction was performed using IntelliSpace Medicina Scientia (ISMS) v2.7.0 (Philips Healthcare, China), adhering to the Imaging Biomarker Standardization Initiative (IBSI) for most features. To capture high-order variants of image characteristics, high- and low-frequency wavelet, exponential, log-sigma, square, square root, gradient, and logarithm transformations were applied. The extracted features in this study included shape, first-order (mean, median, maximum, skewness), GLCM (gray-level co-occurrence matrix), GLRLM (gray-level run length matrix), GLSZM (gray-level size zone matrix), GLDM (gray-level dependence matrix), and NGTDM (neighborhood gray-tone difference matrix). Subsequently, all features were standardized based on the average value and standard deviation of the training set. Feature selection operations were conducted on the training set. Firstly, 30 cases were randomly selected from the training set to evaluate the observer repeatability of the features. During the same period, the two radiologists separately outlined the 3D ROIs for 30 cases to assess the inter-observer repeatability. One month later, reader 1 delineated the 3D ROIs for the same 30 cases again to assess the intra-observer repeatability. The inter-observer and intra-observer correlation coefficients (ICC) were calculated to assess the agreement. Features showing inter- and intra-observer ICCs both exceeding 0.75 were considered to have satisfactory reliability and reproducibility and were included in the subsequent analysis. Secondly, a Pearson correlation coefficient threshold of 0.85 was utilized to remove redundant features. Finally, the least absolute shrinkage and selection operator (LASSO) was employed to recognize the most relevant features for efficient model training. Model building and evaluation The model training process employed a fully automated pipeline. Nine candidate classifiers were evaluated, including nonlinear support vector machine (SVM), naive Bayes, logistic regression (LR), k-nearest neighbors (KNN), multi-layer perceptron (MLP), random forest, decision tree, ADABoost, and XGBoost. For each classifier, a hyperparameter search space was defined, followed by a grid search on the training set to determine the optimal hyperparameter configuration. The detailed search space is displayed in Supplementary Table S2. Thereafter, a five-fold cross-validation scheme was implemented, and the model exhibiting the maximum validation AUC score was selected as the final model. The training process used Python code and the scikit-learn library (version 1.1.2). The receiver operating characteristic (ROC) curves and precision-recall (PR) curves were constructed to assess the diagnostic capabilities of the models. F1-scores were calculated to measure binary classification performance, defined as the harmonic mean of positive predictive value (PPV) and sensitivity. Pearson correlation was performed using Scipy (version 1.7.1). Model performance was compared using the area under the curve (AUC) of ROC curves and average precision (AP) of PR curves. The value of the model in clinical practice was tested by conducting a decision curve analysis (DCA). Figure 2 shows the workflow of this study for constructing and validating the radiomics models. Statistical analysis The SPSS software (version 27.0; IBM, New York, USA) was used to perform statistical analyses. Continuous variables conforming to a normal distribution were expressed in the form of mean ± standard deviation (SD) and compared using an independent sample t-test. Categorical variables were expressed as raw numbers and compared using either the chi-square (X 2 ) test or Fisher's exact test. AUCs of the models were compared by DeLong’s test. In this study, a P-value of less than 0.05 was considered statistically significant. Results Patient characteristics This study included a total of 334 CRC patients(131 PNI + and 203 PNI-) who underwent contrast-enhanced CT in two hospitals from July 2020 to June 2023. Significant differences in clinical characteristics were observed between the PNI + and PNI - groups, including CEA, cT stage, cLN status, pT stage, pN stage, LVI, and tumor deposits (p 0.05). The patients were divided into three subsets: the training set (n = 132), the internal validation set (n = 33), and the external testing set (n = 169). Table 1 compares the clinical characteristics of patients in the three cohorts. Feature selection and model building A total of 1473 features, including 16 shape features, 288 first-order features, and 1169 textural features, presented good reliability and reproducibility, with the inter- and intra-observer ICCs both exceeding 0.75. After deleting redundant features, the 25 most significant radiomics features were selected by analysis of variance (ANOVA) and LASSO LR, comprising 6 first-order features and 19 textural features (Figure S1 ). The Rad-score was constructed through logistic regression and was identified as an independent predictor significantly related to the PNI status in CRC patients (p < 0.001). Subsequently, the radiomics model was built upon the Rad-score through linear combination of the selected features and their corresponding LASSO coefficients. Through univariate and multivariate analysis, CEA, cT stage, and cLN status were screened as independent risk factors for predicting PNI (Odds ratio [OR] = 2.26, 10.99, and 1.96; P = 0.01, 0.008, and 0.033, respectively), as shown in Table 2 . Therefore, a clinical model based on CEA, cT stage, and cLN status was constructed. To build a more robust prediction model, an ensemble model was established by integrating Rad-score with three clinical independent predictors. Table 2 Independent predictors of PNI status in CRC Variables β Nomogram Standard error Wald OR (95% CI) p CEA 0.817 0.316 6.67 2.26 0.01 (1.23–4.25) cTstage(T3) 2.522 0.855 8.7 12.45 0.003 (2.84–91.94) cTstage(T4) 2.397 0.899 7.11 10.99 0.008 (2.25–86.49) cLNstatus 0.671 0.314 4.57 1.96 0.033 (1.06–3.65) Rad-Score 1.834 0.211 75.53 6.26 <0.001 (4.24–9.73) Abbreviations: CEA, carcinoembryonic antigen; cT stage, CT-reported T stage; cLN status, CT-reported lymph node status; β, the regression coefficient; OR, odds ratio; CI, confidence interval Model comparison and validation In this study, five-fold leave-group-out cross-validation (LGOCV) analysis was applied for the purpose of comparing the robustness and reliability of the radiomics models. The radiomics model built upon the Rad-score demonstrated satisfactory discriminative ability, with an AUC of 0.847 (95% confidence interval [CI]: 0.690–1.000) in the internal validation set and 0.921 (95% CI: 0.877–0.966) in the external testing set (Table 3 ). The clinical model achieved AUCs of 0.713 (95%CI: 0.516–0.910) in the internal validation set and 0.597 (95%CI: 0.508–0.685) in the external testing set. Combining the Rad-score with the clinical factors led to improved performance. As presented in Table 3 and Fig. 3 , the AUCs of the ensemble model were 0.864 (95% CI: 0.714–1.000) in the internal validation set and 0.907 (95% CI: 0.858–0.955) in the external testing set, were significantly greater than those of the clinical model (P < 0.05). However, no notable distinction was found when comparing the ensemble model and radiomics model in the internal validation set (AUC = 0.864 vs 0.847, P = 0.466) and external validation set (AUC = 0.907 vs 0.921, P = 0.648). The diagnostic performance of the aforementioned models was also compared using the AP of PR curves, as shown in Figure S2. The DCA indicated that the radiomics model had a higher overall net benefit than that of the clinical model and the ensemble model in predicting PNI, as shown in Fig. 4 . The calibration curve of the radiomics model is shown in Figure S3, with the actual observed probability of PNI represented on the y-axis, and the predicted probability of PNI shown on the x-axis. The general trend of the predicted risk was demonstrated by the locally weighted regression line (solid line) of calibration plots. In this study, the radiomics model demonstrated a favorable consistency between the observed and predicted probabilities. Notably, the solid lines in the internal validation set (the green solid line) and the external testing set (the orange solid line) were in close proximity to the reference line (dotted line). The result of the Hosmer-Lemeshow test was in accordance with this conclusion. Table 3 Comparisons of models in the internal training and external testing sets. Subsets Diagnostic Index Models Clinical Model Radiomics Model Ensemble Model Internal Validation set AUC 0.713 0.847 0.864 (95%CI) (0.516–0.910) (0.690-1.000) (0.714-1.000) Accuracy(%) 0.788 0.788 0.879 Sensitivity(%) 0.727 1.000 0.818 Specificity(%) 0.818 0.682 0.909 External Testing set AUC 0.597 0.921 0.907 (95%CI) (0.508–0.685) (0.877–0.966) (0.858–0.955) Accuracy(%) 0.615 0.845 0.763 Sensitivity(%) 0.641 0.883 0.949 Specificity(%) 0.590 0.808 0.577 Abbreviations: AUC, area under the curve; CI, confidence interval; CI, confidence interval Feature importance and visualization In order to explain and visualize the features that were important for the model, a hierarchically-clustered heatmap was plotted (Fig. 5 ) for the radiomics model, illustrating the 25 retained radiomics features along with their corresponding coefficients following dimensionality reduction through LASSO regression analysis. In addition, the SHapley Additive exPlanation (SHAP) summary chart was plotted for the ensemble model, which indicated the direction and degree of the influence of each feature on the model, as shown in Fig. 6 . Nomogram construction We established the nomogram through the training set and then validated its effectiveness in the internal validation and external testing sets. The nomogram was created to visualize the ensemble model (Fig. 7 ). A greater total score on the nomogram indicates an increased predicted probability of PNI. Discussion The objective of this research was to construct and validate a contrast-enhanced CT-based radiomics nomogram from multicenter datasets for the preoperative prediction of the PNI status in CRC patients. The main findings are summarized as follows: (1) The discriminative capacity of the radiomics model built upon the Rad-score was proved to be highly satisfactory in both the internal validation and external testing sets; (2) CT-reported T (cT) stage, lymph node (cLN) status, and the level of carcinoembryonic antigen (CEA) were clinical predictors for PNI status, which were utilized for building the clinical model; (3) The diagnostic performance of the ensemble model constructed by integrating Rad-score with cT stage, cLN status, and CEA was better than that of the clinical model, but was not significantly different from that of the radiomics model. Recent studies have indicated that PNI is not merely the spread of cancer cells along the connective tissues enveloping the nerve sheath. Instead, it is a complex biological process where various neurotrophic factors and chemokines interact with cancer cells and the surrounding microenvironment [ 18 ]. PNI is associated with cancer invasion, local recurrence, and metastasis, leading to unfavorable clinical outcomes. Growing evidence shows that PNI is an independent risk factor for poor prognosis of colorectal cancer. Therefore, accurately predicting the PNI status of CRC patients contributes to the development of an individualized therapeutic plan and prognosis evaluation [ 19 ]. In comparison with CT and MRI, radiomics can yield favorable results for predicting PNI. This is accomplished by extracting high-throughput features that can capture subtle tissue differences undetectable by the naked eye. Different MR-based radiomics models have been documented for predicting the PNI status in RC [ 14 , 15 , 18 ]. These studies enrolled patients with rectal cancer from single institutions, thereby limiting the predictive value and generalizability of the results. Recently, some multicenter studies [ 20 , 21 ] have reported promising efficacy using machine learning or nomograms based on CT radiomics to predict PNI in colorectal cancer patients. However, the sample sizes of the previous studies were relatively small. In this study, 334 patients were enrolled from two hospitals, and the internal and external validations were performed for the established models, which enhanced the generalization capability of the model and the reliability of the conclusions. In this study, multivariable analysis results revealed that the cT stage and cLN status were independent clinical factors for predicting PNI in CRC, indicating that the tumor's invasive depth and lymph node metastasis were closely related to PNI. With the progression of the cT stage and cLN status, the tumor shows higher aggressiveness and malignancy, increasing the risk of PNI for CRC patients. These findings are in accordance with the observations of previous studies [ 17 , 22 ]. Similar to the results of Que Y et al. [ 23 ], the present study identified CEA as another independent predictor for PNI, highlighting its significance as a serum tumor marker. CEA may indirectly enhance tumor invasiveness by promoting the interaction between tumor cells and the neural microenvironment. Nevertheless, the findings of this research indicated that the clinical models based on cT stage, cLN status, and CEA showed moderate predictive performance in the internal validation set, possibly because cT stage and cLN status were easily influenced by the subjectivity and experience between observers, leading to limited predictive value [ 20 ]. In our study, the Pearson correlation coefficient and LASSO algorithm were utilized to remove the most redundant features and select the most relevant features. Five-fold cross-validation was performed to ensure the stability of the selected radiomics features. Thereafter, a collection of selected radiomics features was weighted by the regression coefficients to construct the Rad-score [ 14 ]. Moreover, the radiomics model based on Rad-score exhibited high discrimination capability, with an AUC of 0.847 in the internal validation and 0.921 in the external testing sets. The radiomics features selected for building the radiomics model included first-order statistical features and texture features. Among them, texture features accounted for the highest proportion as they reflected the tissue heterogeneity by quantifying the spatial distribution pattern of gray intensity in the image. To improve the predictive performance, Rad-score was combined with independent clinical predictors to construct the ensemble model. The diagnostic performance of the ensemble model was much better than that of the clinical model, but was not obviously different from that of the radiomics model. This finding indicates that Rad-score contributed the most to the model construction process. The radiomics model based on Rad-score alone yielded similar predictive value to the ensemble model. To facilitate personalized medical treatment, the ensemble model was converted into a nomogram, predicting an individual risk probability of PNI. As evidenced by its longer axis in the nomogram, Rad-score held greater significance compared to independent clinical predictors. This user-friendly system assists physicians in conducting preoperative assessments of PNI status and customizing clinical management strategies [ 14 ]. Nevertheless, the limitations of the present study should be acknowledged. First, potential selection bias may be present due to the retrospective study design. Second, the manual delineation of the ROIs was performed by skilled radiologists, but it remains a time-consuming process. In the future, the image segmentation process should be simplified. Deep learning, with its potential for automatic segmentation, offers a viable solution for this purpose. Conclusion The radiomics nomogram integrating Rad-score and clinical predictors can help to evaluate the PNI status before surgery and guide personalized treatment of CRC patients. Declarations Acknowledgements We thank Home for Researchers editorial team (www.home-for-researchers.com) for language editing service. Author contributions Conception/design: Chao Su. Provision of study material or patients: Fajuan Shen, Lingling Feng, Qianqian Xia. Collection and/or assembly of data: Yafei Wang, Peijie Shen, Song Tian. Data analysis and interpretation: Yafei Wang, Peijie Shen, Song Tian. Manuscript writing: Juan Zhang, Chao Su. Final approval of manuscript: All authors. Data availability statement The data underlying this article will be shared on reasonable request to the corresponding author. Competing interests statement The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Ethics declarations and consent statement The ethics committee of the Second Affiliated Hospital of Shandong First Medical University has approved our study. The need to obtain informed consent was waived by the ethics committee of the Second Affiliated Hospital of Shandong First Medical University due to the retrospective nature of the study. Consent to publish All authors have agreed to publish this manuscript. Funding The authors declare that no funds, grants, or other support were received during the preparation of this manuscript. References Siegel, R. L. et al. Cancer statistics, 2021. CA Cancer J. Clin. 71 , 7–33. https://doi.org/10.3322/caac.21654 (2021). Kim, C. H. et al. Prognostic impact of perineural invasion in rectal cancer after neoadjuvant chemoradiotherapy. World J. Surg. 43 , 260–272. https://doi.org/10.1007/s00268-018-4774-8 (2019). Dhadda, A. S. et al. Mandard tumour regression grade, perineural invasion, circumferential resection margin and post-chemoradiation nodal status strongly predict outcome in locally advanced rectal cancer treated with preoperative chemoradiotherapy. Clin. Oncol. 26 , 197–202. https://doi.org/10.1016/j.clon.2014.01.001 (2014). Kim, Y. I. et al. Clinical implication of perineural and lymphovascular invasion in rectal cancer patients who underwent surgery after preoperative chemoradiotherapy. Dis. Colon Rectum . 65 , 1325–1334. https://doi.org/10.1097/DCR.0000000000002219 (2022). Zhang, B. et al. Combining perineural invasion with staging improve the prognostic accuracy in colorectal cancer: a retrospective cohort study. BMC Cancer . 23 , 675. https://doi.org/10.1186/s12885-023-11114-8 (2023). Song, J. H. et al. Significance of perineural and lymphovascular invasion in locally advanced rectal cancer treated by preoperative chemoradiotherapy and radical surgery: Can perineural invasion be an indication of adjuvant chemotherapy? Radiother. Oncol 133 , 125–131. https://doi.org/10.1016/j.radonc.2019.01.002 (2019). Chen, T. et al. The incidence and prognosis value of perineural invasion in rectal carcinoma: From meta-analyses and real-world clinical pathological features. Clin. Oncol. 35 , e611–e621. https://doi.org/10.1016/j.clon.2023.05.008 (2023). Qin, L. et al. Perineural invasion affects prognosis of patients undergoing colorectal cancer surgery: a propensity score matching analysis. BMC Cancer . 23 , 452. https://doi.org/10.1186/s12885-023-10936-w (2023). García-Figueiras, R. et al. Advanced imaging techniques in evaluation of colorectal cancer. Radiographics 38 , 740–765. https://doi.org/10.1148/rg.2018170044 (2018). Glynne-Jones, R. et al. Rectal cancer: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann. Oncol. 28 , iv22–iv40. https://doi.org/10.1093/annonc/mdx224 (2017). Taghavi, M. et al. Machine learning-based analysis of CT radiomics model for prediction of colorectal metachronous liver metastases. Abdom. Radiol. 46 , 249–256. https://doi.org/10.1007/s00261-020-02624-1 (2021). Chen, J. et al. Pretreatment MR-based radiomics signature as potential imaging biomarker for assessing the expression of topoisomerase II alpha (TOPO-IIα) in rectal cancer. J. Magn. Reson. Imaging . 51 , 1881–1889. https://doi.org/10.1002/jmri.26972 (2020). Cui, Y. et al. Development and validation of a MRI-based radiomics signature for prediction of KRAS mutation in rectal cancer. Eur. Radiol. 30 , 1948–1958. https://doi.org/10.1007/s00330-019-06572-3 (2020). Guo, Y. et al. Preoperative prediction of perineural invasion with multi-modality radiomics in rectal cancer. Sci. Rep. 11 , 9429. https://doi.org/10.1038/s41598-021-88831-2 (2021). Yang, Y. S. et al. High resolution MRI-based radiomic nomogram in predicting perineural invasion in rectal cancer. Cancer Imaging . 21 , 40. https://doi.org/10.1186/s40644-021-00408-4 (2021). Weiser, M. R. AJCC 8th edition: Colorectal cancer. Ann. Surg. Oncol. 25 , 1454–1455. https://doi.org/10.1245/s10434-018-6462-1 (2018). Knijn, N. et al. Perineural invasion is a strong prognostic factor in colorectal cancer: A systematic review. Am. J. Surg. Pathol. 40 , 103–112. https://doi.org/10.1097/PAS.0000000000000518 (2016). Chen, J. et al. Pretreatment MR-based radiomics nomogram as potential imaging biomarker for individualized assessment of perineural invasion status in rectal cancer. Abdom. Radiol. 46 , 847–857. https://doi.org/10.1007/s00261-020-02710-4 (2021). Wang, H. et al. Perineural invasion in colorectal cancer: mechanisms of action and clinical relevance. Cell. Oncol. 47 (1), 1–17. https://doi.org/10.1007/s13402-023-00857-y (2024). Chen, Q. et al. Computed tomography-based radiomics nomogram for the preoperative prediction of perineural invasion in colorectal cancer: a multicentre study. Abdom. Radiol. 47 , 3251–3263. https://doi.org/10.1007/s00261-022-03620-3 (2022). Liu, N. J. et al. The value of machine learning based on CT radiomics in the preoperative identification of peripheral nerve invasion in colorectal cancer: a two-center study. Insights Imaging . 15 , 101. https://doi.org/10.1186/s13244-024-01664-1 (2024). Li, Y. et al. Preoperative prediction of perineural invasion and KRAS mutation in colon cancer using machine learning. J. Cancer Res. Clin. Oncol. 146 , 3165–3174. https://doi.org/10.1007/s00432-020-03354-z (2020). Que, Y. et al. A prediction nomogram for perineural invasion in colorectal cancer patients: a retrospective study. BMC Surg. 24 , 80. https://doi.org/10.1186/s12893-024-02364-9 (2024). Additional Declarations No competing interests reported. Supplementary Files SupplementaryMaterial.pdf Cite Share Download PDF Status: Posted Version 1 posted 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-6970474","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":539442448,"identity":"b7cfe880-31c6-4209-84b2-13446d45dbf2","order_by":0,"name":"Juan Zhang","email":"","orcid":"","institution":"The Second Affiliated Hospital of Shandong First Medical University","correspondingAuthor":false,"prefix":"","firstName":"Juan","middleName":"","lastName":"Zhang","suffix":""},{"id":539442450,"identity":"aa1bd21f-8612-4745-b0fd-f47f93508339","order_by":1,"name":"Yafei Wang","email":"","orcid":"","institution":"Shandong Cancer Hospital and Institute","correspondingAuthor":false,"prefix":"","firstName":"Yafei","middleName":"","lastName":"Wang","suffix":""},{"id":539442453,"identity":"42aab7d9-0a99-430f-8b8c-3c5770465920","order_by":2,"name":"Fajuan Shen","email":"","orcid":"","institution":"The Second Affiliated Hospital of Shandong First Medical University","correspondingAuthor":false,"prefix":"","firstName":"Fajuan","middleName":"","lastName":"Shen","suffix":""},{"id":539442455,"identity":"2e7cebf0-4abc-42d4-8edb-503769e1e890","order_by":3,"name":"Song Tian","email":"","orcid":"","institution":"Philips Healthcare","correspondingAuthor":false,"prefix":"","firstName":"Song","middleName":"","lastName":"Tian","suffix":""},{"id":539442456,"identity":"66c3454b-eca0-4614-a689-9f37d46e1e3d","order_by":4,"name":"Lingling Feng","email":"","orcid":"","institution":"The Second Affiliated Hospital of Shandong First Medical University","correspondingAuthor":false,"prefix":"","firstName":"Lingling","middleName":"","lastName":"Feng","suffix":""},{"id":539442457,"identity":"21b3ab7c-ae14-41ef-86ac-edd6584b0c37","order_by":5,"name":"Xin Chen","email":"","orcid":"","institution":"The Second Affiliated Hospital of Shandong First Medical University","correspondingAuthor":false,"prefix":"","firstName":"Xin","middleName":"","lastName":"Chen","suffix":""},{"id":539442460,"identity":"a8e33bde-e37d-40b9-bb8d-6ed12c0365d8","order_by":6,"name":"Qianqian Xia","email":"","orcid":"","institution":"The Second Affiliated Hospital of Shandong First Medical University","correspondingAuthor":false,"prefix":"","firstName":"Qianqian","middleName":"","lastName":"Xia","suffix":""},{"id":539442462,"identity":"344f910e-ac23-444c-bfc0-3f45c2df30f4","order_by":7,"name":"Peijie Shen","email":"","orcid":"","institution":"ShanDong Provincial TaiShan Hospital","correspondingAuthor":false,"prefix":"","firstName":"Peijie","middleName":"","lastName":"Shen","suffix":""},{"id":539442463,"identity":"584a86ce-f751-4926-939f-a1b518d0fb26","order_by":8,"name":"Chao Su","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAzklEQVRIiWNgGAWjYFACxoYDCQwHGBjYGxsffCBOC3PjgQ8gLTyHmw1nEKeFvfngDJAWifQ2aQ5iNMjPSGw4zFNzR55f8mGDNAODnZxuAwEtBjdAWo49M5w5O7HBuIAh2djsACEtEkAtvA2HGTfcTmxIBrowcRshLWCHAbXYb7h5EGgdMVoYgA47OLPhcOKGG4yNzURpMTjzsOHAh2OHk2f2JDYzzjAgwi/y7emPPyTUHLbtZz/+/MeHCjs5glrQLSVN+SgYBaNgFIwCHAAAPtlP7kxfMwQAAAAASUVORK5CYII=","orcid":"","institution":"The Second Affiliated Hospital of Shandong First Medical University","correspondingAuthor":true,"prefix":"","firstName":"Chao","middleName":"","lastName":"Su","suffix":""}],"badges":[],"createdAt":"2025-06-25 04:38:05","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6970474/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6970474/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":95119411,"identity":"96df7c4d-3ba2-4898-b876-4b3ad0c30bb6","added_by":"auto","created_at":"2025-11-04 13:41:54","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":48125,"visible":true,"origin":"","legend":"","description":"","filename":"Manuscriptrevised.docx","url":"https://assets-eu.researchsquare.com/files/rs-6970474/v1/0b8c495301d917e4e894dc75.docx"},{"id":95119415,"identity":"e9942611-a981-480b-868d-57d6530b0924","added_by":"auto","created_at":"2025-11-04 13:41:54","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1286731,"visible":true,"origin":"","legend":"","description":"","filename":"Figures.docx","url":"https://assets-eu.researchsquare.com/files/rs-6970474/v1/d4c06afe08673b2798098e7d.docx"},{"id":95119414,"identity":"0dda09d1-dbcf-4b8f-b384-9e44505869d9","added_by":"auto","created_at":"2025-11-04 13:41:54","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":20963,"visible":true,"origin":"","legend":"","description":"","filename":"Tables.docx","url":"https://assets-eu.researchsquare.com/files/rs-6970474/v1/1c139b04a0229fa6fdec7f5d.docx"},{"id":95119426,"identity":"e4ae27aa-de03-4149-a2d9-aed47a65b598","added_by":"auto","created_at":"2025-11-04 13:41:54","extension":"json","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9533,"visible":true,"origin":"","legend":"","description":"","filename":"a48d83efdb9941cfa6de5ea259b2a433.json","url":"https://assets-eu.researchsquare.com/files/rs-6970474/v1/67e9561382cc76e65de7d1b5.json"},{"id":95225759,"identity":"d401f4de-86aa-4091-958c-b0d654a5c7e9","added_by":"auto","created_at":"2025-11-05 16:25:29","extension":"pdf","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":301253,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterial.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6970474/v1/1b9325ba9ce03839f03e4424.pdf"},{"id":95119440,"identity":"1a86c523-1d5a-4619-b533-95b6acf76ba3","added_by":"auto","created_at":"2025-11-04 13:41:55","extension":"xml","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":101494,"visible":true,"origin":"","legend":"","description":"","filename":"a48d83efdb9941cfa6de5ea259b2a4331enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-6970474/v1/24288247990536447a671a73.xml"},{"id":95225361,"identity":"5f991584-2aca-46ee-a42c-ef7c5abe4ba9","added_by":"auto","created_at":"2025-11-05 16:24:54","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":97895,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-6970474/v1/e5237819c231baf767a108bc.png"},{"id":95119427,"identity":"349227b2-eea6-463d-8bd2-de5f664bda5e","added_by":"auto","created_at":"2025-11-04 13:41:54","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":302359,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-6970474/v1/ce3ac4ea59cffc23292ca5d8.png"},{"id":95119434,"identity":"ae7bc934-24b1-4584-be63-0aa4e4b64dd6","added_by":"auto","created_at":"2025-11-04 13:41:55","extension":"jpeg","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":412844,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-6970474/v1/3ce7b3fe14dc1676c80e59d4.jpeg"},{"id":95119438,"identity":"e58aebec-5d60-49bb-8313-bf75c288f388","added_by":"auto","created_at":"2025-11-04 13:41:55","extension":"jpeg","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":317148,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-6970474/v1/865388aa7f22f2a76c721fc5.jpeg"},{"id":95224642,"identity":"198019f2-65f3-4f30-af4d-fef4a2d57dec","added_by":"auto","created_at":"2025-11-05 16:24:05","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":215666,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-6970474/v1/05d34f02306c6eaf80cb1aa0.png"},{"id":95225931,"identity":"a729aac3-cec3-4468-bf05-9b50bac1eb8c","added_by":"auto","created_at":"2025-11-05 16:25:47","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":60799,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-6970474/v1/ea1614edbb27b6dc2267f042.png"},{"id":95119429,"identity":"9c4304ab-6a0a-4a21-9d13-2231d47ead6e","added_by":"auto","created_at":"2025-11-04 13:41:54","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":8035,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-6970474/v1/e2ede6fbbfd034f0f369ee92.png"},{"id":95119432,"identity":"e8ab1602-6171-4137-a6ea-df1480a9f003","added_by":"auto","created_at":"2025-11-04 13:41:55","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":39304,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-6970474/v1/7de4c01c91880b0544ce5a04.png"},{"id":95225372,"identity":"b2c3c85b-0cb3-418c-adb4-8638e4ae3e62","added_by":"auto","created_at":"2025-11-05 16:24:56","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":46399,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-6970474/v1/65f1d2203b473aad49c4a5b2.png"},{"id":95227136,"identity":"333df332-f8a4-48e1-b5d7-84a872301656","added_by":"auto","created_at":"2025-11-05 16:32:09","extension":"png","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":90383,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-6970474/v1/3784d8d3e50fbeb1a8b0f87c.png"},{"id":95119439,"identity":"71b52feb-064f-42e9-b3c7-ba5b14a65fd2","added_by":"auto","created_at":"2025-11-04 13:41:55","extension":"png","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":66901,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-6970474/v1/f1dc0f150933b57d417ca0d6.png"},{"id":95225433,"identity":"6487c6e8-0135-4a01-b94f-445285e250ca","added_by":"auto","created_at":"2025-11-05 16:25:05","extension":"png","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":57847,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-6970474/v1/466653c4848e2599cc04906a.png"},{"id":95119441,"identity":"2ea98555-c26a-4e67-a173-054c33fb1227","added_by":"auto","created_at":"2025-11-04 13:41:55","extension":"png","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":16416,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-6970474/v1/cd7d3c087e843cbf6c67f974.png"},{"id":95225470,"identity":"40e9e730-ec10-4986-ba34-a62fcba192ee","added_by":"auto","created_at":"2025-11-05 16:25:05","extension":"png","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6413,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-6970474/v1/cb400762bf0239dae779e689.png"},{"id":95119433,"identity":"b9bd66c9-119e-4929-a80a-e58851ccaf4d","added_by":"auto","created_at":"2025-11-04 13:41:55","extension":"xml","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":98682,"visible":true,"origin":"","legend":"","description":"","filename":"a48d83efdb9941cfa6de5ea259b2a4331structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-6970474/v1/e9e793f9ac87f9a0b5af6fe5.xml"},{"id":95119435,"identity":"c27e42fc-1054-44f7-8396-86528279a145","added_by":"auto","created_at":"2025-11-04 13:41:55","extension":"html","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":112571,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-6970474/v1/146ec930eb71d022a4c2b5d5.html"},{"id":95119412,"identity":"6f63c049-9ebc-4b9b-b903-45118756974f","added_by":"auto","created_at":"2025-11-04 13:41:54","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":100674,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"Figures321.png","url":"https://assets-eu.researchsquare.com/files/rs-6970474/v1/68525cc0ac269dc31cd83c07.png"},{"id":95225651,"identity":"36f2fa74-88a9-404c-b1c3-cdc84b7cdac6","added_by":"auto","created_at":"2025-11-05 16:25:21","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":424259,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"Figures322.png","url":"https://assets-eu.researchsquare.com/files/rs-6970474/v1/5d130b03132eaa245fd2136d.png"},{"id":95225664,"identity":"2e3858cd-43f6-42fc-a60a-c0d631733b48","added_by":"auto","created_at":"2025-11-05 16:25:23","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":316791,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"Figures323.png","url":"https://assets-eu.researchsquare.com/files/rs-6970474/v1/76f8c41eec678a985a4e694d.png"},{"id":95119419,"identity":"600b210e-26ad-4a56-be07-cf265a82f555","added_by":"auto","created_at":"2025-11-04 13:41:54","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":234309,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"Figures324.png","url":"https://assets-eu.researchsquare.com/files/rs-6970474/v1/339a55370569a443965b42d8.png"},{"id":95225761,"identity":"b4bae0b0-100d-4586-a40a-dc493b7a1953","added_by":"auto","created_at":"2025-11-05 16:25:29","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":303296,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"Figures325.png","url":"https://assets-eu.researchsquare.com/files/rs-6970474/v1/863ad0f7ad65a00ea8e1f7c5.png"},{"id":95226232,"identity":"dff87c46-d55c-452d-abe6-8fe4901fc0e9","added_by":"auto","created_at":"2025-11-05 16:30:44","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":90843,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"Figures326.png","url":"https://assets-eu.researchsquare.com/files/rs-6970474/v1/55a064ec6b4663242d0f3740.png"},{"id":95119422,"identity":"1cb642c6-8a31-44de-b716-75e3072cf6cd","added_by":"auto","created_at":"2025-11-04 13:41:54","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":55110,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"Figures327.png","url":"https://assets-eu.researchsquare.com/files/rs-6970474/v1/08768516b120f285c99385ac.png"},{"id":98628149,"identity":"06d85c0f-a14c-4984-89b9-dcc15e9e5d03","added_by":"auto","created_at":"2025-12-19 17:11:03","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2079728,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6970474/v1/5cb6ef84-de27-4459-912b-37e0af5d7664.pdf"},{"id":95224500,"identity":"01ba3ee3-a6a1-4d66-99c2-b1243ad82f7d","added_by":"auto","created_at":"2025-11-05 16:23:49","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":301253,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterial.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6970474/v1/3882ed977986565874b30196.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Contrast-enhanced CT-based radiomics nomogram for the preoperative prediction of perineural invasion in colorectal cancer: a two-center study","fulltext":[{"header":"Introduction","content":"\u003cp\u003eColorectal cancer (CRC) is the third most widespread cancer worldwide and has an increasing morbidity and mortality, posing a significant threat to public health [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. In the case of early- and mid-stage CRC patients, standardized radical surgery is still regarded as the preferred treatment approach. However, postoperative recurrence and metastasis are the major factors contributing to unfavorable patient outcomes.\u003c/p\u003e\u003cp\u003eApart from the commonly-known pathways of direct extension, lymphatic metastasis, and hematogenous metastasis, perineural invasion (PNI) has been increasingly recognized as a possible pathway of tumor spread [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. As a pathological variable associated with poor prognosis [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e], perineural invasion (PNI) is defined as the biological process in which cancer cells invade the nerves and spread along the nerve sheaths.\u003c/p\u003e\u003cp\u003eGrowing evidence suggests that PNI has a substantial impact on the long-term survival and prognosis of CRC patients [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. Several studies have demonstrated that adjuvant therapy could offer potential benefits for PNI-positive patients. Therefore, the PNI status should be considered when assessing the need for adjuvant therapy in rectal cancer patients [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. Other studies have shown that neoadjuvant therapy could reduce the PNI-positive rate, and postoperative chemotherapy led to a significant improvement in the overall survival (OS) among PNI-positive patients [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. Nevertheless, the status of PNI can only be identified by pathological evaluation of surgical specimens. Specifically, biopsy and imaging examinations provide limited accuracy in determining the PNI status of CRC patients [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. Postoperative pathology is capable of accurately determining the PNI status but it cannot be used to guide preoperative treatment strategies. Notably, accurate prediction of the PNI status for CRC patients before treatment may enable more personalized clinical decision-making.\u003c/p\u003e\u003cp\u003eContrast-enhanced CT is of great significance in localizing colorectal cancer, formulating treatment plans, and monitoring longitudinal responses; in contrast to preoperative biopsy and serum tests, contrast-enhanced CT is a noninvasive examination [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. A growing number of studies have employed radiomics to monitor tumor phenotypes in medical images [\u003cspan additionalcitationids=\"CR12\" citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. Several articles have assessed the effectiveness of MRI-based radiomics to predict PNI in RC [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. However, the sample sizes of the prior studies were rather limited and lacked external validation. Currently, CT-based radiomics studies for PNI prediction in colon cancer are still lacking. Therefore, this study aimed to develop and validate a nomogram integrating CT-based radiomics features with clinical factors, facilitating the preoperative prediction of PNI in a larger group of CRC patients.\u003c/p\u003e"},{"header":"Materials and methods","content":"\u003cp\u003e\u003cb\u003ePatients\u003c/b\u003e\u003c/p\u003e\u003cp\u003e The ethics committee of the Second Affiliated Hospital of Shandong First Medical University approved our study and all methods were performed in accordance with the relevant guidelines and regulations. Owing to the retrospective design of the study, the need to obtain informed consent was waived by the ethics committee of the Second Affiliated Hospital of Shandong First Medical University. A total of 334 patients who underwent surgical resection for CRC in two hospitals from July 2020 to June 2023 were enrolled in this retrospective study. The following were the inclusion criteria: (1) CRC patients confirmed through pathological examination; (2) contrast-enhanced CT was carried out within 2 weeks before surgery; (3) pathological reports for definite PNI status. The exclusion criteria were: (1) CT images of low quality; (2) patients with malignancies other than CRC; (3) any anti-tumor treatment was given to the patient before surgery. The patient recruitment pathway is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e\u003cp\u003eFinally, 165 CRC patients were recruited from Hospital 1 and were randomly divided into the training set (n\u0026thinsp;=\u0026thinsp;132) and the internal validation set (n\u0026thinsp;=\u0026thinsp;33) at a ratio of 8:2 (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Following the same inclusion and exclusion criteria, 169 CRC patients were recruited from Hospital 2 and were assigned to the external validation set (n\u0026thinsp;=\u0026thinsp;169). The medical records were reviewed to acquire the clinical and pathological data of every patient. The baseline data included the age, gender, family history, history of alcoholism, carcinoembryonic antigen (CEA), location, T stage (cT stage) and lymph node status (cLN status) reported by CT, histological grade, pathological T stage (pT) and N stage (pN), lymphovascular invasion (LVI), and tumor deposits, as shown in 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\u003eClinical characteristics of the study population\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"6\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" 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=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCharacteristics\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePNI+ (n\u0026thinsp;=\u0026thinsp;131)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003ePNI- (n\u0026thinsp;=\u0026thinsp;203)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eP value\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eCross - validation set\u003c/p\u003e\u003cp\u003e(n\u0026thinsp;=\u0026thinsp;165)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003etesting set\u003c/p\u003e\u003cp\u003e(n\u0026thinsp;=\u0026thinsp;169)\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAge (Mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD,years)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e64.22\u0026thinsp;\u0026plusmn;\u0026thinsp;11.33\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e62.38\u0026thinsp;\u0026plusmn;\u0026thinsp;10.30\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.126\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e60.66\u0026thinsp;\u0026plusmn;\u0026thinsp;10.41\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e65.49\u0026thinsp;\u0026plusmn;\u0026thinsp;10.54\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGender (Male/Female)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e82/49\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e126/77\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.923\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e96/69\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e112/57\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFamily history(Yes/No)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e11/120\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e18/185\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.882\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e24/141\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e5/164\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHistory of alcoholism(Yes/No)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e27/104\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e60/143\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.069\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e27/138\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e60/109\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCEA(+/-) (positive\u0026thinsp;\u0026ge;\u0026thinsp;5 ng/ml)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e76/55\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e77/126\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\u003cp\u003e64/101\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e89/80\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLocation (Rectum/Left Hemicolon/Right Hemicolon)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e71/33/27\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e98/48/57\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.302\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e78/50/37\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e91/31/47\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ecT stage(T1-2/T3/T4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2/95/34\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e30/129/44\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\u003cp\u003e13/110/42\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e19/114/36\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ecLN status(+/-)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e88/43\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e94/109\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\u003cp\u003e92/73\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e90/79\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHistological grade (Moderate-high/Poor)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e94/37\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e148/55\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.818\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e99/66\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e143/26\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003epT stage (T1/T2/T3/T4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0/2/95/34\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e3/28/141/31\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\u003cp\u003e2/11/117/35\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1/19/119/30\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003epN stage (N0/N1/N2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e48/48/35\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e119/52/32\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\u003cp\u003e80/49/36\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e87/51/31\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLVI(+/-)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e65/66\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e33/170\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\u003cp\u003e45/120\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e53/116\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTumor deposits(+/-)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e50/81\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e33/170\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\u003cp\u003e39/126\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e44/125\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003cp\u003eAbbreviations: CEA, carcinoembryonic antigen; cT stage, CT-reported T stage; cLN status, CT-reported lymph node status; pT stage, pathological T stage; pN stage, pathological N stage; LVI, lymphovascular invasion. \u003c/p\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eReference standard for pathology\u003c/b\u003e\u003c/p\u003e\u003cp\u003eHistopathology results of all CRC patients were obtained from electronic medical records. According to the 8th Edition American Joint Committee on Cancer (AJCC) staging system [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e], all the pathological features were assessed by using the hematoxylin and eosin staining of the resected specimens. When cancer cells invaded any layer of the nerve sheath or surrounded more than 33% of the nerve circumference, PNI status was considered positive; otherwise, PNI status was considered negative [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. Since a systematic review showed hardly any difference between studies relying on PNI data from pathology reports and those with dedicated review for PNI [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e], re - examination of the pathological slides of all enrolled patients was not conducted.\u003c/p\u003e\u003c/li\u003e\u003c/ul\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eImaging acquisition and evaluation\u003c/b\u003e\u003c/p\u003e\u003cp\u003eFor all the recruited patients, abdomen-pelvis enhanced CT was performed to find the primary tumor and metastatic lesions. The scanning range extended from the top of the diaphragm down to the symphysis pubis. The arterial and portal phase images were captured at 25 seconds and 60 seconds after the iodinated contrast agent was injected. Considering that the portal phase provided the most optimal visualization of CRC, portal phase images were chosen for target delineation. Supplementary Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e comprehensively details the CT scanning parameters and contrast agents employed by the two centers.\u003c/p\u003e\u003cp\u003eThe picture archiving and communication system provided CT images in the digital imaging and communications in medicine (DICOM) format. Two radiologists were assigned to review the CT images; reader 1 and reader 2 had 5 years and 10 years of experience in abdominal imaging, respectively. Blinded to all clinical and pathological information, the radiologists assessed the location of the primary tumor and TNM staging in accordance with the 8th AJCC staging system. The T stage (cT stage) and lymph node status (cLN status) were determined from the CT images by integrating the results of the two radiologists (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). In this study, cT1 and cT2 tumors were grouped as cT1-2 due to the limitations of contrast-enhanced CT in differentiating T1 from T2 in CRC patients. Regional lymph node diameter more than 10 mm or at least 3 clustered lymph nodes were defined as cLN status positive. The weighted kappa statistics test was applied to assess the inter-observer variability regarding cT/cN. During the assessment of CT images, any disagreements between the radiologists were resolved through joint discussion.\u003c/p\u003e\u003cp\u003e\u003cb\u003eImaging segmentation\u003c/b\u003e\u003c/p\u003e\u003cp\u003eThe open-source software ITK-SNAP version 3.8.0 (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.itksnap.org\u003c/span\u003e\u003cspan address=\"http://www.itksnap.org\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) was employed to import the portal-phase CT images to delineate the region of interest (ROI). The 3D ROI was defined as the full-volume area covering the entire tumor [26]. The two radiologists manually delineated each 3D ROI along the area of abnormal enhancement on each primary tumor slice. During the delineation process, the radiologists made efforts to exclude intestinal gas, intestinal juice, and feces from the ROIs.\u003c/p\u003e\u003cp\u003e\u003cb\u003eRadiomics feature extraction and selection\u003c/b\u003e\u003c/p\u003e\u003cp\u003eTo ensure the consistency in original image thickness between the training and testing sets, all scans were resampled using the Simple ITK BSpline interpolation method. Feature extraction was performed using IntelliSpace Medicina Scientia (ISMS) v2.7.0 (Philips Healthcare, China), adhering to the Imaging Biomarker Standardization Initiative (IBSI) for most features. To capture high-order variants of image characteristics, high- and low-frequency wavelet, exponential, log-sigma, square, square root, gradient, and logarithm transformations were applied. The extracted features in this study included shape, first-order (mean, median, maximum, skewness), GLCM (gray-level co-occurrence matrix), GLRLM (gray-level run length matrix), GLSZM (gray-level size zone matrix), GLDM (gray-level dependence matrix), and NGTDM (neighborhood gray-tone difference matrix). Subsequently, all features were standardized based on the average value and standard deviation of the training set.\u003c/p\u003e\u003cp\u003eFeature selection operations were conducted on the training set. Firstly, 30 cases were randomly selected from the training set to evaluate the observer repeatability of the features. During the same period, the two radiologists separately outlined the 3D ROIs for 30 cases to assess the inter-observer repeatability. One month later, reader 1 delineated the 3D ROIs for the same 30 cases again to assess the intra-observer repeatability. The inter-observer and intra-observer correlation coefficients (ICC) were calculated to assess the agreement. Features showing inter- and intra-observer ICCs both exceeding 0.75 were considered to have satisfactory reliability and reproducibility and were included in the subsequent analysis. Secondly, a Pearson correlation coefficient threshold of 0.85 was utilized to remove redundant features. Finally, the least absolute shrinkage and selection operator (LASSO) was employed to recognize the most relevant features for efficient model training.\u003c/p\u003e\u003cp\u003e\u003cb\u003eModel building and evaluation\u003c/b\u003e\u003c/p\u003e\u003cp\u003eThe model training process employed a fully automated pipeline. Nine candidate classifiers were evaluated, including nonlinear support vector machine (SVM), naive Bayes, logistic regression (LR), k-nearest neighbors (KNN), multi-layer perceptron (MLP), random forest, decision tree, ADABoost, and XGBoost. For each classifier, a hyperparameter search space was defined, followed by a grid search on the training set to determine the optimal hyperparameter configuration. The detailed search space is displayed in Supplementary Table S2. Thereafter, a five-fold cross-validation scheme was implemented, and the model exhibiting the maximum validation AUC score was selected as the final model. The training process used Python code and the scikit-learn library (version 1.1.2).\u003c/p\u003e\u003cp\u003eThe receiver operating characteristic (ROC) curves and precision-recall (PR) curves were constructed to assess the diagnostic capabilities of the models. F1-scores were calculated to measure binary classification performance, defined as the harmonic mean of positive predictive value (PPV) and sensitivity. Pearson correlation was performed using Scipy (version 1.7.1). Model performance was compared using the area under the curve (AUC) of ROC curves and average precision (AP) of PR curves. The value of the model in clinical practice was tested by conducting a decision curve analysis (DCA). Figure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e shows the workflow of this study for constructing and validating the radiomics models.\u003c/p\u003e\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003eStatistical analysis\u003c/h2\u003e\u003cp\u003eThe SPSS software (version 27.0; IBM, New York, USA) was used to perform statistical analyses. Continuous variables conforming to a normal distribution were expressed in the form of mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation (SD) and compared using an independent sample t-test. Categorical variables were expressed as raw numbers and compared using either the chi-square (X\u003csup\u003e2\u003c/sup\u003e) test or Fisher's exact test. AUCs of the models were compared by DeLong\u0026rsquo;s test. In this study, a P-value of less than 0.05 was considered statistically significant.\u003c/p\u003e\u003c/div\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cb\u003ePatient characteristics\u003c/b\u003e\u003c/p\u003e\u003cp\u003eThis study included a total of 334 CRC patients(131 PNI\u0026thinsp;+\u0026thinsp;and 203 PNI-) who underwent contrast-enhanced CT in two hospitals from July 2020 to June 2023. Significant differences in clinical characteristics were observed between the PNI\u0026thinsp;+\u0026thinsp;and PNI - groups, including CEA, cT stage, cLN status, pT stage, pN stage, LVI, and tumor deposits (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). The PNI\u0026thinsp;+\u0026thinsp;and PNI- groups showed no significant difference in respect of age, gender, family history, history of alcoholism, location, and histological grade (p\u0026thinsp;\u0026gt;\u0026thinsp;0.05). The patients were divided into three subsets: the training set (n\u0026thinsp;=\u0026thinsp;132), the internal validation set (n\u0026thinsp;=\u0026thinsp;33), and the external testing set (n\u0026thinsp;=\u0026thinsp;169). Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e compares the clinical characteristics of patients in the three cohorts.\u003c/p\u003e\u003cp\u003e\u003cb\u003eFeature selection and model building\u003c/b\u003e\u003c/p\u003e\u003cp\u003eA total of 1473 features, including 16 shape features, 288 first-order features, and 1169 textural features, presented good reliability and reproducibility, with the inter- and intra-observer ICCs both exceeding 0.75. After deleting redundant features, the 25 most significant radiomics features were selected by analysis of variance (ANOVA) and LASSO LR, comprising 6 first-order features and 19 textural features (Figure \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e). The Rad-score was constructed through logistic regression and was identified as an independent predictor significantly related to the PNI status in CRC patients (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Subsequently, the radiomics model was built upon the Rad-score through linear combination of the selected features and their corresponding LASSO coefficients.\u003c/p\u003e\u003cp\u003eThrough univariate and multivariate analysis, CEA, cT stage, and cLN status were screened as independent risk factors for predicting PNI (Odds ratio [OR]\u0026thinsp;=\u0026thinsp;2.26, 10.99, and 1.96; P\u0026thinsp;=\u0026thinsp;0.01, 0.008, and 0.033, respectively), as shown in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. Therefore, a clinical model based on CEA, cT stage, and cLN status was constructed. To build a more robust prediction model, an ensemble model was established by integrating Rad-score with three clinical independent predictors.\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\u003eIndependent predictors of PNI status in CRC\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"6\"\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=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" 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=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eVariables\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eβ\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"4\" nameend=\"c6\" namest=\"c3\"\u003e\u003cp\u003eNomogram\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eStandard error\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eWald\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eOR (95% CI)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003ep\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eCEA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e0.817\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e0.316\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e6.67\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e2.26\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e0.01\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e(1.23\u0026ndash;4.25)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003ecTstage(T3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e2.522\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e0.855\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e8.7\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e12.45\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e0.003\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e(2.84\u0026ndash;91.94)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003ecTstage(T4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e2.397\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e0.899\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e7.11\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e10.99\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e0.008\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e(2.25\u0026ndash;86.49)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003ecLNstatus\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e0.671\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e0.314\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e4.57\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e1.96\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e0.033\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e(1.06\u0026ndash;3.65)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eRad-Score\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e1.834\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e0.211\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e75.53\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e6.26\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e(4.24\u0026ndash;9.73)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003cp\u003eAbbreviations: CEA, carcinoembryonic antigen; cT stage, CT-reported T stage; cLN status, CT-reported lymph node status; β, the regression coefficient; OR, odds ratio; CI, confidence interval\u003c/p\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eModel comparison and validation\u003c/b\u003e\u003c/p\u003e\u003cp\u003eIn this study, five-fold leave-group-out cross-validation (LGOCV) analysis was applied for the purpose of comparing the robustness and reliability of the radiomics models. The radiomics model built upon the Rad-score demonstrated satisfactory discriminative ability, with an AUC of 0.847 (95% confidence interval [CI]: 0.690\u0026ndash;1.000) in the internal validation set and 0.921 (95% CI: 0.877\u0026ndash;0.966) in the external testing set (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). The clinical model achieved AUCs of 0.713 (95%CI: 0.516\u0026ndash;0.910) in the internal validation set and 0.597 (95%CI: 0.508\u0026ndash;0.685) in the external testing set. Combining the Rad-score with the clinical factors led to improved performance. As presented in Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e and Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e, the AUCs of the ensemble model were 0.864 (95% CI: 0.714\u0026ndash;1.000) in the internal validation set and 0.907 (95% CI: 0.858\u0026ndash;0.955) in the external testing set, were significantly greater than those of the clinical model (P\u0026thinsp;\u0026lt;\u0026thinsp;0.05). However, no notable distinction was found when comparing the ensemble model and radiomics model in the internal validation set (AUC\u0026thinsp;=\u0026thinsp;0.864 vs 0.847, P\u0026thinsp;=\u0026thinsp;0.466) and external validation set (AUC\u0026thinsp;=\u0026thinsp;0.907 vs 0.921, P\u0026thinsp;=\u0026thinsp;0.648). The diagnostic performance of the aforementioned models was also compared using the AP of PR curves, as shown in Figure S2. The DCA indicated that the radiomics model had a higher overall net benefit than that of the clinical model and the ensemble model in predicting PNI, as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e. The calibration curve of the radiomics model is shown in Figure S3, with the actual observed probability of PNI represented on the y-axis, and the predicted probability of PNI shown on the x-axis. The general trend of the predicted risk was demonstrated by the locally weighted regression line (solid line) of calibration plots. In this study, the radiomics model demonstrated a favorable consistency between the observed and predicted probabilities. Notably, the solid lines in the internal validation set (the green solid line) and the external testing set (the orange solid line) were in close proximity to the reference line (dotted line). The result of the Hosmer-Lemeshow test was in accordance with this conclusion.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eComparisons of models in the internal training and external testing sets.\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=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eSubsets\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eDiagnostic Index\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"3\" nameend=\"c5\" namest=\"c3\"\u003e\u003cp\u003eModels\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eClinical Model\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eRadiomics Model\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eEnsemble Model\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"4\" rowspan=\"5\"\u003e\u003cp\u003eInternal Validation set\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAUC\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.713\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.847\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.864\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e(95%CI)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e(0.516\u0026ndash;0.910)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e(0.690-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e(0.714-1.000)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAccuracy(%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.788\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.788\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.879\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSensitivity(%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.727\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1.000\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.818\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSpecificity(%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.818\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.682\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.909\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"4\" rowspan=\"5\"\u003e\u003cp\u003eExternal Testing set\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAUC\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.597\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.921\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.907\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e(95%CI)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e(0.508\u0026ndash;0.685)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e(0.877\u0026ndash;0.966)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e(0.858\u0026ndash;0.955)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAccuracy(%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.615\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.845\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.763\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSensitivity(%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.641\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.883\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.949\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSpecificity(%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.590\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\u003cp\u003e0.577\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003ctfoot\u003e\u003ctr\u003e\u003ctd colspan=\"5\"\u003eAbbreviations: AUC, area under the curve; CI, confidence interval; CI, confidence interval\u003c/td\u003e\u003c/tr\u003e\u003c/tfoot\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eFeature importance and visualization\u003c/b\u003e\u003c/p\u003e\u003cp\u003eIn order to explain and visualize the features that were important for the model, a hierarchically-clustered heatmap was plotted (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e) for the radiomics model, illustrating the 25 retained radiomics features along with their corresponding coefficients following dimensionality reduction through LASSO regression analysis. In addition, the SHapley Additive exPlanation (SHAP) summary chart was plotted for the ensemble model, which indicated the direction and degree of the influence of each feature on the model, as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003cb\u003eNomogram construction\u003c/b\u003e\u003c/p\u003e\u003cp\u003eWe established the nomogram through the training set and then validated its effectiveness in the internal validation and external testing sets. The nomogram was created to visualize the ensemble model (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e). A greater total score on the nomogram indicates an increased predicted probability of PNI.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThe objective of this research was to construct and validate a contrast-enhanced CT-based radiomics nomogram from multicenter datasets for the preoperative prediction of the PNI status in CRC patients. The main findings are summarized as follows: (1) The discriminative capacity of the radiomics model built upon the Rad-score was proved to be highly satisfactory in both the internal validation and external testing sets; (2) CT-reported T (cT) stage, lymph node (cLN) status, and the level of carcinoembryonic antigen (CEA) were clinical predictors for PNI status, which were utilized for building the clinical model; (3) The diagnostic performance of the ensemble model constructed by integrating Rad-score with cT stage, cLN status, and CEA was better than that of the clinical model, but was not significantly different from that of the radiomics model.\u003c/p\u003e\u003cp\u003eRecent studies have indicated that PNI is not merely the spread of cancer cells along the connective tissues enveloping the nerve sheath. Instead, it is a complex biological process where various neurotrophic factors and chemokines interact with cancer cells and the surrounding microenvironment [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. PNI is associated with cancer invasion, local recurrence, and metastasis, leading to unfavorable clinical outcomes. Growing evidence shows that PNI is an independent risk factor for poor prognosis of colorectal cancer. Therefore, accurately predicting the PNI status of CRC patients contributes to the development of an individualized therapeutic plan and prognosis evaluation [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eIn comparison with CT and MRI, radiomics can yield favorable results for predicting PNI. This is accomplished by extracting high-throughput features that can capture subtle tissue differences undetectable by the naked eye. Different MR-based radiomics models have been documented for predicting the PNI status in RC [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. These studies enrolled patients with rectal cancer from single institutions, thereby limiting the predictive value and generalizability of the results. Recently, some multicenter studies [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e] have reported promising efficacy using machine learning or nomograms based on CT radiomics to predict PNI in colorectal cancer patients. However, the sample sizes of the previous studies were relatively small. In this study, 334 patients were enrolled from two hospitals, and the internal and external validations were performed for the established models, which enhanced the generalization capability of the model and the reliability of the conclusions.\u003c/p\u003e\u003cp\u003eIn this study, multivariable analysis results revealed that the cT stage and cLN status were independent clinical factors for predicting PNI in CRC, indicating that the tumor's invasive depth and lymph node metastasis were closely related to PNI. With the progression of the cT stage and cLN status, the tumor shows higher aggressiveness and malignancy, increasing the risk of PNI for CRC patients. These findings are in accordance with the observations of previous studies [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. Similar to the results of Que Y et al. [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e], the present study identified CEA as another independent predictor for PNI, highlighting its significance as a serum tumor marker. CEA may indirectly enhance tumor invasiveness by promoting the interaction between tumor cells and the neural microenvironment. Nevertheless, the findings of this research indicated that the clinical models based on cT stage, cLN status, and CEA showed moderate predictive performance in the internal validation set, possibly because cT stage and cLN status were easily influenced by the subjectivity and experience between observers, leading to limited predictive value [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eIn our study, the Pearson correlation coefficient and LASSO algorithm were utilized to remove the most redundant features and select the most relevant features. Five-fold cross-validation was performed to ensure the stability of the selected radiomics features. Thereafter, a collection of selected radiomics features was weighted by the regression coefficients to construct the Rad-score [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. Moreover, the radiomics model based on Rad-score exhibited high discrimination capability, with an AUC of 0.847 in the internal validation and 0.921 in the external testing sets. The radiomics features selected for building the radiomics model included first-order statistical features and texture features. Among them, texture features accounted for the highest proportion as they reflected the tissue heterogeneity by quantifying the spatial distribution pattern of gray intensity in the image.\u003c/p\u003e\u003cp\u003eTo improve the predictive performance, Rad-score was combined with independent clinical predictors to construct the ensemble model. The diagnostic performance of the ensemble model was much better than that of the clinical model, but was not obviously different from that of the radiomics model. This finding indicates that Rad-score contributed the most to the model construction process. The radiomics model based on Rad-score alone yielded similar predictive value to the ensemble model.\u003c/p\u003e\u003cp\u003eTo facilitate personalized medical treatment, the ensemble model was converted into a nomogram, predicting an individual risk probability of PNI. As evidenced by its longer axis in the nomogram, Rad-score held greater significance compared to independent clinical predictors. This user-friendly system assists physicians in conducting preoperative assessments of PNI status and customizing clinical management strategies [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eNevertheless, the limitations of the present study should be acknowledged. First, potential selection bias may be present due to the retrospective study design. Second, the manual delineation of the ROIs was performed by skilled radiologists, but it remains a time-consuming process. In the future, the image segmentation process should be simplified. Deep learning, with its potential for automatic segmentation, offers a viable solution for this purpose.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThe radiomics nomogram integrating Rad-score and clinical predictors can help to evaluate the PNI status before surgery and guide personalized treatment of CRC patients.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe thank Home for Researchers editorial team (www.home-for-researchers.com) for language editing service.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eConception/design: Chao Su. Provision of study material or patients: Fajuan Shen, Lingling Feng, Qianqian Xia. Collection and/or assembly of data: Yafei Wang, Peijie Shen, Song Tian. Data analysis and interpretation: Yafei Wang, Peijie Shen, Song Tian. Manuscript writing: Juan Zhang, Chao Su. Final approval of manuscript: All authors.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe data underlying this article will be shared on reasonable request to the corresponding author.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests statement\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics declarations and consent statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe ethics committee of the Second Affiliated Hospital of Shandong First Medical University has approved our study. The need to obtain informed consent was waived by the ethics committee of the Second Affiliated Hospital of Shandong First Medical University due to the retrospective nature of the study. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent to publish\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll authors have agreed to publish this manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that no funds, grants, or other support were received during the preparation of this manuscript.\u0026nbsp;\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eSiegel, R. L. et al. Cancer statistics, 2021. \u003cem\u003eCA Cancer J. Clin.\u003c/em\u003e \u003cb\u003e71\u003c/b\u003e, 7\u0026ndash;33. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3322/caac.21654\u003c/span\u003e\u003cspan address=\"10.3322/caac.21654\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKim, C. H. et al. Prognostic impact of perineural invasion in rectal cancer after neoadjuvant chemoradiotherapy. \u003cem\u003eWorld J. Surg.\u003c/em\u003e \u003cb\u003e43\u003c/b\u003e, 260\u0026ndash;272. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s00268-018-4774-8\u003c/span\u003e\u003cspan address=\"10.1007/s00268-018-4774-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2019).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDhadda, A. S. et al. Mandard tumour regression grade, perineural invasion, circumferential resection margin and post-chemoradiation nodal status strongly predict outcome in locally advanced rectal cancer treated with preoperative chemoradiotherapy. \u003cem\u003eClin. Oncol.\u003c/em\u003e \u003cb\u003e26\u003c/b\u003e, 197\u0026ndash;202. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.clon.2014.01.001\u003c/span\u003e\u003cspan address=\"10.1016/j.clon.2014.01.001\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2014).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKim, Y. I. et al. Clinical implication of perineural and lymphovascular invasion in rectal cancer patients who underwent surgery after preoperative chemoradiotherapy. \u003cem\u003eDis. Colon Rectum\u003c/em\u003e. \u003cb\u003e65\u003c/b\u003e, 1325\u0026ndash;1334. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1097/DCR.0000000000002219\u003c/span\u003e\u003cspan address=\"10.1097/DCR.0000000000002219\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang, B. et al. Combining perineural invasion with staging improve the prognostic accuracy in colorectal cancer: a retrospective cohort study. \u003cem\u003eBMC Cancer\u003c/em\u003e. \u003cb\u003e23\u003c/b\u003e, 675. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12885-023-11114-8\u003c/span\u003e\u003cspan address=\"10.1186/s12885-023-11114-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSong, J. H. et al. Significance of perineural and lymphovascular invasion in locally advanced rectal cancer treated by preoperative chemoradiotherapy and radical surgery: Can perineural invasion be an indication of adjuvant chemotherapy? Radiother. \u003cem\u003eOncol\u003c/em\u003e \u003cb\u003e133\u003c/b\u003e, 125\u0026ndash;131. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.radonc.2019.01.002\u003c/span\u003e\u003cspan address=\"10.1016/j.radonc.2019.01.002\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2019).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChen, T. et al. The incidence and prognosis value of perineural invasion in rectal carcinoma: From meta-analyses and real-world clinical pathological features. \u003cem\u003eClin. Oncol.\u003c/em\u003e \u003cb\u003e35\u003c/b\u003e, e611\u0026ndash;e621. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.clon.2023.05.008\u003c/span\u003e\u003cspan address=\"10.1016/j.clon.2023.05.008\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eQin, L. et al. Perineural invasion affects prognosis of patients undergoing colorectal cancer surgery: a propensity score matching analysis. \u003cem\u003eBMC Cancer\u003c/em\u003e. \u003cb\u003e23\u003c/b\u003e, 452. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12885-023-10936-w\u003c/span\u003e\u003cspan address=\"10.1186/s12885-023-10936-w\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGarc\u0026iacute;a-Figueiras, R. et al. Advanced imaging techniques in evaluation of colorectal cancer. \u003cem\u003eRadiographics\u003c/em\u003e \u003cb\u003e38\u003c/b\u003e, 740\u0026ndash;765. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1148/rg.2018170044\u003c/span\u003e\u003cspan address=\"10.1148/rg.2018170044\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2018).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGlynne-Jones, R. et al. Rectal cancer: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. \u003cem\u003eAnn. Oncol.\u003c/em\u003e \u003cb\u003e28\u003c/b\u003e, iv22\u0026ndash;iv40. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1093/annonc/mdx224\u003c/span\u003e\u003cspan address=\"10.1093/annonc/mdx224\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2017).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTaghavi, M. et al. Machine learning-based analysis of CT radiomics model for prediction of colorectal metachronous liver metastases. \u003cem\u003eAbdom. Radiol.\u003c/em\u003e \u003cb\u003e46\u003c/b\u003e, 249\u0026ndash;256. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s00261-020-02624-1\u003c/span\u003e\u003cspan address=\"10.1007/s00261-020-02624-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChen, J. et al. Pretreatment MR-based radiomics signature as potential imaging biomarker for assessing the expression of topoisomerase II alpha (TOPO-IIα) in rectal cancer. \u003cem\u003eJ. Magn. Reson. Imaging\u003c/em\u003e. \u003cb\u003e51\u003c/b\u003e, 1881\u0026ndash;1889. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/jmri.26972\u003c/span\u003e\u003cspan address=\"10.1002/jmri.26972\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCui, Y. et al. Development and validation of a MRI-based radiomics signature for prediction of KRAS mutation in rectal cancer. \u003cem\u003eEur. Radiol.\u003c/em\u003e \u003cb\u003e30\u003c/b\u003e, 1948\u0026ndash;1958. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s00330-019-06572-3\u003c/span\u003e\u003cspan address=\"10.1007/s00330-019-06572-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGuo, Y. et al. Preoperative prediction of perineural invasion with multi-modality radiomics in rectal cancer. \u003cem\u003eSci. Rep.\u003c/em\u003e \u003cb\u003e11\u003c/b\u003e, 9429. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41598-021-88831-2\u003c/span\u003e\u003cspan address=\"10.1038/s41598-021-88831-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYang, Y. S. et al. High resolution MRI-based radiomic nomogram in predicting perineural invasion in rectal cancer. \u003cem\u003eCancer Imaging\u003c/em\u003e. \u003cb\u003e21\u003c/b\u003e, 40. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s40644-021-00408-4\u003c/span\u003e\u003cspan address=\"10.1186/s40644-021-00408-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWeiser, M. R. AJCC 8th edition: Colorectal cancer. \u003cem\u003eAnn. Surg. Oncol.\u003c/em\u003e \u003cb\u003e25\u003c/b\u003e, 1454\u0026ndash;1455. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1245/s10434-018-6462-1\u003c/span\u003e\u003cspan address=\"10.1245/s10434-018-6462-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2018).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKnijn, N. et al. Perineural invasion is a strong prognostic factor in colorectal cancer: A systematic review. \u003cem\u003eAm. J. Surg. Pathol.\u003c/em\u003e \u003cb\u003e40\u003c/b\u003e, 103\u0026ndash;112. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1097/PAS.0000000000000518\u003c/span\u003e\u003cspan address=\"10.1097/PAS.0000000000000518\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2016).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChen, J. et al. Pretreatment MR-based radiomics nomogram as potential imaging biomarker for individualized assessment of perineural invasion status in rectal cancer. \u003cem\u003eAbdom. Radiol.\u003c/em\u003e \u003cb\u003e46\u003c/b\u003e, 847\u0026ndash;857. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s00261-020-02710-4\u003c/span\u003e\u003cspan address=\"10.1007/s00261-020-02710-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang, H. et al. Perineural invasion in colorectal cancer: mechanisms of action and clinical relevance. \u003cem\u003eCell. Oncol.\u003c/em\u003e \u003cb\u003e47\u003c/b\u003e (1), 1\u0026ndash;17. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s13402-023-00857-y\u003c/span\u003e\u003cspan address=\"10.1007/s13402-023-00857-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChen, Q. et al. Computed tomography-based radiomics nomogram for the preoperative prediction of perineural invasion in colorectal cancer: a multicentre study. \u003cem\u003eAbdom. Radiol.\u003c/em\u003e \u003cb\u003e47\u003c/b\u003e, 3251\u0026ndash;3263. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s00261-022-03620-3\u003c/span\u003e\u003cspan address=\"10.1007/s00261-022-03620-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLiu, N. J. et al. The value of machine learning based on CT radiomics in the preoperative identification of peripheral nerve invasion in colorectal cancer: a two-center study. \u003cem\u003eInsights Imaging\u003c/em\u003e. \u003cb\u003e15\u003c/b\u003e, 101. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s13244-024-01664-1\u003c/span\u003e\u003cspan address=\"10.1186/s13244-024-01664-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLi, Y. et al. Preoperative prediction of perineural invasion and KRAS mutation in colon cancer using machine learning. \u003cem\u003eJ. Cancer Res. Clin. Oncol.\u003c/em\u003e \u003cb\u003e146\u003c/b\u003e, 3165\u0026ndash;3174. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s00432-020-03354-z\u003c/span\u003e\u003cspan address=\"10.1007/s00432-020-03354-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eQue, Y. et al. A prediction nomogram for perineural invasion in colorectal cancer patients: a retrospective study. \u003cem\u003eBMC Surg.\u003c/em\u003e \u003cb\u003e24\u003c/b\u003e, 80. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s12893-024-02364-9\u003c/span\u003e\u003cspan address=\"10.1186/s12893-024-02364-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"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":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Computed tomography, Perineural invasion, Radiomics, Colorectal cancer","lastPublishedDoi":"10.21203/rs.3.rs-6970474/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6970474/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eObjectives/Purpose\u003c/h2\u003e\u003cp\u003eTo develop and validate a radiomics nomogram for preoperative prediction of perineural invasion (PNI) in colorectal cancer (CRC) using multi-center contrast-enhanced CT datasets.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e\u003cp\u003eThis retrospective study enrolled 334 CRC patients from two hospitals, divided into the training, internal validation, and external validation sets. Radiomics features were extracted and selected by the least absolute shrinkage and selection operator (LASSO) logistic regression to build a radiomics model. Clinical independent risk factors were used to construct a clinical model. An ensemble model was established by integrating the Rad-score with clinical predictors. Model performance was evaluated using ROC curves and decision curve analysis (DCA).\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e\u003cp\u003eThe radiomics model was built from whole-tumor radiomics features, while the clinical model incorporated CEA, CT-reported T stage, and lymph node status. The ensemble model achieved the best performance in the internal validation set, whereas the radiomics model demonstrated greater stability in the external validation set. A nomogram combining the Rad-score and clinical predictors was developed for preoperative PNI prediction.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e\u003cp\u003eThe contrast-enhanced CT-based radiomics nomogram is a promising non-invasive tool for preoperative PNI assessment in CRC patients, showing robust performance across multi-center datasets.\u003c/p\u003e","manuscriptTitle":"Contrast-enhanced CT-based radiomics nomogram for the preoperative prediction of perineural invasion in colorectal cancer: a two-center study","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-11-04 13:41:50","doi":"10.21203/rs.3.rs-6970474/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"582846e9-7d3c-4810-bc43-105d736c956e","owner":[],"postedDate":"November 4th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":57372452,"name":"Biological sciences/Cancer"},{"id":57372453,"name":"Health sciences/Gastroenterology"}],"tags":[],"updatedAt":"2025-12-19T12:09:21+00:00","versionOfRecord":[],"versionCreatedAt":"2025-11-04 13:41:50","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6970474","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6970474","identity":"rs-6970474","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}