Developing a Nomogram to Stratify Intracranial Solitary Fibrous Tumor Recurrence.

RATIONALE AND OBJECTIVES: To develop a nomogram to stratify tumor recurrence (TR) in intracranial solitary fibrous tumors (ISFTs) based on the clinical, radiological, and pathological features. MATERIALS AND METHODS: A total of 215 patients from Beijing Tiantan Hospital, Capital Medical University and 48 patients from Lanzhou University Second Hospital, diagnosed with ISFT based on histopathological findings, were included. The patients were randomly divided into training and test cohorts at a ratio of 8:2. Information regarding clinical, radiological, and histopathological features, and the clinical outcomes was retrospectively analyzed. Univariate and multivariate analyses were performed using the Cox proportional hazard model for TR in the training cohort. A nomogram incorporating the independent risk factors was developed in the training cohort and validated in the test cohort. Its predictive performance was analyzed using the Harrell C-index. Decision curve analysis (DCA) was used to evaluate the net clinical benefit. RESULTS: The Harrell C-indices for TR at 3 and 5 years were 0.845 (0.578-0.944) and 0.807 (0.612-0.901) for the test cohort, respectively. In the test cohort, the nomogram provided a net clinical benefit in the DCA over the TR scheme or non-TR scheme. Although postoperative radiotherapy (PORT) was useful for TR prevention, high doses (≥46 Gy) were not superior to lower doses in prolonging the progression-free survival. CONCLUSION: The nomogram obtained in our study had a good predictive performance and could be used for ISFT patients.

Vasari-Based Features Nomogram to Predict the Tumor-Infiltrating CD8+ T Cell Levels in Glioblastoma.

RATIONALE AND OBJECTIVES: Tumor-infiltrating CD8 + T cells play a key role in glioblastoma (GB) development, malignant progression, and recurrence. The aim of the study was to establish nomograms based on the Visually AcceSAble Rembrandt Images (VASARI) features of multiparametric magnetic resonance imaging (MRI) to determine the expression levels of tumor-infiltrating CD8 + T cells in patients with GB. MATERIALS AND METHODS: Pathological and imaging data of 140 patients with GB confirmed by surgery and pathology were retrospectively analyzed. The levels of tumor-infiltrating CD8 + T cells in tumor tissue samples obtained from patients were quantified using immunohistochemical staining. Patients were divided into high and low CD8 expression groups. The MRI images of patients with GB were analyzed by two radiologists using the VASARI scoring system. RESULTS: A total of 25 MRI-based VASARI imaging features were evaluated by two neuroradiologists. The features with the greatest predictive power for CD8 expression levels were, cystic (OR, 3.063; 95% CI: 1.387, 6.766; P = 0.006), hemorrhage (OR, 2.980; 95% CI: 1.172, 7.575; P = 0.022), and ependymal extension (OR, 0.257; 95% CI: 0.114 0.581; P = 0.001). A logistic regression model based on these three features showed better sample predictive performance (AUC=0.745; 95% CI: 0.665, 0.825; Sensitivity=0.527; Specificity=0.857). CONCLUSION: The VASARI feature-based nomogram model can show promise to predict the level of infiltrative CD8 expression in GB tumors non-invasively for earlier tissue diagnosis and more aggressive treatment.

Development of an MRI-Based Comprehensive Model Fusing Clinical, Radiomics and Deep Learning Models for Preoperative Histological Stratification in Intracranial Solitary Fibrous Tumor.

BACKGROUND: Accurate preoperative histological stratification (HS) of intracranial solitary fibrous tumors (ISFTs) can help predict patient outcomes and develop personalized treatment plans. However, the role of a comprehensive model based on clinical, radiomics and deep learning (CRDL) features in preoperative HS of ISFT remains unclear. PURPOSE: To investigate the feasibility of a CRDL model based on magnetic resonance imaging (MRI) in preoperative HS in ISFT. STUDY TYPE: Retrospective. POPULATION: Three hundred and ninety-eight patients from Beijing Tiantan Hospital, Capital Medical University (primary training cohort) and 49 patients from Lanzhou University Second Hospital (external validation cohort) with ISFT based on histopathological findings (237 World Health Organization [WHO] tumor grade 1 or 2, and 210 WHO tumor grade 3). FIELD STRENGTH/SEQUENCE: 3.0 T/T1-weighted imaging (T1) by using spin echo sequence, T2-weighted imaging (T2) by using fast spin echo sequence, and T1-weighted contrast-enhanced imaging (T1C) by using two-dimensional fast spin echo sequence. ASSESSMENT: Area under the receiver operating characteristic curve (AUC) was used to assess the performance of the CRDL model and a clinical model (CM) in preoperative HS in the external validation cohort. The decision curve analysis (DCA) was used to evaluate the clinical net benefit provided by the CRDL model. STATISTICAL TESTS: Cohen's kappa, intra-/inter-class correlation coefficients (ICCs), Chi-square test, Fisher's exact test, Student's t-test, AUC, DCA, calibration curves, DeLong test. A P value <0.05 was considered statistically significant. RESULTS: The CRDL model had significantly better discrimination ability than the CM (AUC [95% confidence interval, CI]: 0.895 [0.807-0.912] vs. 0.810 [0.745-0.874], respectively) in the external validation cohort. The CRDL model can provide a clinical net benefit for preoperative HS at a threshold probability >20%. DATA CONCLUSION: The proposed CRDL model holds promise for preoperative HS in ISFT, which is important for predicting patient outcomes and developing personalized treatment plans. LEVEL OF EVIDENCE: 3 TECHNICAL EFFICACY: Stage 2.

Whole-tumor histogram analysis of multi-parametric MRI for differentiating brain metastases histological subtypes in lung cancers: relationship with the Ki-67 proliferation index.

This study aims to investigate the predictive value of preoperative whole-tumor histogram analysis of multi-parametric MRI for histological subtypes in patients with lung cancer brain metastases (BMs) and explore the correlation between histogram parameters and Ki-67 proliferation index. The preoperative MRI data of 95 lung cancer BM lesions obtained from 73 patients (42 men and 31 women) were retrospectively analyzed. Multi-parametric MRI histogram was used to distinguish small-cell lung cancer (SCLC) from non-small cell lung cancer (NSCLC), and adenocarcinoma (AC) from squamous cell carcinoma (SCC), respectively. The T1-weighted contrast-enhanced (T1C) and apparent diffusion coefficient (ADC) histogram parameters of the volumes of interest (VOIs) in all BMs lesions were extracted using FireVoxel software. The following histogram parameters were obtained: maximum, minimum, mean, standard deviation (SD), variance, coefficient of variation (CV), skewness, kurtosis, entropy, and 1st-99th percentiles. Then investigated their relationship with the Ki-67 proliferation index. The skewness-T1C, kurtosis-T1C, minimum-ADC, mean-ADC, CV-ADC and 1st - 90th ADC percentiles were significantly different between the SCLC and NSCLC groups (all p < 0.05). When the 10th-ADC percentile was 668, the sensitivity, specificity, and accuracy (90.80%, 76.70% and 86.32%, respectively) for distinguishing SCLC from NSCLC reached their maximum values, with an AUC of 0.895 (0.824 - 0.966). Mean-T1C, CV-T1C, skewness-T1C, 1st - 50th T1C percentiles, maximum-ADC, SD-ADC, variance-ADC and 75th - 99th ADC percentiles were significantly different between the AC and SCC groups (all p < 0.05). When the CV-T1C percentiles was 3.13, the sensitivity, specificity and accuracy (75.00%, 75.60% and 75.38%, respectively) for distinguishing AC and SCC reached their maximum values, with an AUC of 0.829 (0.728-0.929). The 5th-ADC and 10th-ADC percentiles were strongly correlated with the Ki-67 proliferation index in BMs. Multi-parametric MRI histogram parameters can be used to identify the histological subtypes of lung cancer BMs and predict the Ki-67 proliferation index.

Correlation of tumor-associated macrophage infiltration in glioblastoma with magnetic resonance imaging characteristics: a retrospective cross-sectional study.

Background: Glioblastoma (Gb) is the most common primary malignant tumor of brain with poor prognosis. Immune cells are the main factors affecting the prognosis of Gb, tumor-associated macrophages (TAMs) are the predominant infiltrating immune cell population in the immune microenvironment of Gb. Analyzing the relationship between magnetic resonance imaging (MRI) features and TAMs of Gb, and using imaging features to characterize the infiltration level of TAMs in tumor tissue may provide indicators for clinical decision-making and prognosis evaluation of Gb. Methods: Data from 140 in patients with isocitrate dehydrogenase (IDH) wild-type Gb diagnosed via histopathology and molecular diagnosis in the Second Hospital of Lanzhou University from January 2018 to April 2022 were collected in this retrospective, cross-sectional study. MRI images were reviewed for lesion location, cyst, necrosis, hemorrhage, contrast-enhanced T1-weighted MRI signal intensity, average apparent diffusion coefficient (ADCmean), and minimum apparent diffusion coefficient (ADCmin). Immunohistochemical staining with anti-CD163 and anti-CD68 antibodies was employed for macrophage detection. The positive cell percentage was estimated in 9 microscopic fields at 400× magnification per whole-slide image with ImageJ software (National Institutes of Health). Additionally, the relationship between MRI features, molecular, states and the positive CD68 and CD163 expression was analyzed. Results: Our study discovered that the mean or median values of CD68+ and CD163+ TAMs were 7.39% and 14.98%, respectively. There was an obvious correlation between CD163+ TAMs and CD68+ TAMs (r=0.497; P=0.000). CD68+ and CD163+ macrophage infiltration correlated with age at diagnosis in patients with Gb (CD68+: r=0.230, P=0.006; CD163+: r=0.172, P=0.042). The levels of Gb TAM infiltration in different tumor locations varied, with the temporal lobe having the highest CD163+ macrophage and CD68+ macrophage infiltration (18.58% and 9.46%, respectively). CD163+ macrophage infiltration was positively correlated with ADCmean (r=0.208; P=0.014). The infiltration of CD68+ macrophages differed significantly between groups with varying degrees of tumor enhancement (H =4.228; P=0.017). There was a significant difference in CD68+ TAMs and CD163+ TAMs between the wild-type and mutant-type telomerase reverse transcriptase (TERT) types (P=0.004 and P=0.031, respectively). Conclusions: Age, location of the tumor, degree of tumor enhancement, ADC value, and TERT mutation status were associated with macrophage infiltration. These findings may serve as an effective tool for characterizing the tumor microenvironment in patients with Gb.

Utility of Apparent Diffusion Coefficient Histogram Analysis in Differentiating Microcystic Meningioma from Intracranial Solitary Fibrous Tumor.

BACKGROUND: To investigate the possibility of histogram analysis of apparent diffusion coefficient (ADC) maps in differentiating microcystic meningioma (MM) from intracranial solitary fibrous tumor (SFT). METHODS: Eighteen patients with MM and 23 patients with SFT were enrolled in this retrospective study. Conventional magnetic resonance imaging (MRI) features and 9 ADC histogram parameters (including mean, first (ADC1), 10th (ADC10), 50th (ADC50), 90th (ADC90), and 99th (ADC99) percentiles ADC, as well as variance, skewness, and kurtosis) between MM and SFT were compared. The diagnostic performance of the optimal parameter was determined by the receiver operating characteristic analysis. RESULTS: SFT showed a significantly lower mean, ADC1, ADC10, ADC50, ADC90, and ADC99 than MM (all P < 0.05), while no significant difference was found in conventional MRI features or other ADC histogram parameters (all P > 0.05). ADC1 was identified as the optimal parameter in differentiating between MM and SFT, which achieved an area under the curve of 0.861, with sensitivity, specificity, and accuracy of 78.26%, 88.89%, and 82.93%, respectively. CONCLUSIONS: MM and SFT show overlapping conventional MRI features. ADC histogram analysis helps to differentiate between MM and SFT, with ADC1 being the optimal parameter with the best discrimination performance.

Apparent diffusion coefficient values effectively predict cell proliferation and determine oligodendroglioma grade.

This study aims to evaluate the value of conventional magnetic resonance imaging (MRI) features and apparent diffusion coefficient (ADC) values in differentiating oligodendroglioma of various grades and explore the correlation between ADC and Ki-67. The preoperative MRI data of 99 patients with World Health Organization (WHO) grades 2 (n = 42) and 3 (n = 57) oligodendroglioma confirmed by surgery and pathology were retrospectively analyzed. Conventional MRI features, ADCmean, ADCmin, and normalized ADC (nADC) were compared between the two groups. A receiver operating characteristic curve was used to evaluate each parameter's diagnostic efficacy in differentiating the two tumor types. Each tumor's Ki-67 proliferation index was also measured to explore its relationship with the ADC value. Compared with WHO2 grade tumors, WHO3 grade tumors had a larger maximum diameter and more significant cystic degeneration/necrosis, edema, and moderate/severe enhancement (all P < 0.05). The ADCmin, ADCmean, and nADC values of the WHO3 and WHO2 grade tumors were significantly different, and the ADCmin value most accurately distinguished the two tumor types, yielding an area under the curve value of 0.980. When 0.96 × 10-3 mm2/s was used as the differential diagnosis threshold, the sensitivity, specificity, and accuracy of the two groups were 100%, 93.00%, and 96.96%, respectively. The ADCmin (r = -0.596), ADCmean (r = - 0.590), nADC (r = - 0.577), and Ki-67 proliferation index values had significantly negative correlations (all P < 0.05). Conventional MRI features and ADC values are beneficial in the noninvasive prediction of the WHO grade and tumor proliferation rate of oligodendroglioma.

Correlation analysis of tumor purity with clinicopathological, molecular, and imaging features in high-grade gliomas.

High-grade gliomas (HGG) have high malignancy, high heterogeneity, and a poor prognosis. Tumor purity is an intrinsic feature of the HGG microenvironment and an independent prognostic factor. The purpose of this study was to analyze the correlation of tumor purity with clinicopathological, molecular, and imaging features. We performed a retrospective analysis of 112 patients diagnosed with HGG (grades III and IV) in our center. Eleven regions of interest (ROI) were randomly selected on whole-slide images (WSI, 40 × magnification) based on HGG tissue paraffin sections and hematoxylin-eosin (H&E) staining. Of these 11 ROIs, five ROIs were visually estimated by pathologists and six ROIs were automatically analyzed using ImageJ software. Last, the average tumor purity (%) of the 11 ROIs was calculated. Correlation analysis of tumor purity with clinicopathological, molecular, and imaging features was conducted. Of the 112 patients included in the study, the mean tumor purity of HGG was 70.96%. There were differences in tumor purity between WHO grades III and IV; the tumor purity of grade IV patients (67.59%) was lower than that of grade III patients (76.00%) (p < 0.001). There were also differences in tumor purity between IDH1 mutant and wild type, and the tumor purity of IDH1 mutant patients was higher than that of IDH1 wild-type patients (p = 0.006). The average range of peritumoral edema was about 19.18 mm, and the diameter of edema, ADCmean, and ADCmin were negatively correlated with tumor purity(r = - 0.236, r = - 0.306, and r = - 0.242; p < 0.05). The grade of HGG, IDH1 mutant/wild type, peritumoral edema, and ADC value were correlated with tumor purity. HGG grade, IDH1 mutant/wild type, peritumoral edema, and ADC value can predict tumor purity and indirectly reflect patient prognosis.

A Nomogram for Predicting Early Recurrence in Patients with High-Grade Gliomas.

OBJECTIVE: The objective of the study was to develop a nomogram to predict early recurrence of high-grade glioma (HGG) based on clinical pathology, genetic factors, and magnetic resonance imaging parameters. METHODS: One hundred fifty-four patients with HGG were classified into recurrence and nonrecurrence groups based on the pathological diagnosis and Response Assessment in Neuro-Oncology criteria. Clinical pathology information included age, sex, preoperative Karnofsky performance status scores, grade, and cell proliferation index (Ki-67). Gene information included P53, isocitrate dehydrogenase 1 (IDH1), O6-methylguanine-DNA methyltransferase, and telomerase reverse transcriptase expression status. All patients underwent baseline magnetic resonance imaging before treatment, including T1-weighted imaging, T2-weighted imaging, contrast-enhanced T1WI, fluid attenuated inversion recovery, and diffusion-weighted imaging examinations. Tumor location, single/multiple tumors, tumor diameter, peritumoral edema, necrotic cyst, hemorrhage, average apparent diffusion coefficient value, and minimum apparent diffusion coefficient values were evaluated. Univariate and multivariate logistic regression analyses were used to determine the predictors of early recurrence and build a nomogram. RESULTS: Univariate analysis showed that the number of tumors (odds ratio [OR], 0.258; 95% confidence interval [CI]: 0.104, 0.639; P = 0.003) and peritumoral edema (OR, 0.965; 95% CI: 0.942, 0.988; P = 0.003; mean in the recurrence group = 22.04 ± 17.21 mm; mean in the nonrecurrence group = 14.22 ± 12.84 mm) were statistically significantly different in patients with early recurrence. Genetic factors associated with early recurrence included IDH1 (OR, 4.405; 95% CI: 1.874, 10.353; P = 0.001) and O6-methylguanine-DNA methyltransferase (OR, 2.389; 95% CI: 1.234, 4.628; P = 0.010). Multivariate logistic regression analysis revealed that the number of tumors (OR, 0.227; 95% CI: 0.084, 0.616; P = 0.004), peritumoral edema (OR, 0.969; 95% CI: 0.945, 0.993; P = 0.013), and IDH1 (OR, 4.200; 95% CI: 1.602, 10.013; P = 0.004) were independent risk factors for early recurrence. The nomogram showed the highest net benefit when the threshold probability was less than 60%. CONCLUSIONS: A nomogram prediction model can effectively aid in clinical treatment decisions for patients with newly diagnosed HGG.

Treatment Response and Prognosis Evaluation in High-Grade Glioma: An Imaging Review Based on MRI.

In recent years, the development of advanced magnetic resonance imaging (MRI) technology and machine learning (ML) have created new tools for evaluating treatment response and prognosis of patients with high-grade gliomas (HGG); however, patient prognosis has not improved significantly. This is mainly due to the heterogeneity between and within HGG tumors, resulting in standard treatment methods not benefitting all patients. Moreover, the survival of patients with HGG is not only related to tumor cells, but also to noncancer cells in the tumor microenvironment (TME). Therefore, during preoperative diagnosis and follow-up treatment of patients with HGG, noninvasive imaging markers are needed to characterize intratumoral heterogeneity, and then to evaluate treatment response and predict prognosis, timeously adjust treatment strategies, and achieve individualized diagnosis and treatment. In this review, we summarize the research progress of conventional MRI, advanced MRI technology, and ML in evaluation of treatment response and prognosis of patients with HGG. We further discuss the significance of the TME in the prognosis of HGG patients, associate imaging features with the TME, indirectly reflecting the heterogeneity within the tumor, and shifting treatment strategies from tumor cells alone to systemic therapy of the TME, which may be a major development direction in the future. LEVEL OF EVIDENCE: 5 TECHNICAL EFFICACY STAGE: 4.

Predicting Masaoka-Koga Clinical Stage of Thymic Epithelial Tumors Using Preoperative Spectral Computed Tomography Imaging.

OBJECTIVES: To investigate the utility of spectral computed tomography (CT) parameters for the prediction of the preoperative Masaoka-Koga stage of thymic epithelial tumors (TETs). MATERIALS AND METHODS: Fifty-four patients with TETs, aged from 37 to 73 years old, an average age of 55.56 ± 9.79 years, were included in the study.According to the Masaoka-Koga staging method, there were 19 cases of stage I, 15 cases of stage II, 8 cases of stage III, and 12 cases of stage IV disease. All patients underwent dual-phase enhanced energy spectral CT scans. Regions of interest (ROIs) were defined in sections of the lesion with homogeneous density, the thoracic aorta at the same level as the lesion, the outer fat layer of the lesion, and the anterior chest wall fat layer. The single-energy CT value at 40-140 keV, iodine concentration, and energy spectrum curve of all lesion and thoracic aorta were obtained. The energy spectrum CT parameters of the lesions, extracapsular fat of the lesions, and anterior chest wall fat in stage I and stage II were obtained. The energy spectrum CT parameters of the lesions, enlarged lymph nodes and intravascular emboli in the 3 groups were obtained. The slope of the energy spectrum curve and the normalized iodine concentration were calculated. RESULTS: In stage I lesions, there was a statistically significant difference between the slope of the energy spectrum curve for the lesion and those of the fat outside the lesion and the anterior chest wall in the arteriovenous phase (P<0.001, P<0.001). The energy spectrum curve of the tumor parenchyma was the opposite of that of the extracapsular fat. In stage II lesions, there was a statistically significant difference between the slope of the energy spectrum curve for the anterior chest wall and those of the lesion and the fat outside the lesion in the arteriovenous phase(P<0.001, P<0.001). The energy spectrum curve of the tumor parenchyma was consistent with that of the extracapsular fat. Distinction between stage I and II tumors be evaluated by comparing the energy spectrum curves of the mass and the extracapsular fat of the mass. The accuracy rate of is 79.4%. For stages III and IV, there was no significant difference in the slope of the energy spectrum curve of the tumor parenchyma, metastatic lymph node, and intravascular embolism (P>0.05). The energy spectrum curve of the tumor parenchyma was consistent with that of the enlarged lymph nodes and intravascular emboli. The two radiologists have strong consistency in evaluating TETs Masaoka-Koga staging, The Kappa coefficient is 0.873,(95%CI:0.768-0.978). CONCLUSION: Spectral CT parameters, especially the energy spectrum curve and slope, are valuable for preoperative TET and can be used in preoperative staging prediction.

Evaluation of the anti-angiogenic effect of bevacizumab on rat C6 glioma by spectral computed tomography.

BACKGROUND: Anti-angiogenic drugs have become a research hotspot in recent years. However, dynamically observing their therapeutic effect at different time points during treatment is a clinical problem. PURPOSE: To explore the feasibility of the quantitative parameters of spectral computed tomography (CT) in evaluating the anti-angiogenic effect of bevacizumab on rat C6 glioma. MATERIAL AND METHODS: Twenty-six male Sprague-Dawley rats were used to establish the C6 glioma model. The rats were randomly divided into the experimental group (n = 13) and control group (n = 13). The experimental group was intraperitoneally injected with 0.2 µL/g bevacizumab every day, whereas the control group was injected with the same dose of normal saline every day for one week. Spectral CT scanning was performed on the 4th and 8th days after treatment; meanwhile, the brain tissues were collected by heart perfusion for H&E staining, and VEGF and HIF-1α immunohistochemical staining. RESULTS: On the 4th and 8th days, significant differences in the 70-keV single-energy CT value, slope of the energy spectrum curve, and iodine concentration were found between the experimental group and the control group. Correlation analysis between immunohistochemistry and quantitative parameters of spectral CT showed that the single energy CT value of 70 keV, slope of the energy spectrum curve, and concentration of iodine were positively correlated with VEGF and HIF-1α at different time points in the experimental group and the control group. CONCLUSION: Spectral CT multi-parameter imaging can be employed as a new method to evaluate the anti-angiogenic effect of bevacizumab on rat C6 glioma.

Ectopic thymoma: Retrospective analysis of eight cases with clinical features and computed tomography findings.

OBJECTIVE: To analyze the computed tomography (CT) and clinical features of ectopic thymoma, and to be familiar with the CT diagnosis of this disease. MATERIALS AND METHODS: CT data, clinical data, and pathological data of eight cases of ectopic thymoma, confirmed by pathology from September 2013 to June 2019, were retrospectively analyzed. RESULTS: Eight cases of thymoma were diagnosed, which included three in mediastinum (one of B1 type, two of C type), two in pericardium (both of B3 type), one in lung (B1 type), one in pleura (AB type), and one in right atrium (B2 type). Among the eight cases, four were men and four were women, aged 36-70 years. The clinical manifestations were chest tightness, shortness of breath and cough, and one case of myasthenia gravis. Six of the 8 patients were misdiagnosed as lymphoma, solitary fibrous tumor, malignant teratoma by CT. CT showed the following: the long diameter of tumor was 4.2 cm-19.5 cm, the shape was elliptical or round, and one case of ectopic thymoma grew in the atrium. The density was homogeneous in two cases and heterogeneous in six cases. The boundary was clear in three cases and unclear in five cases. Among the eight cases, three showed pleural effusion, two showed pericardial effusion and three showed calcification. CONCLUSIONS: Ectopic thymoma is rare and often misdiagnosed due to abnormal position. However, CT findings of ectopic thymoma are similar to those of the anterior superior mediastinal thymoma.

Development and validation of a CT-based radiomic nomogram for preoperative prediction of early recurrence in advanced gastric cancer.

BACKGROUND: In the clinical management of advanced gastric cancer (AGC), preoperative identification of early recurrence after curative resection is essential. Thus, we aimed to create a CT-based radiomic model to predict early recurrence in AGC patients preoperatively. MATERIALS AND METHODS: We enrolled 669 consecutive patients (302 in the training set, 219 in the internal test set and 148 in the external test set) with clinicopathologically confirmed AGC from two centers. Radiomic features were extracted from preoperative diagnostic CT images. Machine learning methods were applied to shrink feature size and build a predictive radiomic signature. We incorporated the radiomic signature and clinical risk factors into a nomogram using multivariable logistic regression analysis. The area under the curve (AUC) of operating characteristics (ROC), accuracy, and calibration curves were assessed to evaluate the nomogram's performance in discriminating early recurrence. RESULTS: A radiomic signature, including three hand crafted features and six deep learning features, was significantly associated with early recurrence (p-value <0.0001 for all sets). In addition, clinical N stage, carbohydrate antigen 199 levels, carcinoembryonic antigen levels, and Borrmann type were considered useful predictors for early recurrence. The nomogram, combining all these predictors, showed powerful prognostic ability in the training set and two test sets with AUCs of 0.831 (95% CI, 0.786-0.876), 0.826 (0.772-0.880) and 0.806 (0.732-0.881), respectively. The predicted risk yielded good agreement with the observed recurrence probability. CONCLUSIONS: By incorporating a radiomic signature and clinical risk factors, we created a radiomic nomogram to predict early recurrence in patients with AGC, preoperatively, which may serve as a potential tool to guide personalized treatment.