Preoperative prediction of malignant potential of 2-5 cm gastric gastrointestinal stromal tumors by computerized tomography-based radiomics

Preoperative CT-based radiomics and deep learning model for predicting risk stratification of gastric gastrointestinal stromal tumors

BACKGROUND

Gastrointestinal stromal tumors (GISTs) are clinically heterogeneous with various malignant potential in different individuals. It is crucial to explore a reliable method for preoperative risk stratification of gastric GISTs noninvasively.

PURPOSE

To establish and evaluate a machine learning model using the combination of computed tomography (CT) morphology, radiomics, and deep learning features to predict the risk stratification of primary gastric GISTs preoperatively.

METHODS

The 193 gastric GISTs lesions were randomly divided into training set, validation set, and test set in a ratio of 6:2:2. The qualitative and quantitative CT morphological features were assessed by two radiologists. The tumors were segmented manually, and then radiomic features were extracted using PyRadiomics and the deep learning features were extracted using pre-trained Resnet50 from arterial phase and venous phase CT images, respectively. Pearson correlation analysis and recursive feature elimination were used for feature selection. Support vector machines were employed to build a classifier for predicting the risk stratification of GISTs. This study compared the performance of models using different pre-trained convolutional neural networks (CNNs) to extract deep features for classification, as well as the performance of modeling features from single-phase and dual-phase images. The arterial phase, venous phase and dual-phase machine learning models were built, respectively, and the morphological features were added to the dual-phase machine learning model to construct a combined model. Receiver operating characteristic (ROC) curves were used to evaluate the efficacy of each model. The clinical application value of the combined model was determined through the decision curve analysis (DCA) and the net reclassification index (NRI) was analyzed.

RESULTS

The area under the curve (AUC) of the dual-phase machine learning model was 0.876, which was higher than that of the arterial phase model or venous phase model (0.813, 0.838, respectively). The combined model had best predictive performance than the above models with an AUC of 0.941 (95% CI: 0.887-0.974) (p = 0.012, Delong test). DCA demonstrated that the combined model had good clinical application value with an NRI of 0.575 (95% CI: 0.357-0.891).

CONCLUSION

In this study, we established a combined model that incorporated dual-phase morphology, radiomics, and deep learning characteristics, which can be used to predict the preoperative risk stratification of gastric GISTs.

Prediction of High-Risk Gastrointestinal Stromal Tumor Recurrence Based on Delta-CT Radiomics Modeling: A 3-Year Follow-up Study After Surgery

Background

Medium- to high-risk classification-gastrointestinal stromal tumors (MH-GIST) have a high recurrence rate and are difficult to treat. This study aims to predict the recurrence of MH-GIST within 3 years after surgery based on clinical data and preoperative Delta-CT Radiomics modeling.

Methods

A retrospective analysis was conducted on clinical imaging data of 242 cases confirmed to have MH-GIST after surgery, including 92 cases of recurrence and 150 cases of normal. The training set and test set were established using a 7:3 ratio and time cutoff point. In the training set, multiple prediction models were established based on clinical data of MH-GIST and the changes in radiomics texture of enhanced computed tomography (CT) at different time periods (Delta-CT radiomics). The area under curve (AUC) values of each model were compared using the Delong test, and the clinical net benefit of the model was tested using decision curve analysis (DCA). Then, the model was externally validated in the test set, and a novel nomogram predicting the recurrence of MH-GIST was finally created.

Results

Univariate analysis confirmed that tumor volume, tumor location, neutrophil-lymphocyte ratio (NLR), platelet lymphocyte ratio (PLR), diabetes, spicy hot pot, CT enhancement mode, and Radscore 1/2 were predictive factors for MH-GIST recurrence (P < .05). The combined model based on these above factors had significantly higher predictive performance (AUC = 0.895, 95% confidence interval [CI] = [0.839-0.937]) than the clinical data model (AUC = 0.735, 95% CI = [0.6 62-0.800]) and radiomics model (AUC = 0.842, 95% CI = [0.779-0.894]). Decision curve analysis also confirmed the higher clinical net benefit of the combined model, and the same results were validated in the test set. The novel nomogram developed based on the combined model helps predict the recurrence of MH-GIST.

Conclusions

The nomogram of clinical and Delta-CT radiomics has important clinical value in predicting the recurrence of MH-GIST, providing reliable data reference for its diagnosis, treatment, and clinical decision-making.

Development and validation of a preoperative risk nomogram prediction model for gastric gastrointestinal stromal tumors

BACKGROUND AND STUDY AIMS

Gastrointestinal stromal tumors (GIST) carry a potential risk of malignancy, and the treatment of GIST varies for different risk levels. However, there is no systematic preoperative assessment protocol to predict the malignant potential of GIST. The aim of this study was to develop a reliable and clinically applicable preoperative nomogram prediction model to predict the malignant potential of gastric GIST.

PATIENTS AND METHODS

Patients with a pathological diagnosis of gastric GIST from January 2015 to December 2021 were screened retrospectively. Univariate and multivariate logistic analyses were used to identify independent risk factors for gastric GIST with high malignancy potential. Based on these independent risk factors, a nomogram model predicting the malignant potential of gastric GIST was developed and the model was validated in the validation group.

RESULTS

A total of 494 gastric GIST patients were included in this study and allocated to a development group (n = 345) and a validation group (n = 149). In the development group, multivariate logistic regression analysis revealed that tumor size, tumor ulceration, CT growth pattern and monocyte-to- lymphocyte ratio (MLR) were independent risk factors for gastric GIST with high malignancy potential. The AUC of the model were 0.932 (95% CI 0.890-0.974) and 0.922 (95% CI 0.868-0.977) in the development and validation groups, respectively. The best cutoff value for the development group was 0.184, and the sensitivity and specificity at this value were 0.895 and 0.875, respectively. The calibration curves indicated good agreement between predicted and actual observed outcomes, while the DCA indicated that the nomogram model had clinical application.

CONCLUSIONS

Tumor size, tumor ulceration, CT growth pattern and MLR are independent risk factors for high malignancy potential gastric GIST, and a nomogram model developed based on these factors has a high ability to predict the malignant potential of gastric GIST.

Machine learning-based predictive model for the differential diagnosis of ≤ 5 cm gastric stromal tumor and gastric schwannoma based on CT images

The clinical symptoms of ≤ 5 cm gastric stromal tumor (GST) and gastric schwannoma (GS) are similar, but the treatment regimens are different. This study explored the value of computed tomography (CT) combined with machine learning (ML) algorithms to find the best model to discriminate them. A total of 126 patients with GST ≤ 5 cm and 35 patients with GS ≤ 5 during 2013-2022 were included. CT imaging features included qualitative data (tumor location, growth pattern, lobulation, surface ulcer status, necrosis, calcification, and surrounding lymph nodes) and quantitative data [long diameter (LD); short diameter (SD); LD/SD ratio; degree of enhancement (DE); heterogeneous degree (HD)]. Patients were randomly divided into a training set (n = 112) and test set (n = 49) using 7:3 stratified sampling. The univariate and multivariate logistic regression analysis were used to identify independent risk factors. Five ML algorithms were used to build prediction models: Support Vector Machine, k-Nearest Neighbor, Random Forest, Extra Trees, and Extreme Gradient Boosting Machine. The analysis identified that HDv, lobulation, and tumor growth site were independent risk factors (P < 0.05). We should focus on these three imaging features of tumors, which are relatively easy to obtain. The area under the curve for the SVM, KNN, RF, ET, and XGBoost prediction models were, respectively, 0.790, 0.895, 0.978, 0.988, and 0.946 for the training set, and were, respectively, 0.848, 0.892, 0.887, 0.912, and 0.867 for the test set. The CT combined with ML algorithms generated predictive models to improve the differential diagnosis of ≤ 5 cm GST and GS which has important clinical practical value. The Extra Trees algorithm resulted in the optimal model.