Multiparametric MRI-based radiomics combined with deep learning for preoperative risk stratification of endometrial carcinoma
4 affiliations
- 1Department of Radiology, The First People’s Hospital of Qinzhou
- 2Department of Medical Imaging, Jinjiang Municipal Hospital
- 3Department of Radiology, Guigang City People’s Hospital
- 4Department of Radiology, Affiliated Hospital of Youjiang Medical University for Nationalities
Abstract
Objective: To develop and compare radiomics, deep learning, clinical, and combined models based on multiparametric MRI radiomics and 2.5D deep learning features for preoperative risk stratification of endometrial carcinoma (EC).
Methods: This retrospective study included 246 patients with pathologically confirmed EC, with preoperative clinical and imaging data collected. Clinical data, T2-weighted imaging (T2WI), and contrast-enhanced T1-weighted imaging (CE-T1WI) were collected. Radiomics and deep learning models were constructed based on T2WI, CE-T1WI, and their combination. A combined model integrating multiparametric radiomics features, deep learning features, and clinical parameters was subsequently developed. Model performance and clinical utility were evaluated using the area under the receiver operating characteristic curve (AUC), calibration curves, decision curve analysis (DCA), and the DeLong test for pairwise model comparisons. Gradient-weighted Class Activation Mapping (Grad-CAM) was employed to enhance model interpretability.
Results: Univariate and multivariate logistic regression analysis identified maximum tumor diameter, MR-DMI, BMI and ER status as independent predictive factors. Comparison of radiomics, deep learning, clinical, and combined models demonstrated that the multiparametric MRI model outperformed single-sequence models based on radiomics or deep learning, while the combined model exhibited the best predictive performance in the validation set, achieving an AUC of 0.891, accuracy of 0.86, sensitivity of 0.909, specificity of 0.765, and F1-score of 0.896. DeLong test results showed that, in the training set, the combined model had higher predictive performance than all other models, and the differences were statistically significant (P < 0.05). In the validation set, the combined model showed statistically significant improvements compared with the clinical model and the T2+CET1-Rad model (P<0.05). Decision curve analysis indicated that the combined model provided the highest clinical net benefit.
Conclusion: The combination of multiparametric MRI-based radiomics and 2.5D deep learning demonstrated favorable predictive performance in stratifying risk for EC, suggesting its potential utility as a non-invasive tool and offering a promising adjunctive approach to support personalized treatment planning and prognostic evaluation for EC patients in clinical practice.
Related
- Integrating Systemic Inflammation, Nutrition and Skeletal Muscle into a Composite Score (the HALP–Sarcopenia Combined Index) for Preoperative Risk Stratification of Lymph-Node Metastasis and Recurrence in Endometrial Carcinoma: A Single-Centre Retrospective Study
- Multiparametric MRI radiomics model for preoperative prediction of IDH1 R132H immunopositivity status in adult diffuse Glioma
- Multiparametric MRI-based habitat and peritumoral radiomics integrating semi-automated segmentation for preoperative prediction of high-grade prostate cancer: a dual-center study
- Multiparametric MRI radiomics and clinical features for preoperative prediction of high-grade meningioma: a two-centre retrospective study
- Multiparametric MRI radiomics combined with clinical-radiologic features improves preoperative prediction of muscle invasion in bladder cancer
BibTeX
@article{ouyang2026multiparametric,
title = {{Multiparametric MRI-based radiomics combined with deep learning for preoperative risk stratification of endometrial carcinoma}},
author = {Yanshan Ouyang and Mianlei Lin and Yiying Cao and Nanxuan He and Xiao Zhong and Yingning Wu},
journal = {Frontiers in Oncology},
year = {2026},
month = oct,
doi = {10.3389/fonc.2026.1942153},
url = {https://www.frontiersin.org/articles/10.3389/fonc.2026.1942153/full},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Multiparametric MRI-based radiomics combined with deep learning for preoperative risk stratification of endometrial carcinoma
AU - Yanshan Ouyang
AU - Mianlei Lin
AU - Yiying Cao
AU - Nanxuan He
AU - Xiao Zhong
AU - Yingning Wu
PY - 2026
DA - 2026/10/22
JO - Frontiers in Oncology
DO - 10.3389/fonc.2026.1942153
UR - https://www.frontiersin.org/articles/10.3389/fonc.2026.1942153/full
AB - Objective: To develop and compare radiomics, deep learning, clinical, and combined models based on multiparametric MRI radiomics and 2.5D deep learning features for preoperative risk stratification of endometrial carcinoma (EC). Methods: This retrospective study included 246 patients with pathologically confirmed EC, with preoperative clinical and imaging data collected. Clinical data, T2-weighted imaging (T2WI), and contrast-enhanced T1-weighted imaging (CE-T1WI) were collected. Radiomics and deep learning models were constructed based on T2WI, CE-T1WI, and their combination. A combined model integrating multiparametric radiomics features, deep learning features, and clinical parameters was subsequently developed. Model performance and clinical utility were evaluated using the area under the receiver operating characteristic curve (AUC), calibration curves, decision curve analysis (DCA), and the DeLong test for pairwise model comparisons. Gradient-weighted Class Activation Mapping (Grad-CAM) was employed to enhance model interpretability. Results: Univariate and multivariate logistic regression analysis identified maximum tumor diameter, MR-DMI, BMI and ER status as independent predictive factors. Comparison of radiomics, deep learning, clinical, and combined models demonstrated that the multiparametric MRI model outperformed single-sequence models based on radiomics or deep learning, while the combined model exhibited the best predictive performance in the validation set, achieving an AUC of 0.891, accuracy of 0.86, sensitivity of 0.909, specificity of 0.765, and F1-score of 0.896. DeLong test results showed that, in the training set, the combined model had higher predictive performance than all other models, and the differences were statistically significant (P < 0.05). In the validation set, the combined model showed statistically significant improvements compared with the clinical model and the T2+CET1-Rad model (P<0.05). Decision curve analysis indicated that the combined model provided the highest clinical net benefit. Conclusion: The combination of multiparametric MRI-based radiomics and 2.5D deep learning demonstrated favorable predictive performance in stratifying risk for EC, suggesting its potential utility as a non-invasive tool and offering a promising adjunctive approach to support personalized treatment planning and prognostic evaluation for EC patients in clinical practice.
SN - 2234-943X
ER -