Multiparametric MRI-based radiomics machine learning nomogram for predicting aggressive histology in endometrial cancer.
Authors
Affiliations (4)
Affiliations (4)
- Radiology Department, Fujian Maternity and Child Health Hospital, College of Clinical Medicine for Obstetrics and Gynecology and Pediatrics, Fujian Medical University, Fuzhou, China. [email protected].
- Department of Pathology, Fujian Maternity and Child Health Hospital, College of Clinical Medicine for Obstetrics and Gynecology and Pediatrics, Fujian Medical University, Fuzhou, China.
- Radiology Department, Fujian Maternity and Child Health Hospital, College of Clinical Medicine for Obstetrics and Gynecology and Pediatrics, Fujian Medical University, Fuzhou, China.
- Department of Radiation Oncology, Fujian Maternity and Child Health Hospital, College of Clinical Medicine for Obstetrics and Gynecology and Pediatrics, Fujian Medical University, Fuzhou, China. [email protected].
Abstract
To develop and validate a radiomics-based machine learning nomogram using multiparametric MRI for preoperative prediction of aggressive histology in endometrial cancer (EC) patients. This dual-center study retrospectively analyzed histologically confirmed EC patients who underwent preoperative MRI. Radiomics features were trained and tested to predict aggressive histology with a support vector machine (SVM) algorithm. Clinical data and conventional MRI findings were collected. A multivariable logistic regression analysis was conducted to create a predictive fusion model, which was displayed as a nomogram for the training set and validated on an independent external test set. Calibration curves and Hosmer-Lemeshow tests were used for goodness-of-fit evaluation. Three predictive models were constructed, namely M1 (original biopsy alone), M2 (radiomics alone), and M3 (combined nomogram). The model's performance was evaluated using ROC analysis, and pairwise comparisons of AUCs were conducted via DeLong's test. DCA was used for net benefit comparison. 283 women were enrolled (training: 198; test: 85). The M3 achieved AUCs of 0.900 (95% CI: 0.850-0.938) and 0.890 (95% CI: 0.803-0.948) for the training and test sets, respectively, demonstrating good fit according to Hosmer-Lemeshow tests (P > 0.05). Delong tests with Bonferroni correction indicated that the fusion model's AUCs of M3 surpassed those of M1in predicting aggressive histology (adjusted P < 0.05). Additionally, DCA demonstrated a higher net benefit for the M3 model, with IDIs of 0.126 and 0.176 (P < 0.01) in both sets. A multiparametric MRI-based radionics machine learning nomogram improves the preoperative diagnosis of aggressive histology in EC patients.