Dosiomics-based prediction of the occurrence of bone marrow suppression in patients with pelvic tumors
10.3760/cma.j.cn113030-20230718-00007
- VernacularTitle:基于剂量组学预测盆腔部肿瘤患者发生骨髓抑制的研究
- Author:
Yanchun TANG
1
;
Jingyi TANG
;
Jinkai LI
;
Qin QIN
;
Hualing LI
;
Zhigang CHANG
;
Tianyu ZHANG
;
Yaru PANG
;
Xinchen SUN
Author Information
1. 南京医科大学第一附属医院，南京 210000
- Keywords:
Pelvic neoplasms;
Bone marrow suppression;
Dosiomics;
Machine learning
- From:
Chinese Journal of Radiation Oncology
2024;33(7):620-626
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To assess the predictive value of dosiomics in predicting the occurrence of bone marrow suppression (BMS) in patients with pelvic tumors during radiotherapy.Methods:A retrospective analysis was conducted on the clinical data and radiotherapy planning documents of 129 patients with pelvic region tumors who underwent radiotherapy at the First Affiliated Hospital of Nanjing Medical University from January 2019 to January 2023. The region of interest (ROI) was outlined for bone marrow in the pelvic region by Accu Contour software in planning CT, and the ROI was exported together with the dose distribution file. According to a stratified randomization grouping method, the patients were divided into the training set and test set in an 8 vs. 2 ratio. The dosiomic features were extracted from the ROI, and the two independent samples t-test and the least absolute shrinkage and selection operator (LASSO) algorithm was employed to identify the best predictive characteristics. Subsequently, the dosiomic scores were calculated. Clinical predictors were identified through both univariant and multivariate logistic regression analyses. Predictive models were constructed by using clinical predictors alone and combining clinical predictors and dosiomic scores. The efficacy of predictive model was assessed by plotting the receiver operating characteristic (ROC) curve and evaluating its performance through the area under the ROC curve (AUC), the calibration curve, and decision curve analysis (DCA). Results:Fourteen dosiomic features that showed a strong correlation with the occurrence of BMS were screened and utilized to calculate the dosiomic scores. Based on both univariant and multivariate logistic regression analyses, chemotherapy, planning target volume (PTV) and V 5 Gy were identified as clinical predictors. According to the combined model, the AUC values for the training set and test set were 0.911 and 0.868, surpassing those of the clinical model (AUC=0.878 and 0.824). Furthermore, the analysis of both the calibration curve and DCA suggested that the combined model had higher calibration and net clinical benefit. Conclusion:The combined model has a high diagnostic value for predicting BMS in patients with pelvic tumors during radiotherapy.
