Dosiomics-based prediction of the occurrence of bone marrow suppression during radiotherapy for esophageal cancer
10.3760/cma.j.cn113030-20241015-00399
- VernacularTitle:基于剂量组学预测食管癌放疗期间发生骨髓抑制的研究
- Author:
Yilin LIU
1
;
Yanchun TANG
;
Ziyue SUN
;
Jinkai LI
;
Yaru PANG
;
Xinchen SUN
Author Information
1. 南京医科大学第一附属医院，南京210000
- Publication Type:Journal Article
- Keywords:
Esophageal neoplasms;
Dosiomics;
Bone marrow suppression;
Machine learning
- From:
Chinese Journal of Radiation Oncology
2025;34(7):684-691
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To study the risk factors and dosiomics-based prediction model of bone marrow suppression in patients with esophageal cancer during radiotherapy.Methods:Clinic data and radiotherapy planning documents of 107 patients with oesophageal cancer who underwent radiotherapy at the First Affiliated Hospital of Nanjing Medical University from January 2021 to May 2024 were retrospectively analyzed. Blood test results before and during radiotherapy were collected, and patients were classified into myelosuppressive groups (≤grade 1 and ≥grade 2). Clinical features, traditional dosimetric features and dosiomics features were collected, respectively. According to the stratified randomization grouping method, all patients were divided into the training and test sets in a 7 vs. 3 ratio. The region of interest was obtained by automatically outlining the thoracic skeleton (including the sternum, thoracic vertebrae and ribs) by AccuContour software. Dosiomics features were extracted from the dose distribution of the thoracic skeleton, and these features were screened using the independent samples t-test, the muse selector and the least absolute shrinkage operator. Subsequently, the dosiomic scores were calculated. Statistically significant clinical characteristics were screened using univariate and multivariate logistic regression analyses. Support vector machine method was used to construct a clinical model and a clinical combined with dosiomic model. Subsequently, nomogram was drawn for clinical prediction. The clinical efficacy and clinical benefit of predictive model were 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:Thirteen dosiomic features associated with bone marrow suppression were screened. Based on both univariate and multivariate logistic regression analyses, simultaneous chemotherapy, V 35 Gy and the average dose to bone were identified as statistically significant clinical predictors (all P<0.05). The AUC values of the combined model in the training and test sets were 0.800 and 0.776, higher than 0.709 and 0.650 of the clinical model. The calibration curves showed good agreement between the predicted and actual probabilities of the combined model. The DCA results showed that the net clinical benefit of the combined model was higher than that of the clinical model. Conclusions:The combined dosiomics-based model is effective in improving the predictive performance of bone marrow suppression occurring after radiotherapy for esophageal cancer.
