The following is a summary of “Predictive value of delta radiomics in xerostomia after chemoradiotherapy in patients with stage III-IV nasopharyngeal carcinoma,” published in the February 2024 issue of Oncology by Wang et al.
Xerostomia, a prevalent complication following chemoradiotherapy in nasopharyngeal carcinoma (NPC) patients, presents significant clinical challenges. This study aimed to develop a Delta radiomics model utilizing magnetic resonance T1-weighted imaging (T1WI) sequences to predict xerostomia post-chemoradiotherapy and assess its diagnostic efficacy.
Retrospective data from 255 NPC patients with stage III-IV were collected, including clinical information and magnetic resonance imaging (MRI) data from pre-treatment and post-induction chemotherapy (IC). Xerostomia severity was graded within one week post-concurrent chemoradiotherapy (CCRT) as mild (92 cases) or severe (163 cases). Radiomics features were extracted from delineated parotid glands in T1WI sequences pre- and post-IC, and Delta radiomics features were calculated. Univariate logistic analysis, correlation analysis, and Gradient Boosting Decision Tree (GBDT) methods were employed for dimension reduction, optimal feature selection, and model establishment, respectively, for pretreatment, post-IC, and Delta radiomics xerostomia prediction models. Predictive efficacy was evaluated using receiver operating characteristic (ROC) curves and decision curves.
Optimal features were identified, with 15, 10, and 12 selected for pretreatment, post-IC, and Delta radiomics, respectively. The constructed models demonstrated promising performance, with area under the curve (AUC) values of 0.738, 0.751, and 0.843 in the training set, respectively. Clinical data analysis revealed age as the only statistically significant factor in both xerostomia severity groups (P < 0.05).
Delta radiomics emerges as a valuable tool for predicting post-chemoradiotherapy xerostomia severity in NPC patients, offering the potential for early clinical intervention strategies. These findings underscore the significance of integrating radiomics-based predictive models into clinical practice to enhance patient care and treatment planning.
Source: ro-journal.biomedcentral.com/articles/10.1186/s13014-024-02417-6