ESTRO 2025 - Abstract Book

S3921

Radiobiology - Normal tissue radiobiology

ESTRO 2025

2245

Proffered Paper Multi-Omics Analysis on the Pituitary and Thyroid Gland on Predicting the Radiation-Induced Hypothyroidism for NPC patients Jiaming WU 1 , Jiang ZHANG 1 , Xinzhi TENG 1 , Xinyu ZHANG 1 , Jiachen SUN 1 , Yuanpeng ZHANG 1 , Tian LI 1 , Jing CAI 1,2,3 , Victor H.F. LEE 4 , Kenneth C.W. LEE 5 , Kwok Hung 6 , Ka Man CHEUNG 6 , James C.H. CHOW 6 , Jian Zang 7 1 Department of Health Technology and Informatics, Hong Kong Polytechnic University, Hong Kong, China. 2 Research Institute for Smart Ageing, Hong Kong Polytechnic University, Hong Kong, China. 3 Hong Kong Polytechnic University Shenzhen Research Institute, Hong Kong Polytechnic University, Shenzhen, China. 4 Department of Clinical Oncology, Hong Kong University, Hong Kong, China. 5 Department of Clinical Oncology, Prince of Wales Hospital, Hong Kong, China. 6 Department of Clinical Oncology, Queen Elizabeth Hospital, Hong Kong, China. 7 Department of Radiation Oncology, Xijing Hospital, Fourth Military Medical University, Xi An, China Purpose/Objective: Hypothyroidism is a common side effect of radiotherapy for nasopharyngeal carcinoma(NPC) patients [1]. An early event is characterized by elevated thyroid-stimulating hormone(TSH) levels, known as subclinical hypothyroidism(sHT). Previous studies have focused on the thyroid gland itself. However, in addition to thyroid, pituitary gland plays a critical role in the feedback loop of thyroid hormones, whereas its role in sHT has rarely been explored, only research on the situation decreased TSH levels. This study aims to investigate the contributions of the anatomy and dose deposition of both thyroid and pituitary to sHT by radiomics and dosiomics modeling. Material/Methods: 332 patients from Queen Elizabeth Hospital and United Christian Hospital were included. Contrast-enhanced CT images, planning dose distributions, and clinical records and dose-volume histogram (DVH) were collected retrospectively. sHT was defined as elevated TSH levels(>4.2 mIU/L). Radiomics and dosiomics features were extracted from CT and dose maps within thyroid and pituitary by pyradiomics in Python, and workflow is shown in Figure.1. Highly repeatable features were selected using a perturbation method, followed by mRMR and LASSO for feature selection. Models including Logistic Regression, Random Forest and xGBoost were trained and compared using different combinations of six feature groups: thyroid radiomics(TR), thyroid dosiomics(TD), pituitary radiomics(PR), pituitary dosiomics(PD), clinical factors, and DVH. Modalities were combined in various ways to assess prediction. T-tests evaluated clinical factors, the AUC was used to evaluate performance.