Click here to


Are you sure ?

Yes, do it No, cancel

Prediction of Acute Xerostomia Based On Delta Radiomics From CT Images During Radiation Therapy for Nasopharyngeal Cancer

Yanxia LIU1*, Hongyu SHI1, Sijuan Huang2, Xiaochuan CHEN1, Huimin ZHOU2,3, Hui CHANG2, Yunfei XIA2, Guohua WANG1, Xin Yang2. (1) School of Software Engineering, South China University of Technology, Guangzhou, Guangdong, 510006, China. (2) Sun Yat-sen University Cancer Center; State Key Laboratory of Oncology in South China; Collaborative Innovation Center for Cancer Medicine; Guangdong Key Laboratory of Nasopharyngeal Carcinoma Diagnosis and Therapy, Guangzhou, Guangdong, 510060, China. (3) Department of Oncology, the Seventy-fourth Group Army Hospital of the Chinese People's Liberation Army, Guangzhou, Guangdong, 510318, China.


(Monday, 7/15/2019) 1:15 PM - 1:45 PM

Room: Exhibit Hall | Forum 2

Purpose: To investigate radiation-induced changes of CT radiomics in parotid glands (PG) and salivary amount (SA), in predicting acute xerostomia during radiotherapy (RT) for Nasopharyngeal Cancer (NPC).

Methods: CT and outcome data for 35 patients with stage I–IVB, randomly from the NPC clinical trial registered on the (ID: NCT01762514), were involved prospectively. All patients received radical treatment based on IMRT with a prescription dose of 68.1 Gy in 30 fractions. The patients ages ranged 24–72y and each patient had five CT sets acquired at treatment position at 0ᵗʰ, 10ᵗʰ, 20ᵗʰ, 30ᵗʰ fractions during RT, and at 3-month after RT. The PGs were delineated on each CT set by a radiation oncologist and verified independently by another one. Patients salivary were also collected every other 10-days during RT. Acute xerostomia was evaluated based on RTOG acute toxicity scoring and SA. 1703 radiomics features were calculated for PG from each CT set, including feature value at 0ᵗʰ fraction (FV₀), FV�₀ and delta FV (ΔFV�₀₋₀) respectively. Extensive experiments were conducted and finally RidgeCV and Recursive Feature Elimination (RFE) were used for feature selection, while Linear Regression was used for SA₃₀ prediction.

Results: Substantial changes in various radiomics metrics of PG during RT were observed. 14 normalized FV (NFV), selected from △NFV�₀₋₀, NFV₀ and NFV�₀, could have the best predictive ability with a Mean Squared Error (MSE) of 0.0569. The model predicts the level of acute xerostomia with a precision of 92.2% and a sensitivity of 100%, as compared to the clinical observed SA.

Conclusion: This study demonstrates that radiation-induced acute xerostomia can be early predicted based on the radiomics changes of the PG during RT. SA, NFV₀, NFV�₀ and especially △NFV�₀₋₀ could have the best predictive ability of acute xerostomia level for individual patient based on the delta radiomics.

Funding Support, Disclosures, and Conflict of Interest: Student's Platform for Innovation and Entrepreneurship Training Program - 201813902075; 201813902071; 201713902050; Hubei Key Laboratory of Medical Information Analysis and Tumor Diagnosis and Treatment - PJS140011504; Natural Science Foundation of Guangdong Province - 2017A030310217; Pearl River S&T Nova Program of Guangzhou - 201710010162


CT, Image Analysis, Feature Selection


IM/TH- Image Analysis (Single modality or Multi-modality): Computer-aided decision support systems (detection, diagnosis, risk prediction, staging, treatment response assessment/monitoring, prognosis prediction)

Contact Email