ESTRO meets Asia 2024 - Abstract Book

S160

Interdisciplinary – Head & neck

ESTRO meets Asia 2024

Conclusion:

By combining complex ML models with SHAP, the development of interpretable prediction models revealed the correlation between input features and specific sites of distant metastasis, and facilitated clinical decisions for personalized treatment.

Keywords: Machine learning, Distant metastasis

References:

1. Chen YP, Chan ATC, Le QT, Blanchard P, Sun Y, Ma J. Nasopharyngeal carcinoma. Lancet . 2019;394(10192):64 80. doi:10.1016/S0140-6736(19)30956-0

2. Aw L, Bb M, Wt N, At C. Management of Nasopharyngeal Carcinoma: Current Practice and Future Perspective. Journal of clinical oncology : official journal of the American Society of Clinical Oncology . 2015;33(29). doi:10.1200/JCO.2015.60.9347 3. X C, Y L, X L, et al. An interpretable machine learning prognostic system for locoregionally advanced nasopharyngeal carcinoma based on tumor burden features. Oral oncology . 2021;118. doi:10.1016/j.oraloncology.2021.105335

4. Lundberg SM, Lee SI. A unified approach to interpreting model predictions. Advances in neural information processing systems . 2017;30.

5. Alabi RO, Elmusrati M, Leivo I, Almangush A, Mäkitie AA. Machine learning explainability in nasopharyngeal cancer survival using LIME and SHAP. Sci Rep . 2023;13(1):8984. doi:10.1038/s41598-023-35795-0

6. L Z, D D, X F, et al. A deep learning-based radiomic nomogram for prognosis and treatment decision in advanced nasopharyngeal carcinoma: A multicentre study. EBioMedicine . 2021;70. doi:10.1016/j.ebiom.2021.103522

57

Proffered Paper

Response-adapted risk index model for prognosis and treatment selection in nasopharyngeal carcinoma Yang Liu, Junlin Yi

Radiation oncology, National Cancer Center/National Clinical Research Center for Cancer/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China

Purpose/Objective:

Dynamic response to therapy is strongly associated with outcomes for various malignancies. This study aims to develop and validate the model response-adapted individualized risk index (RAIRI) as an individualprognostic approach and predictive biomarker for adjuvant chemotherapy (AC) benefit in nonmetastatic nasopharyngeal carcinoma (NPC) based on pretreatment clinical characteristics, longitudinal plasma cell-free Epstein–Barr virus DNA (cfEBV-DNA), and tumor regression measurements collected during treatment.