ESTRO 2024 - Abstract Book

S5015

Physics - Radiomics, functional and biological imaging and outcome prediction

ESTRO 2024

J. Deck, M. et al. Effect of Lymphopenia on Tumor Response and Clinical Outcomes Following Chemoradiotherapy in Stage III Non-Small Cell Lung Cancer, Lung Cancer, 2023, doi: 10.2147/LCTT.S386344.

1116

Digital Poster

Prediction of PD-L1 expression rate in non-small cell lung cancer based on CT-radiomics

Lihong Xi 1 , Lingya Yang 1 , Xiaofeng He 1 , Jingling Huang 1 , Jun Li 1 , Daquan Zhong 1 , Haipeng Huang 2 , Jiabin Li 2 , Tiantian Zhai 1 1 the Cancer Hospital of Shantou University Medical College, Radiation oncology, Shantou, China. 2 Jieyang People's Hospital, Radiation oncology, Jieyang, China

Purpose/Objective:

Recently, immunotherapy has gradually covered the whole process of treatment for non-small cell lung cancer (NSCLC) patients. However, only a part of patients can benefit from immunotherapy. In clinical practice, the expression level of Programmed cell death-1 / Programmed cell death-ligand 1 (PD-1/PD-L1) is often used to screen the population that may benefit from immunotherapy. Numerous studies have shown that NSCLC patients with high PD-1/PD-L1 expression have a better immunotherapeutic response and correlate with better survival. This study aims to explore the efficacy of radiomics features on venous-phase enhanced CT in predicting PD-L1 expression in NSCLC patients and construct a prediction model to predict the PD-L1 expression level of NSCLC patients non-invasively, in real-time, and accurately, thus aiding in clinical treatment decision-making.

Material/Methods:

The study cohort was composed of NSCLC patients primarily treated between December 2018 and May 2023. All cancer patients have been pathologically confirmed by surgical resection or needle biopsy. The clinical parameters and pre-treatment enhanced CT images were collected for analysis. Univariable and multivariable logistic regression analyses were performed to identify the significant features and construct the clinical prediction model, radiomics prediction model, and combined prediction model. The nomogram was made to provide a convenient method for risk prediction.

Results:

One hundred and three patients were enrolled in this study. Clinical characteristics such as gender, age, ECOG score, pathological type, TNM stage and clinical stage were collected. Radiomics features were extracted from each pre treatment contrast-enhanced CT images by Pyradiomics, including 17 first-order histogram features, 22 morphological features and 109 texture features. Depending on the Tumor Cell Proportion Score (TPS) of PD-L1 expression, two different research endpoints were set (TPS < 1% vs. TPS ≥ 1%; TPS < 10% vs. TPS ≥ 10%), and predictive