AUTHOR=Wang Lijie , Liu Ailing , Wang Zhiheng , Xu Ning , Zhou Dandan , Qu Tao , Liu Guiyuan , Wang Jingtao , Yang Fujun , Guo Xiaolei , Chi Weiwei , Xue Fuzhong 

TITLE=A Prognostic Model of Non-Small Cell Lung Cancer With a Radiomics Nomogram in an Eastern Chinese Population

JOURNAL=Frontiers in Oncology

VOLUME=Volume 12 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2022.816766

DOI=10.3389/fonc.2022.816766

ISSN=2234-943X

ABSTRACT=Background: The aim of this study was to build and validate a radiomics nomogram by integrating the radiomics features extracted from the CT images and known clinical variables (TNM staging et al) for individually predicting overall survival (OS) of patients with non-small cell lung cancer (NSCLC).
Methods: 1480 patients with clinical data and pretreatment CT images during January 2013 and May 2018 were enrolled in this study. We randomly assigned the patients into training (N = 1036) and validation cohorts (N = 444). We extracted 1288 quantitative features from the CT images of each patient. The Least Absolute Shrinkage and Selection Operator (LASSO) Cox regression model was applied in feature selection and radiomics signature building. The radiomics nomogram used for the prognosis prediction was built by combining the radiomics signature and clinical variables which derived from clinical data. Calibration ability, and discrimination ability were analyzed in both training and validation cohort.
Results: Eleven radiomics features were selected by LASSO Cox regression derived from CT images and the radiomics signature was built in the training cohort. The radiomics signature was significantly associated with NSCLC patients’ OS (HR=3.913, P < 0.01). The radiomics nomogram combining the radiomics signature with six clinical variables (age, sex, chronic obstructive pulmonary disease, T stage, N stage, M stage) had better prognostic performance than that of the clinical nomogram both in the training cohort (C-index, 0.861, 95% CI: 0.843-0.879) vs. (C-index, 0.851, 95% CI: 0.832-0.870); P < 0.001) and the validation cohort (C-index, 0.868, 95% CI: 0.841-0.896) vs. (C-index, 0.854, 95% CI: 0.824-0.884); P = 0.002). The calibration curves demonstrated optimal alignment between the prediction and actual observation.
Conclusion: The established radiomics nomogram could act as a noninvasive prediction tool for individualized survival prognosis estimation in patients with NSCLC. The radiomics signature derived from CT images may help clinicians in decision making and hold promise to be adopted in the patient care setting as well as the clinical trial setting.