Body composition radiomics integrated with machine learning to predict prognosis in advanced NSCLC treated with first-line immunochemotherapy and concurrent radiotherapy.
Authors
Affiliations (5)
Affiliations (5)
- Department of Medical Imaging, Shanghai East Hospital (East Hospital Affiliated with Tongji University), No. 150 Jimo Road, Pudong New Area, Shanghai, 200120, P. R. China.
- Department of Pain Management, Shanghai East Hospital, School of Medicine, Tongji University, Shanghai, 200120, China.
- School of Medical and Information Engineering, Gannan Medical University, Ganzhou, 341000, China.
- Department of Medical Oncology, Shanghai Pulmonary Hospital, School of Medicine, Tongji University Medical School Cancer Institute, Tongji University, Shanghai, 200433, China. [email protected].
- Department of Pain Management, Shanghai East Hospital, School of Medicine, Tongji University, Shanghai, 200120, China. [email protected].
Abstract
Adipose tissue is a highly heterogeneous and plastic endocrine and immune organ whose characteristics have been linked to prognosis in NSCLC patients receiving First-Line Immunochemotherapy and Concurrent Radiotherapy. This study aimed to investigate whether baseline body composition and radiomics features could serve as prognostic predictors in this patient population. This retrospective study involved the collection of data from 87 patients receiving chemotherapy in conjunction with immunotherapy. Radiomic features were extracted from skeletal muscle and adipose tissue at the L3 level of the lumbar spine. The most reliable radiomics features were selected to develop six machine learning classifier models. Patients were stratified into high-risk and low-risk groups based on the determination of an optimal threshold. Subsequently, TATI(Total adipose tissue index) and radiomics features were integrated to develop comprehensive predictive model, and the model's performance was assessed using ROC and DCA. The results indicated that the AUC for the combined model was 0.857 (95% CI: 0.765-0.949) in the train cohort and 0.834 (95% CI: 0.740-0.934) in the test cohort. Subsequently, the risk stratification analysis revealed that PFS was significantly shorter in high-risk patients compared to low-risk patients. Our model displayed enhanced predictive accuracy and greater net benefit. The research findings indicate that radiomics features derived from body composition hold potential for predicting prognosis in patients with non-small cell lung cancer. Our model may identify high-risk cohorts, thereby assisting in the identification of patients likely to benefit from concurrent radiotherapy through optimised treatment regimens.