Predicting response and survival to first-line treatment with baseline [

Background: Small-cell lung cancer (SCLC) is a highly malignant disease with a propensity for early progression and high mortality. The prognostic value of treatment response and survival has been verified for solely established imaging, clinical, and biochemical markers. There is a lack of evidence for the combination of those parameters with machine learning and integrated models, particularly in the context of molecular imaging. Objectives: The aim of this study was to predict early disease progression and survival using CT-based radiomic features (RF), integrating [ Design: This retrospective pilot study included 62 patients with non-metastatic and metastatic SCLC who underwent stage-based primary treatment following baseline [ Methods: A radiomics signature was generated based on the first-line treatment response by RECIST 1.1 criteria. The RF was integrated using binary logistic regression analysis with the PET parameter metabolic tumor volume (MTV) of the primary tumor and initial disease stage. The integrated model with the highest AUC for predicting early disease progression was evaluated for predicting progression-free survival (PFS) and overall survival (OS) in both non-metastatic and metastatic patients. Results: A single CT-based RF demonstrated predictive capacity (AUC = 0.81). Integration of the MTV and disease stage enhanced the predictive capacity (AUC = 0.9). A Youden index-based threshold of <0.62 was identified as a significant predictor of prolonged PFS: non-metastatic disease with a median PFS of 25 versus 4 months (HR = 0.072; Conclusion: A multiparametric approach based on a Radiomics model may potentially be capable of identifying patients at risk for early disease progression, PFS, and OS in non-metastatic and metastatic SCLC.