PET/CT radiomics and machine learning enable non-invasive survival stratification and histologic tumor risk profiling in patients with lung adenocarcinoma

Meixin Zhao,Kilian Kluge,Laszlo Papp,Marko Grahovac,Shaomin Yang,Chunting Jiang,Denis Krajnc,Clemens P. Spielvogel,Boglarka Ecsedi,Alexander Haug,Shiwei Wang,Marcus Hacker,Weifang Zhang,Xiang Li

DOI: https://doi.org/10.21203/rs.3.rs-771161/v1

2021-08-05

Abstract:Abstract Purpose Risk stratification in patients with lung adenocarcinoma (LUAD) is mandatory for treatment guiding and outcome prediction. Amongst clinical parameters including histological analyses, imaging procedures provide important information. The present study aimed to investigate the ability of machine learning models trained on clinical and 2-deoxy-2-[18F]fluoro-D-glucose ([ 18 F]FDG) positron emission tomography/computed tomography (PET/CT) derived radiomic data to predict overall survival (OS), tumor grade (TG), and histologic growth pattern risk (GPR) in treatment naïve LUAD patients. Methods 421 treatment naïve patients with histologically diagnosed lung adenocarcinoma and available [ 18 F]FDG PET/CT imaging were retrospectively analyzed. Four patient cohorts were generated based on the available data for predicting 4-year OS (n = 276), 3-year OS (n = 280), TG (n = 298) and GPR (n = 265). [ 18 F]FDG-positive lesions were delineated semiautomatically, from which 2082 radiomic features were extracted and combined with endpoint-specific clinical and demographic parameters. Machine learning models were built for the prediction of 4-year OS (M 4OS ), 3-year OS (M 3OS ), tumor grading (M TG ) and histologic growth pattern risk (M GPR ), respectively. Monte Carlo (MC) cross-validation with 100-folds and 80:20 training to validation split was employed as a performance evaluation for all models. Kaplan-Meier survival analysis was performed to assess the association between the M 4OS and M 3OS predictions with OS. Results Area under the receiver operator characteristics curve (AUC) was highest for M 4OS (AUC 0.88), followed by M 3OS (AUC 0.84), while M TG and M GPR performed equally well (AUC 0.76). Predictions of M 4OS (HR 0.128, p < 0.000001) and M 3OS (HR 0.0942, p < 0.000001) were independently associated with OS. Conclusion In our retrospective cohorts, machine learning models demonstrated the ability to prognosticate long-term survival outcomes in patients with lung adenocarcinoma. Furthermore, tumor lesions could be characterized according to their histologic grade and predominant growth pattern risk with high accuracy.