Development of an 18F-fluorodeoxyglucose positron emission tomography/computed tomography-based visceral pleural invasion prediction index in peripheral subpleural non-small cell lung cancer ≤3 cm.

Background: Accurate prediction of the visceral pleural invasion (VPI) status in stage I peripheral non-small cell lung cancers (NSCLCs) may help patients benefit from surgery and chemotherapy strategies. Fluorodeoxyglucose positron emission tomography/computed tomography (FDG PET/CT) is a combination of CT and FDG PET that reflects not only the FDG PET-associated metabolic features but also the CT-associated anatomic features. The FDG PET/CT radiomics-based model has demonstrated greater accuracy in the diagnosis, prognosis, and treatment evaluation of tumors compared with those based on CT only, including in NSCLCs. This study aimed to develop an FDG PET/CT-based VPI prediction (VPIP) index for peripheral subpleural NSCLCs ≤3 cm to help design a more rational treatment strategy for small-size NSCLCs. Methods: This retrospective study included 2,383 patients with peripheral subpleural NSCLCs ≤3 cm between August 2016 and February 2022 diagnosed at Shanghai Chest Hospital. The patients were randomly assigned to the modeling and validation cohorts in a 3:1 ratio. Further, 1,691 PET and 1,691 CT radiomics features of tumors were extracted using the Multi-Modality Radiomics software on the syngo.via workstation. The t-test combined with least absolute shrinkage and selection operator regression was used for feature selection. The regression coefficients were used to formulate the VPIP index. The predictive performance of the VPIP index in both cohorts was assessed using the area under the receiver operating characteristic curve (AUC) and compared using the DeLong test. Results: Eventually, 6 PET and 10 CT features were selected to formulate the FDG PET/CT-based VPIP index. The predictive accuracy of the PET/CT-based VPIP index, with a threshold >-1.13 for VPI, was 83.4% and 83.7% in the modeling and validation cohorts, respectively. This accuracy was substantially higher than that based on PET alone (formulated by 15 PET radiomics features) or CT alone (formulated by 12 CT radiomics features) in both the modeling (AUC, 0.926 vs. 0.906 vs. 0.921) and validation cohorts (AUC, 0.918 vs. 0.892 vs. 0.905) (P=0.00-0.01). Conclusions: The VPIP index formulated in this study based on FDG PET/CT in peripheral subpleural NSCLCs ≤3 cm may help develop more rational surgery strategies for small-size NSCLCs.