Abstract
Purpose
Preoperative prediction of the Lauren classification in gastric cancer (GC) has important clinical significance for improving the prognostic system and guiding personalized treatment. This study investigated the usefulness of deep learning radiomics analysis (DLRA) for preoperatively differentiating Lauren classification in patients with GC, using computed tomography (CT) images.
Methods
A total of 329 patients pathologically diagnosed with GC were recruited from August 2012 and December 2020. Patients (n = 262) recruited from August 2012 to July 2019 were randomly allocated into training cohort (n = 184) and internal validation cohort (n = 78), and patients recruited from August 2019 to December 2020 were included in external validation cohort (n = 67). Information on clinical characteristics were collected. Radiomics features were extracted from CT images at arterial phase (AP) and venous phase (VP). A radiomics nomogram incorporating the radiomics signature and clinical information was built for distinguishing Lauren classification, and its discrimination, calibration, and clinical usefulness were evaluated. Moreover, we also constructed a clinical model using the clinical factors only for baseline comparison.
Results
The nomogram incorporating the two radiomics signatures and clinical characteristics exhibited good discrimination of Lauren classification on all cohorts [overall C-indexes 0.771 (95% CI: 0.709–0.833) in the training cohort, 0.757 (95% CI: 0.698–0.807) in the internal validation cohort, 0.725 (95% CI: 0.655–0.793) in the external validation cohort]. Compared with the conventional clinical model, the deep learning hybrid radiomics nomogram (DHRN) exhibits enhanced predictive ability. Further, the calibration curve and decision curve substantiated the excellent fitness and clinical applicability of the model.
Conclusions
DLRA exhibited good performance in distinguishing Lauren classification in GC. In personalized treatment of GC, this preoperative nomogram could provide baseline information for optimizing the quality of clinical decision-making and therapeutic strategies.