Key points are not available for this paper at this time.
We aimed to explore the value of a PET/CT-based radiomics signature in predicting occult lymph node metastasis (OLM) and outcomes in clinical N0 (cN0) lung adenocarcinoma (LUAD), to uncover the biologic meaning of OLM-associated radiomics phenotypes and to validate the reproducibility of the identified radiomics-correlated key genes. A radiomics signature for OLM prediction was developed in a training cohort and validated across multiple validation and testing cohorts in this multicenter study. Prognostic implications of the radiomics score (Radscore) were assessed by measuring recurrence-free survival. Biologic processes and pathways were enriched and correlated with Radscore and each of the 8 radiomics phenotypes using paired PET/CT and RNA sequencing data. The reproducibility of identified radiomics-associated key genes was validated using a public database, clinical tissue samples, and in vitro experiments. Here we show OLM is detected in 127 (19.9%) of 637 patients. The proposed signature achieves AUCs of 0.82, 0.81, 0.78 and 0.79 in the training, internal validation, prospective testing and external testing cohort, respectively. In addition, Radscore is identified as an independent predictive factor in predicting the risk of recurrence of LUAD. Furthermore, Radscore and OLM-related radiomics features are mostly associated with immune response. Finally, four key genes (MIR600HG, FAM13A-AS1, AQP4 and GRIA1), especially the MIR600HG gene, play important roles in OLM and prognosis of early-stage LUAD. We demonstrate that radiogenomics performed on PET/CT images provides complementary clinical, prognostic and molecular information with great potential for the prediction of OLM and risk stratification in cN0 LUAD. Diagnosing patients with lung cancer metastasis (tumors that have left the primary site) can be challenging even with the most sensitive imaging tools available as they may be very small. Without metastasis, lung adenocarcinoma (LUAD) is called clinical N0 (cN0). Radiogenomics combines medical imaging data with genomic data and has gained a lot of attention due to its excellent performance in confirming image-to-molecular feature association. Therefore, in this study, we aimed to explore the value of a imaging-based radiomics signature in predicting lung metastasis and outcomes in cN0 LUAD patients. Our results indicated that imaging-based radiomics signature could help accurately predict metastasis and stratify prognosis in this patient population. Ouyang et al. use a radiogenomic analysis to predict occult lymph node metastasis (OLM) and survival in clinical N0 (cN0) lung adenocarcinoma (LUAD). The results suggest that radiogenomics provides complementary clinical, prognostic and molecular information with great potential for the prediction of OLM and risk stratification in cN0 LUAD.
Ouyang et al. (Wed,) studied this question.