Preoperative Ki-67 proliferation index prediction with a radiomics nomogram in stage T1a-b lung adenocarcinoma.

OBJECTIVES

To establish a radiomics nomogram for preoperative prediction of Ki-67 proliferation index in stage T1a-b lung adenocarcinoma.

METHODS

A total of 206 patients with pathologically confirmed lung adenocarcinoma who underwent CT scans within 2 weeks preoperatively from January 2016 to June 2020 were retrospectively included. Ki-67 index ≤ 10% was considered low expression, and Ki-67 index > 10% was considered high expression. The primary cohort was randomized with a 7:3 ratio into a training cohort (n = 145) and a validation cohort (n = 61). The minimum redundancy maximum relevance (mRMR) and the least absolute shrinkage and selection operator (LASSO) were used for feature selection, and radiomics signature was constructed. Univariate and multivariate logistic regression analyses were used to identify clinically important risk factors and radiomics signature associated with Ki-67 proliferation index, which were then combined into radiomics nomogram.

RESULTS

Tumor maximum diameter (P = 0.005), lobulation (P = 0.002), absent of vacuole (P < 0.001), and Radscore (P < 0.001) were independent risk predictors of high Ki-67 proliferation index expression. The radiomics nomogram showed good predictive efficacy. The AUC, sensitivity, specificity and accuracy of radiomics nomogram in the training and validation cohorts were 0.91 (95% CI: 0.86-0.96), 87.9%, 80.5%, 83.4% and 0.85 (95% CI: 0.75-0.94), 71.9%, 82.8% and 77.0%. Decision curve analysis further demonstrated the clinical utility of the nomogram.

CONCLUSIONS

Radiomics nomogram provide a non-invasive method to predict Ki-67 proliferation index preoperatively in stage T1a-b lung adenocarcinoma, which might be the supplementary information for clinicians to choose the appropriate treatment program.