Abstract:

Objective To explore the CT features and predictive value of radiomics nomogram of visceral pleural invasion （VPI） in stage Ⅰ lung adenocarcinoma with peripheral solid nodules. Methods One hundred and fifty patients with stage Ⅰ lung adenocarcinoma with peripheral solid nodules treated at the Second Affiliated Hospital of Guizhou University of Traditional Chinese Medicine from August 2022 to November 2023 were selected as the study objects. Patients from August 2022 to March 2023 were defined as the training set （n=112）， and patients from April 2023 to November 2023 were defined as the validation set （n=38）. The training set was used to build the model， and the training set and validation set were used to evaluate the model performance respectively. In the training set， patients were divided into VPI positive group （n=35） and VPI negative group （n=77） based on the occurrence of VPI. Least absolute shrinkage and selection operator （LASSO） regression analysis was used to reduce the dimensionality of features. Multivariate logistic regression analysis was used to predict the influencing factors of VPI， and a radiomics nomogram prediction model was constructed based on the results of the multivariate analysis. Receiver operator characteristic （ROC） curves and calibration curves were used to evaluate the predictive efficacy of the prediction model. Results There were statistically significant differences in pathological types （χ²=11.49， P=0.003）， focal maximum diameter （t=5.83， P<0.001）， lobulation sign （χ²=9.29， P=0.002）， density （χ²=8.32， P=0.004）， intratumoral necrosis （χ²=5.86， P=0.015）， pleural traction （χ²=12.88， P<0.001）， pleural contact （χ²=4.82， P=0.028）， and adjacent pleural thickening （χ²=4.87， P=0.027） between the VPI positive group and negative group in the training set. LASSO regression analysis showed that 8 features were ultimately selected， and radiomics scores were constructed based on the corresponding coefficients of the features. Univariate analysis showed that， focal maximum diameter （OR=1.48， 95%CI：1.09-2.01， P=0.010）， lobulation sign （OR=5.09， 95%CI：2.31-6.00， P=0.001）， density （OR=4.25， 95%CI：1.47-7.18， P=0.004）， intratumoral necrosis （OR=2.27， 95%CI：1.01-5.17， P=0.049）， pleural traction （OR=6.75， 95%CI：1.92-13.68， P<0.001）， pleural contact （OR=3.58， 95%CI：1.18-5.65， P=0.018）， adjacent pleural thickening （OR=3.60， 95%CI：1.18-5.72， P=0.018）， and radiomics score （OR=19 418.06， 95%CI：394.18-957 161.04， P<0.001） were all influencing factors in the prediction of VPI in peripheral solid nodule stage Ⅰ lung adenocarcinoma patients. Multivariate analysis showed that， lobulation sign （OR=6.42， 95%CI：1.42-18.58， P=0.018）， intratumoral necrosis （OR=3.63， 95%CI：1.01-10.01， P=0.046）， pleural traction （OR=4.19， 95%CI：1.17-10.92， P=0.028）， and radiomics score （OR=179 711.20， 95%CI：525.13-61 552 573.59， P<0.001） were independent influencing factors in the prediction of VPI in patients with stage Ⅰ peripheral solid nodules of lung adenocarcinoma. A radiomics nomogram prediction model was established for indicators with statistical significance in multivariate analysis. ROC curve analysis showed that in the training set and validation set， the area under the curve （AUC） of the radiomics nomogram model predicting VPI of patients with peripheral solid nodules in stage Ⅰ lung adenocarcinoma was 0.88 （95%CI：0.82-0.94） and 0.87 （95%CI：0.78-0.97）， respectively， and the sensitivity was 93% and 82%， respectively. The specificity was 72% and 80%， respectively. C-indices of the training and validation set were 0.89 （95%CI：0.84-0.96） and 0.88 （95%CI：0.78-0.99）， respectively， and the calibration curves of both sets fitted well with the ideal curve. Conclusions The CT features of VPI in stage Ⅰ lung adenocarcinoma with peripheral solid nodules are lobular sign， intratumoral necrosis， and pleural traction. The radiomics nomogram model based on CT features of lobular sign， intratumoral necrosis， pleural traction， and radiomics score can predict VPI in patients with peripheral solid nodules in stage Ⅰ lung adenocarcinoma has high predictive efficacy.

Key words: Adenocarcinoma of lung, Pleura, Nomograms, Radiomics