Differentiation of focal organising pneumonia and peripheral adenocarcinoma in solid lung lesions using thin-section CT-based radiomics.

AIM: To evaluate the predictive role of radiomics based on computed tomography (CT) in discriminating focal organising pneumonia (FOP) from peripheral lung adenocarcinoma (LA).

MATERIALS AND METHODS: Institutional research board approval was obtained for this retrospective study. One hundred and seventeen patients with FOP and 109 patients with LA who underwent thin-section CT from January 2011 to August 2017 were reviewed systematically and analysed. The clinical and radiological features were established as model A and multi-feature-based radiomics as model B. The diagnostic performance of model A, model B, and model A+B were evaluated and compared via receiver operating characteristic (ROC) curve analysis and logistic regression analysis.

RESULTS: Sex, symptoms, necrosis, and the halo sign were identified as independent predictors of LA. The area under the ROC curve (Az value), accuracy, sensitivity, and specificity of model A were 0.839, 75.7%, 82.6%, and 69.2% respectively. Model B showed significantly higher accuracy than model A (83.6% versus 75.7%, p=0.032). The top four best-performing features, WavEnLH_s-3, WavEnHH_s-3, Teta3, and Volume, performed as independent factors for discriminating LA. Regression analysis indicated that model B had superior model fit than model A with Akaike information criterion (AIC) values of 73.6% versus 59.1%, respectively. Combining model A with model B is useful in achieving better diagnostic performance in discriminating FOP from LA: the Az value, accuracy, sensitivity, and specificity were 0.956, 87.6%, 85.3%, and 89.7% respectively.

CONCLUSIONS: Radiomics based on CT exhibited better diagnostic accuracy and model fit than clinical and radiological features in discriminating FOP from LA. Combination of both achieved better diagnostic performance.