Photo Credit: freepik

The following is a summary of “Malignancy risk stratification for pulmonary nodules: comparing a deep learning approach to multiparametric statistical models in different disease groups,” published in the January 2025 issue of Pulmonology by Piskorski et al. 

Researchers conducted a retrospective study to evaluate the performance of the lung-cancer-prediction-convolutional-neural-network (LCP-CNN), a deep learning-based approach, compared to multiparametric statistical methods (Brock model and Lung-RADS®) for risk classification of incidentally detected pulmonary nodules across cohorts with varying risk profiles and pulmonary diseases.  

They analyzed non-contrast and contrast-enhanced CT scans of pulmonary nodules (5-30 mm) with ground truth defined by histology or follow-up stability. The final dataset included 297 patients and 422 nodules, 105 of which were malignant. Diagnostic accuracy, including ROC analysis, was used to evaluate the classification performance of LCP-CNN, Brock model, and Lung-RADS® across subcohorts (total, screening, emphysema, and interstitial lung disease).  

The results showed that LCP-CNN outperformed the Brock model in total and screening cohorts (AUC 0.92 [95% CI: 0.89-0.94] and 0.93 [95% CI: 0.89-0.96]). LCP-CNN also exhibited superior sensitivity to both the Brock model and Lung-RADS® in the total screening and emphysema cohorts at a 5% risk threshold. At a 65% threshold, LCP-CNN showed better or equivalent sensitivity compared to the Brock model and Lung-RADS®. No significant performance differences were observed between subcohorts.  

Investigators concluded that the study provided additional evidence for integrating deep learning-based decision support systems into pulmonary nodule classification workflows, regardless of patient risk profile or underlying pulmonary disease.  

Source: pubmed.ncbi.nlm.nih.gov/39747589/