Researchers have created advanced deep learning (DL) models that surpass existing tools in predicting the malignant potential of incidental lung lesions. A study published in European Radiology reveals that these models, which integrate imaging and clinical data, exhibited higher sensitivity and specificity compared to the Brock model, a standard malignancy prediction tool. The research involved a diverse test set of 269 pulmonary nodules from various institutions, showing the DL models maintained consistent performance across different settings. The screening-trained model achieved an area under the curve (AUC) of 0.74, while the combined data model reached an AUC of 0.72, both outperforming the Brock model, which had an AUC of 0.63. The study suggests that these DL models could enhance decision-making in clinical practices surrounding lung cancer management.
Wed, 12 Aug 2026 19:25:11 GMT | Radiology Business