https://doi.org/10.1140/epjs/s11734-024-01446-2
Regular Article
Leveraging machine learning models for enhanced differentiation of hard-diagnosed lung lesions
1
Saint Petersburg Electrotechnical University “LETI”, Prof. Popova, 5, 197022, Saint-Petersburg, Russia
2
Saint-Petersburg State Research Institute of Phthisiopulmonology, Ligovsky Ave. 2-4, 194064, Saint Petersburg, Russia
3
Centre for Nonlinear Chemistry, Immanuel Kant Baltic Federal University, Nevskogo St. 14, 236041, Kaliningrad, Russia
Received:
5
November
2024
Accepted:
12
December
2024
Published online:
7
January
2025
The accurate differentiation of pulmonary lesions (nodule/mass) is crucial for selecting appropriate treatment strategies, particularly for distinguishing benign lesions such as hamartoma and tuberculoma from malignancies like non-small cell lung cancer (NSCLC). This study investigates the use of machine learning (ML) models, including Decision Tree (DT), Random Forest (RF), and CatBoost (CB), to identify key predictors of nodules type based on computed tomography (CT) imaging features. We analyzed CT data from 363 patients with confirmed diagnoses of hamartoma, tuberculoma, or NSCLC to evaluate the models’ predictive performance and identify the most significant diagnostic features. The models demonstrated high accuracy, sensitivity, and specificity, with DT and CB models highlighting changes in surrounding tissue as primary indicators, whereas RF integrated additional predictors, providing a nuanced classification framework. These findings suggest that ML models can enhance diagnostic accuracy for lung lesions and reduce unnecessary invasive procedures, although further validation in larger cohorts is recommended.
Copyright comment Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
© The Author(s), under exclusive licence to EDP Sciences, Springer-Verlag GmbH Germany, part of Springer Nature 2024
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
