The Role of Using Machine Learning Algorithms in Non-contrast CT Based Images in Determining Pathological Subtypes of Lung Cancer.

Authors

Beyza Nur Kuzan, Can Ilgin, Murat Emeç, Hacer Şahika Yildiz, Gonca Gül Geçmen, Tamer Baysal

Published in

Technology in cancer research & treatment. Volume 25. Pages 15330338261483507. Epub Aug 25, 2026.

Abstract

IntroductionLung cancer is a common cancer with a high mortality rate worldwide. Significant differences in treatment, prognosis, and recurrence rates exist depending on lung cancer subtype. The goal of our study is to use machine learning models based on chest computed tomography (CT) scans to predict tumor subtypes.MethodsA total of 99 lung cancer cases with histopathological tissue diagnosis were included in this retrospective, single-center study. Tumor segmentation was performed in three-dimension (3D) using mediastinal and parenchymal windows on a chest CT, and 180 features were extracted, of which 90 were retained after recursive feature elimination (RFE). Due to class imbalance, model performance was tested using a stratified folding method. Machine learning (ML) algorithms CatBoost, XGBoost, and Ensemble models were trained separately on parenchymal and mediastinal window feature sets.ResultsThe most common type of lung cancer among patients was adenocarcinoma (AC) in the non-small cell lung cancer (NSCLC) group (n=57, 57.58%). When compared to XGBoost and Ensemble models, CatBoost demonstrated superior performance in differentiating between small cell lung cancer (SCLC) and NSCLC, achieving 90% accuracy and an 80.39% area under the curve (AUC). The Ensemble model, also trained with the parenchymal window, achieved the highest accuracy in distinguishing adenocarcinoma from other lung cancers (non-adenocarcinoma), with 75% accuracy and 72.92% AUC, respectively.ConclusionIn summary, our research has demonstrated that machine learning (ML) algorithms can accurately determine lung cancer tumor subtypes, eliminating the need for additional invasive procedures.

PMID:
42641643
Bibliographic data and abstract were imported from PubMed on 16 Oct 2026.

Read full publication at:
Please sign in to see all details.

Stats

1-terrible, 9-excellent. How would you rate this publication? Sign in in to submit your rating.

Post a comment

You need to be signed in to post comments. You can sign in here.

Comments

There are no comments yet.