Marta Borowska, Małgorzata Mojsak, Ewelina Bębas, Jolanta Pauk, Marcin Hładuński, Małgorzata Domino
Abstract
Texture analysis is a foundational approach in imaging studies and demonstrates excellent diagnostic performance, with radiomic analysis being the most widely used method. New approaches to texture analysis continue to be developed.
Introduction
Non–small cell lung cancer (NSCLC) is the second most common cancer globally, accounting for 85% of lung cancer cases and approximately 1.4 million deaths annually [1]. Among diagnoses lung cancers, there are four major histological subtypes: adenocarcinoma (ADC, approximately 40%, the most common subtype in passive smokers), squamous cell carcinoma (SCC, approximately 30%, the most common subtype in active smokers), small cell carcinoma (SC, approximately 15%), and large cell carcinoma (LCC, approximately 10%) [2].
Materials and methods
The study analyzed 154 MRI images from 31 patients (20 men (65%) and 11 women (35%); mean age: 65 years, range: 35–82 years) with metabolically active lung tumors confirmed through PET/MRI.
Results
Among 75 entropy–based texture features, returned from combinations of five entropy–based methods (n = 5), five scales (n = 5), and three image types – raw images and two filtrations (n = 3), differences between ADC and SCC groups were presented for SampEn2D (Fig 3), FuzzEn2D (Fig 4), PermEn2D (Fig 5), DispEn2D (Fig 6), and DistEn2D (Fig 7), separately.
Discussion
The proposed protocol for MRI image segmentation, processing, and analysis facilitates the extraction of entropy–based texture features from lung cancer patients. This study employed PET/MRI images to confirm tumor metabolic activity, while feature extraction and analysis were conducted solely on MRI images.
Conclusion
Applying entropy–based ML–supported classification of MRI images enables the differentiation of ADC and SCC. However, combining FOS, SOS, and entropy–based texture features achieved the highest classification performance.
Citation: Borowska M, Mojsak M, Bębas E, Pauk J, Hładuński M, Domino M (2026) Machine learning using entropy–based texture features from MRI to differentiate histological subtypes of non–small cell lung cancer identified as metabolically active on PET/MRI. PLoS One 21(1): e0338373. https://doi.org/10.1371/journal.pone.0338373
Editor: Amgad Muneer, The University of Texas, MD Anderson Cancer Center, UNITED STATES OF AMERICA
Received: July 23, 2025; Accepted: November 20, 2025; Published: January 21, 2026
Copyright: © 2026 Borowska et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are available at: https://doi.org/10.5281/zenodo.15845728.
Funding: The study was supported by the Polish Ministry of Science and Higher Education as a part of the project WZ/WM–IIB/2/2024. Research for patients was supported by the project funded by the National Centre for Research and Development in the framework of Programme “Prevention practices and treatment of civilization diseases” - STRATEGMED (contract no. STRATEGMED2/266484/2/NCBR/2015).
Competing interests: The authors have declared that no competing interests exist.
