Novel Texture-Based Radiomic Signatures for Non-Invasive Prediction of EGFR Mutation Status in Lung Nodules

Background. Accurate identification and analysis of lung nodules via computed tomography are pivotal for lung cancer diagnosis and the detection of genetic alterations, such as epidermal growth factor receptor (EGFR) mutations. While conventional radiomics has become a cornerstone of medical imaging, its predictive power for determining EGFR mutation status remains limited, necessitating innovative approaches to improve diagnostic reliability. Objective: to enhance the accuracy of EGFR mutation status prediction in lung nodules by introducing and integrating novel texture-based radiomics features into conventional radiomics analysis. Material and methods. Three novel radiomic features were developed: Adaptive Texture Contrast (ATC), Directional Texture Uniformity (DTU), and Co-occurrence of Texture Transitions (CTT). They were designed to capture complex texture patterns associated with EGFR mutations. Integrating these features, a classification model was employed to differentiate EGFR mutant from wild-type lung nodules. Results. The incorporation of ATC, DTU, and CTT into the radiomics feature set improved the classification accuracy by 4%. The Minimum Redundancy Maximum Relevance (MRMR) feature selection method further validated the significance of these features, ranking them as the top contributors to the model’s predictive performance. Conclusion. The findings underscore the potential of advanced texture analysis in improving the diagnostic capabilities of radiomics for lung nodule classification. By enabling more accurate predictions of EGFR mutations, the study supports the advancement of personalized medicine and targeted treatment strategies in lung cancer, highlighting the importance of continuous innovation in feature engineering.

1. Shariaty F, Pavlov VA, Zavyalov SV, et al. Application of a texture appearance model for segmentation of lung nodules on computed tomography of the chest. Journal of the Russian Universities. Radioelectronics. 2022; 25(3): 96–117 (in Russ.). https://doi.org/10.32603/1993-8985-2022-25-3-96-117. [Шариати Ф., Павлов В.А., Завьялов С.В. и др. Применение модели внешнего вида текстуры для сегментации легочных узлов при компьютерной томографии грудной клетки. Известия высших учебных заведений России. Радиоэлектроника. 2022; 25(3): 96–117. https://doi.org/10.32603/1993-8985-2022-25-3-96-117.]

2. Shariaty F, Pavlov V, Baranov M. AI-driven precision oncology: integrating deep learning, radiomics, and genomic analysis for enhanced lung cancer diagnosis and treatment. Signal Image Video Process. 2025; 19(9): 693. https://doi.org/10.1007/s11760-025-04244-y.

3. Roy SS, Hsu CH, Kagita V. Deep learning applications in image analysis. Springer; 2023: 224 pp. Available at: https://link.springer.com/content/pdf/10.1007/978-981-99-3784-4.pdf (accessed 02.07.2025).

4. Skalunova M, Shariaty F, Rozov S, Radmard AR. Personalized chemotherapy selection for lung cancer patients using machine learning and computed tomography. In: 2023 International Conference on Electrical Engineering and Photonics (EExPolytech); 2023: 128–31. https://doi.org/10.1109/EExPolytech58658.2023.10318700.

5. Shariaty F, Duan L, Pavlov V, et al. A novel gene assay combined with medical imaging for accurate prognosis and prediction of cancer type. In: 2022 International Conference on Electrical Engineering and Photonics (EExPolytech); 2022: 118–21. https://doi.org/10.1109/EExPolytech56308.2022.9950997.

6. Liu Q, Liu X. Feature extraction of human viruses microscopic images using gray level co-occurrence matrix. In: 2013 International Conference on Computer Sciences and Applications, Wuhan, China; 2013: 619–22. https://doi.org/10.1109/CSA.2013.149.

7. Huang W, Luo M, Liu X, et al. Arterial spin labeling images synthesis from sMRI using unbalanced deep discriminant learning. IEEE Transact Med Imaging. 2019; 38(10): 2338–51. https://doi.org/10.1109/TMI.2019.2906677.

8. Augustyniak K, Chrabaszcz K, Smeda M, et al. High-resolution Fourier transform infrared (FT-IR) spectroscopic imaging for detection of lung structures and cancer-related abnormalities in a murine model. Appl Spectrosc. 2022; 76(4): 439–50. https://doi.org/10.1177/00037028211025540.

9. Bakr S, Gevaert O, Echegaray S, et al. A radiogenomic dataset of non-small cell lung cancer. Sci Data. 2018; 5: 180202. https://doi.org/10.1038/sdata.2018.202.

10. NSCLC Radiogenomics. Available at: https://www. cancerimagingarchive.net/collection/nsclc-radiogenomics/ (accessed 02.07.2025).

11. Peli E. Contrast in complex images. J Opt Soc Am A. 1990; 7(10): 2032–40. https://doi.org/10.1364/JOSAA.7.002032.

12. Haralick R, Shanmugam K, Dinstein I. Textural features for image classification. IEEE Transact Syst Man Cybern. 1973; SMC-3(6): 610–21. https://doi.org/10.1109/TSMC.1973.4309314.

13. Ding C, Peng H. Minimum redundancy feature selection from microarray gene expression data. J Bioinform Comput Biol. 2005; 3(2): 185–205. https://doi.org/10.1142/S0219720005001004.

14. Sun L, Dong Y, Xu S, et al. Predicting multi-gene mutation based on lung cancer CT images and Mut-SeResNet. Appl Sci. 2023; 13(3): 1921. https://doi.org/10.3390/app13031921.

15. Mahajan A, Kania V, Agarwal U, et al. Deep-learning-based predictive imaging biomarker model for EGFR mutation status in non-small cell lung cancer from CT imaging. Cancers. 2024; 16(6): 1130. https://doi.org/10.3390/cancers16061130.