Radiomics Chest CT-Based Prediction of Lymphovascular Invasion in Lung Cancer: Preliminary Findings

All articles of this category (opens in new window) Address for correspondence Maria Eduarda Barbosa da Silva (mebsilva@hcpa.edu.br) Introduction: Radiomics is the automated extraction of quantitative features (texture, shape, intensity) from a volume of interest (VOI) in medical images. Lymphovascular invasion (LVI) is a significant prognostic factor in non-small cell lung cancer (NSCLC). Therefore, evaluating LVI using chest computed tomography (CT) radiomics may serve as a new and valuable strategy to improve prognostic assessment and support treatment decision-making in lung cancer . Objective: To assess the potential presence of LVI based on CT radiomics from patients with NSCLC . Methods: A prospective study that included patients with NSCLC who underwent pulmonary resection between March 2024 and June 2025 at HCPA. The presence of LVI was determined by the histopathological analysis of the surgical specimen. The tumor visible on the patient’s chest CT was segmented as a volume of interest (VOI) using 3D Slicer software. The PyRadiomics plugin was used to extract features after VOI pre-processing. The analysis pipeline consisted of pre-processing with data standardization (Z-score), removal of low-variance features, and feature selection using the SelectKBest. After training a Logistic Regression model with L2 penalty (Ridge), validation was performed using the Leave-One-Out Cross-Validation (LOOCV) strategy. To estimate the variability of the metrics, Bootstrap with 100 resamples was applied . Results: The sample consisted of 24 patients, with 58% being female, showing a median age of 67 years. The mean tumor size was 29 mm (± 22). Adenocarcinoma was the most common subtype (83%), stage IA2 was the most frequent (37%), and 54% (n = 13) presented with LVI. PyRadiomics extracted 1,288 radiomic features. Three Wavelet-transformed texture features were the most frequently selected. The model achieved a ROC AUC of 0.73 (95% CI: 0.45 – 0.94), a weighted F1-score of 0.72 (95% CI: 0.46 – 0.93), a weighted Precision of 0.73 (95% CI: 0.47 – 0.95), and a weighted Recall of 0.73 (95% CI: 0.47 - 0.94). Conclusions The moderate performance of the model suggests a potential association between radiomic features reflecting heterogeneous tumor texture and the presence of LVI. However, the small sample size and lack of external validation represent important limitations that must be addressed to enhance the model’s reliability and generalizability. Publication History Article published online: 26 June 2026 © 2026. The Author(s). This is an open access article published by Thieme under the terms of the Creative Commons Attribution 4.0 International License, permitting copying and reproduction so long as the original work is given appropriate credit (https://creativecommons.org/licenses/by/4.0/) Thieme Revinter Publicações Ltda. Rua Rego Freitas, 175, loja 1, República, São Paulo, SP, CEP 01220-010, Brazil Bibliographical Record Maria Eduarda Barbosa da Silva, João Gabriel Colleto Souza, Cristiano da Lima Hackmann, Lisiane Pentz, Yuri Thomé Machado Petrillo, Bernardo Martins Lavarda, Maurício Guidi Saueressig. Radiomics Chest CT-Based Prediction of Lymphovascular Invasion in Lung Cancer: Preliminary Findings. Brazilian Journal of Oncology 2026; 22. DOI: 10.1055/s-0046-1825475