AB019. Machine learning models from FDG PET/CT for predicting recurrence in thymomas
Original Research

AB019. Machine learning models from FDG PET/CT for predicting recurrence in thymomas

Angelo Castello1, Luigi Manco2, Margherita Cattaneo3, Riccardo Orlandi3, Lorenzo Rosso3, Giorgio Alberto Croci4, Luigia Florimonte1, Giovanni Scribano5, Stefano Ferrero4, Mario Nosotti3, Massimo Castellani1, Gianpaolo Carrafiello6, Paolo Mendogni3

1Department of Nuclear Medicine, Fondazione IRCCS Ca' Granda, Ospedale Maggiore Policlinico, Milan, Italy; 2Medical Physics Unit, University Hospital of Ferrara, Ferrara, Italy; 3Thoracic Surgery, Fondazione IRCCS Ca’ Granda, Ospedale Maggiore Policlinico, Milan, Italy; 4Division of Pathology, Fondazione IRCCS Ca’ Granda, Ospedale Maggiore Policlinico, Milan, Italy; 5Postgraduate School in Medical Physics, Physics Department, University of Bologna, Bologna, Italy; 6Department of Diagnostic and Interventional Radiology, Foundation IRCCS Cà Granda-Ospedale Maggiore Policlinico, Milan, Italy

Correspondence to: Angelo Castello, MD. Department of Nuclear Medicine, Fondazione IRCCS Ca’ Granda, Ospedale Maggiore Policlinico, via Francesco Sforza, 28, Milan, 20151, Italy. Email: angelo.castello@policlinico.mi.it.

Background: An accurate pre-operative strategy is fundamental for patients’ risk stratification and prognostication in order to choose the appropriate treatment for patients with thymomas. This study aimed to develop machine learning (ML) models to predict recurrence in patients with thymomas using conventional and radiomic signatures extracted from fluorodeoxyglucose positron emission tomography/computed tomography (FDG PET/CT).

Methods: A total of 50 patients (25 males, 25 females; mean age 63.3 years) who underwent FDG PET/CT before surgery between 2012 and 2022 were retrospectively analyzed. Radiomic analysis was performed using free-from-recurrence (FFR) status as a reference. A total of 856 radiomic features (RFts) were extracted from PET and CT datasets following IBSI guidelines, and robust RFts were selected. The dataset was split into training (70%) and validation (30%) sets. Two ML models (PET- and CT-based, respectively), each with three classifiers—Random Forest (RF), Support-Vector-Machine (SVM), and Decision Tree, were trained and internally validated using RFts and clinical-metabolic signatures.

Results: A total of 50 regions of interest (ROIs) were selected and segmented. Height patients had recurrence (FFR 1), whereas 42 did not (FFR 0). Forty-three robust RFts were selected from the CT dataset and 16 from the PET dataset, predominantly wavelet-based RFts. Additionally, three metabolic PET parameters [i.e., relative PET (rPET), quantitative PET (qPET), and tumor-to-mediastinum (T/M) ratio] were selected and included in the PET model: tumor SUVmax/LiverSUVmax (rPET), tumor SUVpeak/LiverSUVmean (qPET), and the ratio between the SUVmax of thymomas and the SUVmean of aortic arc (T/M). Both the CT and PET models successfully discriminated FFR after surgery, with the CT Model slightly outperforming the PET Model across different classifiers. The performance metrics of the RF classifier for the CT and PET models were area under the curve (AUC) =0.970/0.949, accuracy =0.880/0.840, precision =0.884/0.842, recall =0.880/0.846, specificity =0.887/0.839, sensitivity =0.920/0.844, true-positive =81.8%/83.3%, true-negative =92.9%/84.6%, respectively (Figure 1).

Conclusions: ML-models trained on PET/CT radiomic features show promising results for predicting recurrence in patients with thymomas, which could be potentially applied in clinical practice for a better personalized treatment strategy.

Keywords: Thymomas; fluorodeoxyglucose positron emission tomography/computed tomography (FDG PET/CT); recurrence; radiomics; machine learning (ML)

Figure 1 Radial plot of RF learner performances scores of ML models. AUC, area under the curve; CT, computed tomography; ML, machine learning; PET, positron emission tomography; RF, Random Forest; TN, true-negative; TP, true-positive.

Acknowledgments

None.

Footnote

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://med.amegroups.com/article/view/10.21037/med-25-ab019/coif). P.M. serves as an unpaid editorial board member of Mediastinum from January 2025 to December 2026. The other authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee and institutional review board of Fondazione IRCCS Ca’ Granda Ospedale Maggiore Policlinico, Milan (study identification: INT121/24) and individual consent for this retrospective analysis was waived.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

doi: 10.21037/med-25-ab019
Cite this abstract as: Castello A, Manco L, Cattaneo M, Orlandi R, Rosso L, Croci GA, Florimonte L, Scribano G, Ferrero S, Nosotti M, Castellani M, Carrafiello G, Mendogni P. AB019. Machine learning models from FDG PET/CT for predicting recurrence in thymomas. Mediastinum 2025;9:AB019.

Article Options

Download Citation

Share