Original Research
AB060. A predictive model for surgical resectability in thymic epithelial tumors using clinical data: a step towards integrating radiomics and histopathology
Mohammed Ghazali1, Mae Shu1, DuyKhanh P. Ceppa2, Rohan Maniar3, Patrick J. Loehrer3
1Indiana University School of Medicine, Indianapolis, IN, USA;
2Division of Cardiothoracic Surgery, Department of Surgery, Indiana University School of Medicine, Indianapolis, IN, USA;
3Division of Hematology/Oncology, Department of Medicine, Indiana University School of Medicine, Indianapolis, IN, USA
Correspondence to: Patrick Loehrer, MD. Division of Hematology/Oncology, Department of Medicine, Indiana University School of Medicine, 535 Barnhill Dr., Suite 210, Indianapolis, IN 46202, USA. Email: ploehrer@iu.edu.
Background: Thymic epithelial tumors (TETs) are rare, heterogeneous malignancies. Surgery is the primary treatment for early-stage TETs, while induction strategies aim to reduce tumor mass and improve R0 resection likelihood. However, treating borderline resectable TETs remains a challenge. Machine learning (ML) can offer insights beyond traditional methods and the field of RaPtomics applies ML to radiologic and pathological images. The purpose of this article is to establish a baseline model that will allow future evaluation of implicit features like texture and homogeneity against current clinical standards. To start, we hypothesize that a clinical model can accurately predict surgical resectability in TET patients.
Methods: From a REDCap database of 1,089 TET patients at Indiana University’s thoracic oncology clinic, 202 patients met the following criteria: confirmed clinical stage I-III TET, confirmed resection status, and surgery with curative intent. Clinical and demographic data points, including tumor size, World Health Organization (WHO) staging, age, and concurrent autoimmune diagnoses, were used to encode 52 features. Patients were divided into cohorts based on resection: R0 resection (135 patients) and R1/R2 resection (67 patients). Five ML models were assessed via stratified cross-validation, evaluating accuracy and area under the curve (AUC).
Results: An ensemble model combining Random Forest, AdaBoost, and XGBoost classifiers achieved 66% accuracy and an AUC of 0.73, indicating moderate discriminatory power in distinguishing R0 from R1/R2 resection cases. A sensitivity of 0.78 for R1/R2 suggests that the model effectively identifies unresectable cases, while a precision-recall AUC of 0.876 supports its utility for ranking patients by resectability risk. Top contributing features included WHO and Masaoka staging, highlighting the predictive power of current oncologic staging systems. Notably, sex assigned at birth and race/ethnicity also emerged as meaningful contributors, suggesting possible demographic associations with resectability.
Conclusions: This study establishes a foundational ML model for predicting surgical resectability in TETs. Future work will address current model limitations by incorporating RaPtomics features and expanding the cohort to differentiate between direct-to-surgery and neoadjuvant chemotherapy patients. This may enhance decision-making for neoadjuvant therapy and improve R0 resection likelihood.
Keywords: Thymic Epithelial Tumor; Machine Learning; Raptomics; Clinical Modeling
Acknowledgments
The William P. Loehrer Thymic Cancer Fund.
Funding: This work was funded, in part, with support from the Indiana Clinical and Translational Sciences Institute Grant funded, in part from the National Institutes of Health (UL1TR002529).
Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://med.amegroups.com/article/view/10.21037/med-25-ab060/coif). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. The 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 board of the Total Cancer protocol IUSCC-0678, IRB #1807389306 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-ab060
Cite this abstract as: Ghazali M, Shu M, Ceppa DP, Maniar R, Loehrer PJ. AB060. A predictive model for surgical resectability in thymic epithelial tumors using clinical data: a step towards integrating radiomics and histopathology. Mediastinum 2025;9:AB060.
