Pisula, Juan I. 
ORCID: 0000-0002-6131-8528, Helbig, Doris ORCID: 0000-0002-5841-4631, Sancéré, Lucas ORCID: 0009-0000-3857-5641, Persa, Oana-Diana, Bürger, Corinna, Fröhlich, Anne, Lorenz, Carina ORCID: 0009-0006-7827-8055, Bingmann, Sandra, Niebel, Dennis, Drexler, Konstantin, Landsberg, Jennifer, Thomas, Roman ORCID: 0000-0001-9132-4876, Bozek, Katarzyna ORCID: 0000-0002-0917-6876 and Brägelmann, Johannes ORCID: 0000-0002-1306-2169
(2025).
Explainable, federated deep learning model predicts disease progression risk of cutaneous squamous cell carcinoma.
npj Precision Oncology, 9 (1).
pp. 1-11.
Springer Nature.
ISSN 2397-768X
![[thumbnail of s41698-025-00997-4.pdf]](http://kups.ub.uni-koeln.de/style/images/fileicons/application_pdf.png "s41698-025-00997-4.pdf") |
PDF
s41698-025-00997-4.pdf Bereitstellung unter der CC-Lizenz: Creative Commons Attribution. Download (3MB) |
Abstract
[Artikel-Nr.: 205] Predicting cancer patient disease progression is a key step towards personalized medicine and secondary prevention. Risk stratification systems based on clinico-pathological criteria aim to identify high-risk patients, but accurate predictions remain challenging. Deep learning models present new opportunities for patient risk prediction, yet their interpretability has been largely unexplored. We developed a transformer-based approach for predicting progression of cutaneous squamous cell carcinoma (cSCC) patients based on diagnostic histopathology tumor slides. Our initial model showed AUROC = 0.92 on a held-out test set, with average AUROC of 0.65 on external validation cohorts. To further increase generalizability and reduce potential privacy concerns, we trained the model in a federated manner across three clinical centers, reaching AUROC = 0.82 across all cohorts, with image-based risk scores achieving hazard ratios up to 7.42 ( p < 0.01) in multivariable analyses. Through interpretability analysis, we identified spatial and morphological features predictive of progression, suggesting that tumor boundary information and tissue heterogeneity characterize progressive cSCCs. Trained exclusively on routine diagnostic slides and offering biological insights, our model can improve secondary prevention and understanding of cSCC while enabling deployment across clinical centers without administrative overheads or privacy concerns.
| Item Type: | Article |
| Creators: | Creators Email ORCID ORCID Put Code Persa, Oana-Diana UNSPECIFIED UNSPECIFIED UNSPECIFIED Bürger, Corinna UNSPECIFIED UNSPECIFIED UNSPECIFIED Fröhlich, Anne UNSPECIFIED UNSPECIFIED UNSPECIFIED Bingmann, Sandra UNSPECIFIED UNSPECIFIED UNSPECIFIED Niebel, Dennis UNSPECIFIED UNSPECIFIED UNSPECIFIED Drexler, Konstantin UNSPECIFIED UNSPECIFIED UNSPECIFIED Landsberg, Jennifer UNSPECIFIED UNSPECIFIED UNSPECIFIED |
| URN: | urn:nbn:de:hbz:38-806593 |
| Identification Number: | 10.1038/s41698-025-00997-4 |
| Journal or Publication Title: | npj Precision Oncology |
| Volume: | 9 |
| Number: | 1 |
| Page Range: | pp. 1-11 |
| Number of Pages: | 11 |
| Date: | 28 June 2025 |
| Publisher: | Springer Nature |
| ISSN: | 2397-768X |
| Language: | English |
| Faculty: | Faculty of Medicine |
| Divisions: | CECAD - Cluster of Excellence Cellular Stress Responses in Aging-Associated Diseases Faculty of Medicine > Dermatologie > Klinik und Poliklinik für Dermatologie und Venerologie Faculty of Medicine > Medizinische Statistik und Bioinformatik > Institut für Medizinische Statistik und Bioinformatik � IMSB Faculty of Medicine > Pathologie und Neuropathologie > Institut für Pathologie Faculty of Medicine > Weitere > Translationale Genomik Zentrum für Molekulare Medizin |
| Subjects: | Medical sciences Medicine |
| ['eprint_fieldname_oa_funders' not defined]: | Publikationsfonds UzK |
| Refereed: | Yes |
| URI: | http://kups.ub.uni-koeln.de/id/eprint/80659 |
https://orcid.org/0000-0002-6131-8528Downloads
Downloads per month over past year
Altmetric
Export
Actions (login required)
 |
View Item |