A multi-data fusion deep learning model for prognostic prediction in upper tract urothelial carcinoma

BackgroundUpper tract urothelial carcinoma (UTUC) is a rare but highly invasive urinary malignancy with a high postoperative recurrence rate.MethodsWe retrospectively collected data from 133 UTUC patients who underwent radical nephroureterectomy between 2005 and 2017. Patients were divided into a tr...

Full description

Saved in:
Bibliographic Details
Main Authors: Hongdi Sun, Siping Chen, Yongxing Bao, Fengyan You, Honghui Zhu, Xin Yao, Lianguo Chen, Jiangwei Miao, Fanggui Shao, Xiaomin Gao, Binwei Lin
Format: Article
Language:English
Published: Frontiers Media S.A. 2025-08-01
Series:Frontiers in Oncology
Subjects:
Online Access:https://www.frontiersin.org/articles/10.3389/fonc.2025.1644250/full
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:BackgroundUpper tract urothelial carcinoma (UTUC) is a rare but highly invasive urinary malignancy with a high postoperative recurrence rate.MethodsWe retrospectively collected data from 133 UTUC patients who underwent radical nephroureterectomy between 2005 and 2017. Patients were divided into a training set (n=103) and a testing set (n=30). A multi-modal deep learning model named Multi-modal Image-Clinical Combination Classifier (MICC) was developed by integrating multi-phase contrast-enhanced CT imaging and clinical data. The model’s prognostic performance was compared with two unimodal models—ImageNet (CT-based) and ClinicalNet (clinical data-based)—and traditional clinical parameters including pathological T stage. Feature importance was evaluated using SHapley Additive exPlanations (SHAP).ResultsThe MICC model achieved superior prognostic accuracy with AUCs of 0.918 and 0.895 in the training and testing sets, respectively, outperforming unimodal models. Classification metrics were robust, with accuracy of 0.854, sensitivity of 0.889, specificity of 0.836, negative predictive value (NPV) of 0.933, and positive predictive value (PPV) of 0.744. Precision-recall analysis confirmed strong identification of high-risk patients despite dataset imbalance. SHAP analysis highlighted that CT imaging features contributed most significantly to the model’s predictions.ConclusionIntegrating multi-phase CT imaging with clinical data, the MICC model provides accurate prognostic prediction for UTUC patients. This approach has potential to assist clinicians in personalized risk stratification and treatment planning, ultimately improving patient outcomes.
ISSN:2234-943X