A habitat-pathology-clinical transformer model for precision prognostication of neoadjuvant immunochemotherapy in ESCC
Esophageal squamous cell carcinoma (ESCC) shows marked inter-patient heterogeneity in response to neoadjuvant immunochemotherapy (NICT), highlighting the need for reliable pre-treatment predictors. In this multicenter study, 102 patients with advanced ESCC from two institutions were enrolled. Pre-treatment contrast-enhanced CT, whole-slide histopathology images, and clinical data were collected. Habitat-based radiomics was selected as the optimal imaging approach, with a Support Vector Machine (SVM) achieving an AUC of 0.723 (95% CI: 0.580–0.866) in the test cohort. For pathomics, the ResNet50-based multi-instance learning model with logistic regression was selected, achieving an AUC of 0.766 (95% CI: 0.639–0.894). These signatures were integrated with clinical features to develop HPTrans, a transformer-based multimodal model. HPTrans outperformed all single-modality approaches, achieving AUC of 0.921 (95% CI: 0.889–0.952) in the training cohort and 0.824 (95% CI: 0.720–0.928) in the test cohort. Collectively, HPTrans may aid pretreatment NICT response prediction and treatment selection in patients with advanced ESCC.
Authors
- Wei Li (ORCID: https://orcid.org/0000-0001-9956-0208)
- Qi Cao (ORCID: https://orcid.org/0000-0001-6970-8654)
- Shenglong Xie
- Gang Feng
- Wenlei Lu
- Zhiyang Li
- Bin He
Institutions
- University of Electronic Science and Technology of China (CN)
- Tianjin Medical University Cancer Institute and Hospital (CN)
- Sichuan Academy of Medical Sciences & Sichuan Provincial People's Hospital (CN)
Publication Details
- Journal
- iScience
- Published
- 2026-11-17
- DOI
- https://doi.org/10.1016/j.isci.2026.118158
- Primary Topic
- Radiomics and Machine Learning in Medical Imaging
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- Department of Science and Technology of Sichuan Province