MRI-Based Habitat Radiomics for Predicting Early Recurrence After Pancreatic Cancer Resection
Journal of Clinical Medicine · 15.11.2026 · 15 (22) · 8956
DOI: 10.3390/jcm15228956 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Clinical Medicine, online erschienen am 15.11.2026. Verfasst von Jiale Li und 9 weiteren Autor:innen. Die Arbeit nennt 40 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Natural Science Foundation of Liaoning Province.
Abstract (Originalsprache)
Purpose
To investigate the predictive performance of multi-sequence MRI habitat radiomics for early recurrence (ER) following curative resection of pancreatic ductal adenocarcinoma (PDAC) and to provide preoperative imaging evidence to guide personalized diagnosis and treatment.
Methods
This study collected retrospective clinical data from a total of 260 PDAC patients at two medical centers. Tumors were segmented into three habitat subregions via unsupervised K-means clustering. Radiomic features from these subregions were used to construct the habitat models. The performance of the traditional clinical model, the whole-tumor radiomics model, the single-sequence habitat models, the combined habitat model, and the Clinic-Habitat combined model was compared. The model’s validity was evaluated using a series of metrics, including the area under the ROC curve (AUC), calibration curve analysis, and decision curve analysis (DCA). The results of the model were interpreted using the additive interpretation method of SHAP. Spearman’s rank-order correlation analysis was performed to explore biological correlations between radiomic scores and pathological markers. Risk-stratified survival analysis was implemented based on model-derived optimal cutoff values.
Results
The Clinic-Habitat model demonstrated the best performance, with AUCs of 0.922 (training), 0.896 (internal validation), and 0.886 (external test), outperforming the clinical baseline model, the clinical-pathological model, the individual-sequence habitat models, and the whole-tumor radiomics model. SHAP analysis indicated that T1WI-Hab-score1 exerted the greatest influence on model predictions. The Hab-radiomic score was significantly correlated with pathological markers reflecting tumor aggressiveness. A cutoff value of 0.556 derived from the Clinic-Habitat model effectively stratified patients into high- and low-risk ER subgroups, with Kaplan–Meier curves showing a significant difference in RFS between the two groups (p < 0.001).
Conclusions
Multi-sequence MRI-based habitat radiomics enables preoperative prediction of ER risk in PDAC and facilitates patient risk stratification and personalized treatment planning.
Förderung
- Natural Science Foundation of Liaoning Province
Zitieren
APA
Li, J., Song, J., Zhu, A., Zhang, L., Wang, W., Guo, W., Jin, J., Wang, X., Fang, X., & Dong, Y. (2026). MRI-Based Habitat Radiomics for Predicting Early Recurrence After Pancreatic Cancer Resection. Journal of Clinical Medicine, 15(22), 8956. https://doi.org/10.3390/jcm15228956
Vancouver
Li J, Song J, Zhu A, Zhang L, Wang W, Guo W, et al. MRI-Based Habitat Radiomics for Predicting Early Recurrence After Pancreatic Cancer Resection. Journal of Clinical Medicine. 2026;15(22):8956. doi:10.3390/jcm15228956
BibTeX
@article{Li2026,
author = {Li, Jiale and Song, Jiadong and Zhu, Aonan and Zhang, Lijie and Wang, Weijia and Guo, Wei and Jin, Jing and Wang, Xiaona and Fang, Xiangming and Dong, Yang},
title = {MRI-Based Habitat Radiomics for Predicting Early Recurrence After Pancreatic Cancer Resection},
journal = {Journal of Clinical Medicine},
year = {2026},
volume = {15},
number = {22},
pages = {8956},
doi = {10.3390/jcm15228956}
}
EUVIMED