Early Post‐ECPR Brain CT for Prediction of Neurological Outcome: An Interpretable Approach
International Journal of Clinical Practice · 2026-01 · 2026 (1)
DOI: 10.1155/ijcp/9152075 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Clinical Practice, online erschienen am 21.06.2026. Verfasst von Man Wang und 6 weiteren Autor:innen. Die Arbeit nennt 44 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Natural Science Foundation of China, Youth Science Foundation of Guangxi Medical University.
Abstract (Originalsprache)
Background
This study aims to develop a radiomics‐driven machine learning model based on early brain CT features to predict neurological outcomes in patients undergoing ECPR and to investigate whether the integration of these radiomic features provides incremental predictive value beyond traditional clinical characteristics.
Methods
We retrospectively analyzed 60 patients who underwent ECPR from January 2020 to September 2025. All patients underwent noncontrast brain CT within 24 h following ECPR. Neurological outcomes were assessed at hospital discharge using the CPC scale (1‐2 favorable and 3–5 unfavorable), as long‐term follow‐up data were unavailable in this retrospective study. Radiomic features were extracted from the bilateral cortex, white matter, and caudate‐putamen on early post‐ECPR brain CT scans using the Harvard‐Oxford atlas. Five machine learning models (LR, RF, KNN, SVM, and XGB) were developed for each region via nested cross‐validation, and the best‐performing cortical region was selected based on predictive performance. Stable radiomic features from the optimal region (selection frequency ≥ 60%) were then used to calculate a radiomic score (Rad‐score) for each patient. Finally, a radiomic‐enhanced model was constructed by integrating this Rad‐score with clinical variables using logistic regression with LASSO feature selection within a nested cross‐validation framework, and its performance was compared against a baseline clinical model.
Results
Among the three brain regions evaluated, the cortex yielded the highest predictive performance for neurological outcomes. The radiomics‐enhanced model demonstrated superior discriminative performance compared with the baseline clinical model (mean AUC: 0.804 vs. 0.648; ΔAUC = 0.156). Significant improvements in risk reclassification were observed (NRI = 0.344, 95% CI: 0.093–0.583, p = 0.006; IDI = 0.259, 95% CI: 0.115–0.388, p < 0.001), along with better calibration (Brier score: 0.1237 vs. 0.2025) and higher clinical net benefit (average improvement: 0.0962). SHAP analysis identified the Rad‐score as the most influential predictor.
Conclusion
Early head CT radiomic features, particularly those derived from the cerebral cortex, may serve as a potential predictor of neurological outcomes in ECPR patients. Compared with clinical features alone, the radiomics‐enhanced model improves neurological outcome prediction, suggesting incremental prognostic value that could offer preliminary support for early clinical decision‐making.
Förderung
- National Natural Science Foundation of China
- Youth Science Foundation of Guangxi Medical University
Zitieren
APA
Wang, M., Wang, F., Pang, G., Wei, W., Zhai, Y., Huang, Z., & Wang, W. (2026). Early Post‐ECPR Brain CT for Prediction of Neurological Outcome: An Interpretable Approach. International Journal of Clinical Practice, 2026(1). https://doi.org/10.1155/ijcp/9152075
Vancouver
Wang M, Wang F, Pang G, Wei W, Zhai Y, Huang Z, et al. Early Post‐ECPR Brain CT for Prediction of Neurological Outcome: An Interpretable Approach. International Journal of Clinical Practice. 2026;2026(1). doi:10.1155/ijcp/9152075
BibTeX
@article{Wang2026,
author = {Wang, Man and Wang, Fan and Pang, Guo-Dong and Wei, Wen-Cai and Zhai, Yi-Ling and Huang, Zhou and Wang, Wei},
title = {Early Post‐ECPR Brain CT for Prediction of Neurological Outcome: An Interpretable Approach},
journal = {International Journal of Clinical Practice},
year = {2026},
volume = {2026},
number = {1},
doi = {10.1155/ijcp/9152075}
}
EUVIMED