Development of an interpretable ventricular disproportion index–disproportionately enlarged subarachnoid space hydrocephalus model for differentiating idiopathic normal pressure hydrocephalus from cerebral small vessel disease
Journal of Brain and Spine · 2026-10
DOI: 10.1016/j.jbs.2026.09.002 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Brain and Spine, 2026-10. Verfasst von Yiwei Feng und 13 weiteren Autor:innen. Die Arbeit nennt 38 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0). Gefördert von: National Natural Science Foundation of China, National Science and Technology Major Project.
Abstract (Originalsprache)
Background and aims Distinguishing idiopathic normal pressure hydrocephalus (iNPH), a treatable neurological disorder, from cerebral small vessel disease (CSVD) is a critical clinical imperative, yet this distinction remains highly challenging owing to substantial overlap in clinical symptoms and radiological manifestations. In this retrospective single-center study, we aimed to develop and internally validate an interpretable, imaging-based machine learning model to differentiate iNPH from CSVD.
Methods
We included a clinically representative, imbalanced cohort of 205 patients (21 with iNPH; 184 with CSVD) and constructed a parsimonious model (ventricular disproportion index-disproportionately enlarged subarachnoid-space hydrocephalus, VDI-DESH) utilizing four core hydrocephalus-related radiological markers: Evans’ index, third ventricle width, temporal horn enlargement, and DESH score. Its performance was rigorously evaluated against a full-feature (18-variable) Random Forest model using repeated stratified 5-fold cross-validation (500 outer test folds), with model interpretability assessed via Shapley Additive exPlanations (SHAP) values.
Results
Patients with iNPH exhibited significantly more pronounced hydrocephalus markers but paradoxically presented with a higher CSVD burden (e.g., more frequent lacunes and cerebral microbleeds (CMBs)) than the CSVD cohort, highlighting severe diagnostic overlap. Across cross-validation folds, the Random Forest model achieved a higher area under the receiver operating characteristic curve (ROC-AUC) (0.9539 vs. 0.9332) and average precision (0.7653 vs. 0.6408). However, the compact VDI-DESH model offered a highly desirable, sensitivity-oriented operating profile (sensitivity: 0.9011 vs. Random Forest: 0.8638), which is crucial to avoid missing treatable iNPH cases. Owing to the low prevalence of iNPH, positive predictive value remained modest across models, and integrating overlapping clinical covariates did not further improve predictive performance. Furthermore, SHAP analysis confirmed that all four VDI-DESH variables contributed synergistically to the predictions.
Conclusions
The VDI-DESH model provides a simple, highly interpretable, and sensitive imaging-based approach for differentiating iNPH from CSVD in imbalanced clinical settings. It serves as a practical screening and decision-support tool to prioritize patients for advanced iNPH evaluations, warranting future external validation.
Förderung
- National Natural Science Foundation of China
- National Science and Technology Major Project
Zitieren
APA
Feng, Y., Chen, G., Zheng, L., Maimaituxun, G., Shi, Z., Liu, X., Liu, S., Xu, H., Men, X., Qin, B., Yan, W., Wang, J., Lu, Z., & Wei, L. (2026). Development of an interpretable ventricular disproportion index–disproportionately enlarged subarachnoid space hydrocephalus model for differentiating idiopathic normal pressure hydrocephalus from cerebral small vessel disease. Journal of Brain and Spine. https://doi.org/10.1016/j.jbs.2026.09.002
Vancouver
Feng Y, Chen G, Zheng L, Maimaituxun G, Shi Z, Liu X, et al. Development of an interpretable ventricular disproportion index–disproportionately enlarged subarachnoid space hydrocephalus model for differentiating idiopathic normal pressure hydrocephalus from cerebral small vessel disease. Journal of Brain and Spine. 2026. doi:10.1016/j.jbs.2026.09.002
BibTeX
@article{Feng2026,
author = {Feng, Yiwei and Chen, Guoming and Zheng, Lu and Maimaituxun, Gulijianaiti and Shi, Zhijing and Liu, Xu and Liu, Sanxin and Xu, Huiming and Men, Xuejiao and Qin, Bing and Yan, Wei and Wang, Jingjing and Lu, Zhengqi and Wei, Lei},
title = {Development of an interpretable ventricular disproportion index–disproportionately enlarged subarachnoid space hydrocephalus model for differentiating idiopathic normal pressure hydrocephalus from cerebral small vessel disease},
journal = {Journal of Brain and Spine},
year = {2026},
doi = {10.1016/j.jbs.2026.09.002}
}
EUVIMED