Related Experiment Video
Updated: Jun 13, 2026
Automated Midline Shift and Intracranial Pressure Estimation based on Brain CT Images
Published on: April 13, 2013
Clinical Subgroups and Treatment Outcomes in Idiopathic Normal Pressure Hydrocephalus: Application of Machine
Emalee J Burrows1,2, Linda D'Antona1,2, James Booker1,3
1UCL Queen Square Institute of Neurology, London, UK.
Background And Objectives:
The variability in clinical phenotype, radiological features and treatment outcomes of idiopathic normal pressure hydrocephalus (iNPH) poses significant challenges for diagnosis and management. This study aimed to use artificial intelligence techniques, specifically unsupervised machine-learning clustering, to identify meaningful clinical subgroups and outcome trajectories after ventriculoperitoneal shunt insertion.
Methods:
In this retrospective single-center case series, clustering methods were applied to characterize patients with shunt-responsive iNPH, based on demographic, clinical, and radiological variables. Clinical data included preoperative symptoms and postoperative shunt response and long-term outcomes. All radiological data were extracted by 2 senior neuro-radiologists blinded to the clinical data.
Results:
A total of 187 patients were identified, with a mean age of 75.6 ± 6.6 years, and comprising 69 (37%) women. Patients were followed up for a median duration of 28 months [IQR, 16-59]. Ward's hierarchical agglomerative clustering identified 4 clearly defined subgroups with distinct clinical outcome trajectories. Individuals with probable iNPH typically aligned with one of the following profiles: (1) disproportionately enlarged subarachnoid space hydrocephalus-positive, (2) dilated sylvian fissure hydrocephalus, (3) small-vessel disease with ventriculomegaly, and (4) marked ventriculomegaly.
Conclusion:
These findings challenge the traditional view that iNPH is a single disease and provide new insights into its pathophysiology. While disproportionately enlarged subarachnoid space hydrocephalus remains a valuable indicator of shunt responsiveness, its low negative predictive value highlights the importance of considering additional iNPH phenotypes. Incorporating cluster-based phenotyping into clinical pathways may improve prognostic prediction, enhance patient counselling before shunt surgery, and enable more targeted management strategies. Attention to such subgroups should be considered in future iNPH research.
