MACHINE LEARNING-DRIVEN RISK STRATIFICATION OF POST TRAUMATIC HYDROCEPHALUS BASED ON EARLY CRANIAL CT MORPHOMETRY

Abstract

Post-traumatic hydrocephalus (PTH) is a frequent complication 
following traumatic brain injury (TBI), potentially leading to delayed neurological 
deficits, cognitive decline, and functional impairment. Early detection is critical to 
enable timely intervention, prevent secondary brain injury, and improve outcomes. 
This study aimed to develop a predictive model integrating early ventricular 
morphometry derived from CT scans and key clinical variables. Retrospective data 
from 180 patients with moderate TBI (GCS 9–12) treated between 2021 and 2024 were 
analyzed. Morphometric indices, including Evans Index (EI), Bicaudate Index (BI), 
and Fronto-Occipital Horn Ratio (FOHR), were measured at the level of the foramen 
of Monro and combined with clinical parameters (age, sex, Glasgow Coma Scale 
[GCS], intraventricular hemorrhage [IVH], and decompressive craniectomy [DC]). 
Machine learning models, including gradient boosting, random forest, and logistic 
regression, predicted PTH occurrence within 6 months. PTH developed in 40 patients 
(22.2%). Gradient boosting achieved the highest predictive performance (AUC=0.88), 
with EI, IVH, DC, and FOHR identified as the most important predictors. This 
integrative approach facilitates early identification of high-risk patients and supports 
individualized monitoring strategies.