Objective: Hydrocephalus which is defined as excessive accumulation of cerebrospinal fluid in brain, is one of the most critical entities in pediatric neurosurgery; as it could result in severe developmental delay and neurologic deficits. Imaging evaluation of brain ventricular system is important for clinical decision-making in pediatric hydrocephalus. Manual measurement of linear ventricular indices is encumbered with human errors. Automatic measurement of these parameters would increase measurement reliability. The authors aimed to develop an accurate automatic algorithm for calculating pediatric, ventricular indices in brain computerized tomography)CT (scan slices. Methods: The study cohort consisted of 30 brain CT scans of hydrocephalic children treated in Mofid Children hospital, Tehran, Iran. Various indices of ventricle were calculated automatically using GrabCut image segmentation algorithm for each patient and compared with manual assessment data which were measured by pediatric neurosurgeon. Data accuracy was assessed using Student’s t-test. Model application in asymmetrical ventricular anatomy and in patients with history of shunt insertion was assessed in the last step. Results: The proposed model was able to measure ventricular indices by accuracy of 94 to 97 %. It was also reliable in asymmetrical brain ventricles and presence of shunt hardware. Our results showed there were not any significant statistical differences between these patients and other patients (asymmetrical brain ventricles with p ≥ 0.3190 and presence of shunt hardware with p ≥ 0.1489). Conclusion: The authors present an automatic intelligence-based model for calculating different ventricular indices in pediatric hydrocephalus population with acceptable accuracy. This model could be adopted to the realtime clinical evaluation of hydrocephalus and improve clinician work flow in emergent neurosurgical situations or in developing countries, where brain CT scan is the most common brain imaging modality.
