Using computer vision techniques on CT scans to measure changes in ventricular volume to aid in the diagnosis of hycrocephalus

Abstract

Hydrocephalus causes the size of the cerebral ventricles to change. Current subjective assessment of ventricles by neuroradiologists and neurosurgeons has limited accuracy, because of the complex shape of the ventricular system.

To calculate the volume of the ventricles, an algorithm (CSV) which combined segmentation and volume calculation was developed. The ventricles of a CT scan of the head were segmented and their volumes were calculated. The segmentation method was tested on 15 clinical cases and the similarity index was above 0.7. The volume calculation algorithm was tested on both analytical and physical phantoms. The accuracy of the volume was controlled within an error of 5%. Finally CSV was tested on 8 more clinical cases where the patient had been imaged on two occasions and the outcome diagnosis was known. A volume change of less than 5% in the normal cases and greater than 20% in the cases diagnosed with hydrocephalus was calculated. This shows our algorithm (CSV) is capable of distinguishing hydrocephalus.