Detecting Hydrocephalus Using CT Scans and Machine Leaning Techniques

Every year, hydrocephalus affects many people, whether they are infants or adults with higher incidence in pediatric group. This means that a large number of families affected by this disease, and automatic determination of hydrocephalus in medical images could help people who are affected with this disease and their condition is at a dangerous stage. In this research, we try to implement an algorithm that could detect hydrocephalus from CT images. Our work is divided into two stages of implementing the proposed algorithm and comparing it with the results of similar researches. We try to design and implement all steps from scratch. The proposed model reaches the result with appropriate accuracy but slower running time. Our Method is able to detect hydrocephaly from 40 patients CT Images and calculate ventricular indices with accuracy of 64% to 93% in comparison with manually extracted ventricular indices by a neurosurgeon. In addition, we also checked the K-nearest neighbor (KNN) and Watershed methods and concluded that Region Growing is the best among this category.