Automatic ventricle segmentation using quantitative MRI

Motivation: In clinical routine the size of the ventricles is generally estimated using subjective visual inspection. A complicating factor for an automated ventricle segmentation is the variability and inhomogeneity of conventional image intensity. Goal(s): A fully automatic ventricle segmentation was developed which uses quantitative CSF maps as input. The aim of the study was to investigate the accuracy of the algorithm compared to manual segmentation. Approach: 47 Normal Pressure Hydrocephalus (NPH) patients were measured with quantitative MRI before and after shunt operation. Also a group of 11 normals was acquired. Results: On average, both methods differed with 1 ml; The Pearson correlation was 1.0. Impact: A fully automatic ventricle segmentation, using quantitative CSF maps as input, showed a high accuracy compared to manual segmentation. This allows a far more objective and accurate follow-up for Normal Pressure Hydrocephalus patients than the current method of visual inspection.