3D Quantitative MRI: A Fast and Reliable Method for Ventricular Volumetry

Abstract

Purpose

Volumetry of cerebral ventricles is a far more sensitive measure for shunt-induced reduction of ventricular size than traditional 2-dimensional (2D) measures, such as Evans index. However, available ventricle segmentation methods are time-consuming, resulting in limited use in clinical practice. Quantitative MRI (qMRI) obtains objective measurements of physical tissue properties, enabling automatic segmentation of white and gray matter and intracranial cerebrospinal fluid. The aim of this study was to evaluate the reliability and processing time of both manual and manually corrected automatic ventricular volumetry through the application of 3D qMRI.

Methods

An independent examiner performed manual ventricular volumetry segmentations on 45 3D qMRI acquisitions (15 healthy individuals, 15 idiopathic normal pressure hydrocephalus (iNPH) patients, 15 shunted iNPH patients) twice. Another independent examiner manually segmented 15 of these acquisitions once. An automatic ventricle segmentation algorithm generated a third set of ventricular segmentations for all 45 data sets. The automatic segmentations were then corrected by both examiners to obtain a fourth set of data. All segmentations were assessed for intra- and interobserver reliability.

Results

Intra- and interobserver reliability for all segmentations, manual, corrected, and automatic, was excellent (intra-class correlation coefficient 1.000, 1.000 and 0.999 respectively). Ventricular volumes were on average 42 ± 18 mL (mean ± SD) in healthy individuals, 140 ± 34 mL in iNPH patients, and 113 ± 35 mL in shunted iNPH patients.

Conclusions

3D qMRI is a reliable and time-efficient method to obtain relevant volumetric measures of intracranial cerebrospinal fluid spaces for both clinical and research purposes. The corrected automatic segmentations provide a feasible time expenditure for clinicians caring for patients with iNPH.