Comparison of Computed Tomography 3-Dimensional Volumetric Analysis of Ventricular Size to Visual Radiological Assessment

Abstract

Interpretation of ventricular volume on computed tomography scans of hydrocephalus patients is usually subjective. The objective of this study was to determine whether radiological assessment of interval change correlates better with an objective calculated volume change or with other objective 2-dimensional estimates of ventricle volume change. Ventricular volume, Evans ratio, and frontal and occipital Horn ratio were retrospectively assessed on 95 pairs of scans from patients with a ventriculoperitoneal shunt. To determine ventricle volume, all voxels of cerebrospinal fluid density were isolated on a 3-dimensional reconstructed computed tomography scan. Voxels of fluid density contiguous with one another in the ventricular system were isolated. Radiological assessments of interval change were divided into 5 groups based on reported findings in the radiology report. The 95% mean confidence intervals were developed for changes in the measured parameters, given a particular radiological assessment. Multinomial regression was subsequently performed to determine which parameter was most closely correlated with the radiological assessment. Significant overlap was found in the confidence intervals for objectively calculated volume change between the different categories of radiological assessment. The frontal and occipital Horn ratio had the most consistent correlation with the radiological assessment, followed by the Evans ratio. Objectively calculated volume change correlated poorly with radiological assessment. Radiological interpretation does not correlate well with objectively calculated volume changes, but correlates better with other parameters that approximate volume and are likely used to visually evaluate interval change. We recommend that ventricle volume be objectively measured to increase consistency between radiological interpretation and actual interval changes.