Key points are not available for this paper at this time.
Presented here is a modification of a previously developed, computed, semiautomated system for quantifying the amount of CSF and tissue on brain CT scans. The technique automatically strips skull from brain and applies a two-dimensional high-pass filter to reduce spectral-shift artifact. The resultant images are presented along with the raw images, section by section, on a screen for fluid-tissue differentiation by a human scorer. This is accomplished by adjusting a one-bit display of the filtered image until a satisfactory separation threshold is found. Data can be summed within or across sections to provide measures of CSF volume in specific regions of interest such as the ventricles and sulci. A major advantage is the improved accuracy of sulcal measurement. This method has been applied to scans of 57 community volunteers, 20-84 years old. High correlations have been established between rankings performed by an experienced clinician and automated rankings (r = 0.88 with rankings based on ventricular size, and r = 0.83 with rankings based on sulcal size). Furthermore, significant correlations were found between computed measures of ventricular size and age and between sulcal widening and age.
Pfefferbaum et al. (Tue,) studied this question.