Objective digital determination of CSF spaces is discussed, with ventricular and subarachnoid spaces handled separately. This method avoids the difficulty of visual definition of ventricular borders in planimetric measurements. The principle is to count automatically all pixels corresponding to CSF in a given region with a Hounsfield unit and to multiply this number by the pixel size. This will give the total surface area of CSF spaces in square millimeters. The calculation of pixel values for CSF spaces and brain tissue is experimentally formulated taking the intersection of the Gaussian curves for ventricular content and brain tissue. In practice, the determination of CSF spaces is done by first calculating a histogram of the total brain in a given slice defining all CSF spaces. Next a histogram of a region including ventricles with adjoining tissue is calculated and the ventricular size is calculated. By subtraction of the ventricle value from the total CSF space value, the subarachnoid space size is obtained. The advantages of this mehtod will be discussed.