Abstract
BACKGROUND: In textbooks and literature, intracranial cerebrospinal fluid (iCSF) volume is typically presented as a generalized range without population-specific references. The volume of iCSF in normal populations varies with age and brain parenchyma volume, and exhibits sex-related differences. This study aimed to develop an automated extraction model for quantitatively measuring iCSF volume using brain non-contrast computed tomography (CT) images, and to establish reference values for iCSF in populations of varying age and brain atrophy groups based on sex. METHODS: Brain non-contrast CT images from 400 individuals were retrospectively collected. The iCSF was delineated manually slice by slice. The UNet Transformers (UNETR) architecture was employed to develop an iCSF extraction and automatic quantitative measurement model. Subsequently, the model was used to extract iCSF automatically from 1,600 cases, and reference values were set for different ages and degrees of brain atrophy for males and females. RESULTS: The automatic extraction and quantitative volume measurement model for iCSF achieved a Dice coefficient 0.78. In age groups of ≤44, 45-59, 60-74, and ≥75 years, the reference values of total iCSF volume for males were 62.0, 91.2, 116.9, and 127.0 mL, respectively, and for females were 59.1, 91.2, 113.6, and 122.4 mL, respectively. In the no, mild, moderate, and severe brain atrophy groups, the reference values of total iCSF volume for males were 58.4, 79.6, 99.1, and 127.4 mL, respectively, and for females were 59.1, 78.4, 98.0, and 125.1 mL, respectively. CONCLUSIONS: Given the good performance of the model, the established reference values for iCSF volume could significantly aid research related to iCSF.