Abstract
Purpose :
High intracranial pressure (ICP) is a condition that causes structural eye changes in the cerebrospinal fluid (CSF) and optic nerve sheath (ONS) that are visible on magnetic resonance images (MRI). These may have relevance for diagnosis and monitoring of high ICP and have been previously measured using manual measurements on 2D images.
Our goal is to develop and validate reliable 3D measurements of CSF volume in the ONS from MRI.
Methods :
Two raters performed measurements of CSF Volume in ONS using MRIs of 4 eyes (2 normal ICP, 2 high ICP) as follows:
Using Scan IP (Synopsys Inc.), a full-width, half-maximum (FWHM) approach was applied to identify boundaries between the sclera/vitreous, ONS/CSF, and optic nerve (ON)/CSF on 3D Fast Imaging with Steady State Acquisition (3D FIESTA-C) MRI scan DICOM images (0.3516mm x 0.3516mm x 0.5mm).
The ellipsoid of best fit was computed over the vitreous and the y-axis Euler angle was used to re-align the image. The centerline of the CSF was calculated to identify the anterior boundary of the CSF, represented as the Y0 point. The CSF was resampled to isometric 0.01mm3 and cropped at 3mm posterior to the anterior boundary. The volume of the cropped CSF mask was measured.
Inter and intra rater reliability were calculated.
Results :
For a single rater performing CSF volume measurement on 2 occasions, 3mm CSF volumes varied by 0.2mm3 (0.192%) and 0.7mm3 (0.562%) respectively in high ICP eyes and 0.7mm3 (0.637%) and 0.2mm3 (0.197%) respectively in normal ICP eyes. For two different raters performing CSF volume measurements, 3mm CSF volumes varied by 0.8mm3 (0.767%) and 0.6mm3 (0.480%) respectively in high ICP eyes and 0.5mm3 (0.455%) and 0.9mm3 (0.882%) respectively in normal ICP eyes.
Conclusions :
Our method to measure CSF Volume from 3D FIESTA-C MRI scans shows excellent inter/intra operator reliability. Further work will apply this methodology to studying the utility of ONS CSF volume for detection and monitoring of high ICP.
This is a 2020 ARVO Annual Meeting abstract.