Phase-Contrast MRI CSF Flow Measurements for the Diagnosis of Normal-Pressure Hydrocephalus: Observer Agreement of Velocity Versus Volume Parameters.

OBJECTIVE: Manual segmentation of the aqueduct for CSF flow analysis may induce measurement variability. The aim of our study was to assess observer agreement of manual segmentation and to compare the repeatability and accuracy of different flow parameters in differentiating normal-pressure hydrocephalus (NPH) from brain atrophy.

SUBJECTS AND METHODS: Thirty-two subjects were included and were divided into three groups: control, NPH, and brain atrophy. Subjects underwent phase-contrast MRI. Quantitative analysis of aqueductal CSF flow using manual ROI placement was performed by two independent readers. Reader 1 repeated measurements 3 months after the first session to assess interobserver and intraobserver agreement. Intraclass correlation coefficients (ICCs), within-subject SD, and repeatability were calculated. Peak systolic velocity (PSV), peak mean velocity, and aqueductal CSF stroke volume, which we refer to as "stroke volume," were recorded and compared between the three patient groups. The ROC curves of diagnostic accuracy for NPH were compared.

RESULTS: PSV was ROI-independent, so only one measurement was obtained. The NPH group had significantly higher PSV, peak mean velocity, and stroke volume values in all readings than both the control and brain atrophy groups. The accuracy of PSV for the diagnosis of NPH was 82.7%, and the accuracy of peak mean velocity was 92.5-93.3% for the three readings. Stroke volume had perfect accuracy of 100% for the three readings. The stroke volume had higher ICCs (0.97 and 0.98) than the peak mean velocity (0.88). The intraobserver repeatability and interobserver repeatability of peak mean velocity were 1.9 cm/s, and the intraobserver repeatability and interobserver repeatability of stroke volume were 27.4 and 19.6 µL/cycle, respectively.

CONCLUSION: Stroke volume had better agreement and repeatability and was more accurate than peak mean velocity for the diagnosis of NPH. PSV lacks variability but was the least accurate.