The relationships among spinal CSF flows, spinal cord geometry, and vascular correlations: evidence of intrathecal sources and sinks.

We studied relationships of cerebral spinal fluid (CSF) pulsatile flow at cervical, thoracic, and lumbar levels using phase-contrast cine MRI (PCCMRI) to determine the following: ) instantaneous and average net flows at cervical, thoracic, and lumbar levels, ) stochastic correlations of CSF flow with major arterial supplies and major draining veins, and ) whether adjustments of cord-flow curves-using cord cross-sectional areas, caudal lengths, and caudal volumes-would normalize flow curves from different levels. We scanned 15 healthy volunteers without anesthesia, ages 23-46 yr, using external, retrocardiac-gated, two-dimensional PCCMRI at 3T. Transverse scans of the subarachnoid space, arteries, and veins were acquired and analyzed at cervical, thoracic, and lumbar levels. Instantaneous CSF flow decreased craniocaudally along the full time course of a cardiac cycle. Downward net flow generally increased craniocaudally. During diastole, instantaneous CSF flow decreased proportionally to cross-sectional area, caudal residual length, and caudal residual volume of the cord. The proportionalities were less consistent during systole. CSF, internal carotid artery (ICA), vertebral artery, and lower aorta temporal correlations were highest in systole and decreased craniocaudally. CSF flow temporally correlated better with lower aorta flow than with the ICA at T7 and L2 during systole but not diastole. Inferior vena cava temporal correlation increased craniocaudally. We conclude that whereas instantaneous flow is attenuated cranial caudally, net downward flow, per cardiac cycle, increases caudally, becoming statistically significant at T7 and below the conus medullaris. We can explain the results with the assumption of cord CSF production and peripheral-dominated CSF absorption.