We present description and design of a CSF flow phantom which should prove useful in development and validation of flow imaging protocols for measurement of CSF flow. Numerous diseases and pathologies hamper flow of the CSF including hydrocephalus, Chiari malformation, and spinal cord injury.

A 3 feet, .625” inside diameter, .75” outside diameter clear polycarbonate tube, was used as an idealized model of the spinal canal (figure 1). A clear tube was chosen to allow for visual inspection of the Phantom for bubbles and overseeing of the flow of fluid. The cord was modeled by a wooden .3125” diameter dowel rod which ran the entire extent of the tube. These dimensions closely match reported diameter ranges of the human spinal canal (~0.67’’) and spinal cord (~0.31’’). To center the dowel rod inside the canal, winged support structures were created in SolidWorks (figure 2a) and 3D printed to the exact dimension of the polycarbonate tube with a hole having the diameter of the dowel rod precisely centered. These support structures were placed on the ends with a third located 1ft from one end of the phantom. The winged support structures were then glued to the dowel rod. The support structures were placed at these locations so that the flow of the fluid would not be obstructed, and so that the flow would be fully developed distal to the structures where imaging was performed. Two dowel rods were utilized to mimic both a healthy spine and one with a stenosis. The occlusion modeling a 75% area stenosis in the spine was constructed using 3D printing (figure 2b) and glued to the second dowel rod. Experiments were carried out using a closed-loop flow system (figure 2c). An MR-compatible, computer-controlled pump (LB Pump; LB Technology LLC, Louisville, KY) was used with the capability to program user-defined flow waveforms. All flow connectors made use of common PVC couplings and connectors so that a pumping system could be connected to the phantom. It was also designed with screw off connectors for easy changing of the two spinal cords.
Imaging was performed on a Philips Achieva 1.5T scanner (Philips Healthcare, Best, NL) using an 8 element SENSE knee coil. To measure the velocity two 4D flow MRI methods were utilized: conventional 4D flow and 4D Spiral flow [1]. For the conventional 4D flow acquisitions, scan parameters were as follows: FOV = 80*80*60 mm³, TR = 11 ms, TE = 7 ms, tip angle = 6^o, Venc = 20-60 cm/s for all 3 directions, in plane resolutions = 1*1 mm², slice thickness = 3 mm, number of slices=20. For 4D Spiral flow, all parameters were identical other than TE = 3.1 ms and TR = 9.9 ms. In order to provide k-space coverage, 36 spiral arms with 4ms readout each were adopted. Image acquisition time for steady flow were (4D conventional: 1:24, 4D spiral: 0:46) and for pulsatile flow were (4D conventional: 27:46, 4D spiral: 15:38).Table 1 shows peak velocity and flow rate for steady flows at 4, 8, and 12 ml/s. Results illustrate that both acquisition methods measure velocity profiles with similar accuracy, however there is a light underestimation in Spiral acquisition. Flow rate variation inside the phantom through z direction was measured as an indicator of noise in the acquisition. Results show that 4D Spiral flow is more robust encountering phase contrast noise.In this work we have reported initial results of using two 4D flow techniques on two CSF flow phantoms designed. At these low velocity and low flow rates, results indicate good agreement between the two 4D flow methods, however there is a slight underestimation of velocities with 4D Spiral Flow. It should be mentioned that relative to 4D conventional flow, 4D spiral flow reduced the scan time by around 40%.