We constructed a phantom for data acquisition and developed a standardized method to calibrate PC-MRI for accurate shunt flow assessment on any MR scanner. Our protocol applies phantom calibration to evaluate variation in PC-MRI measurements introduced by different MR hardware and validates multisite PC-MRI data collection. Rigorous methods for statistical comparison between types of scanner hardware are also provided.

We constructed and evaluated a versatile suite of phantoms (Multimedia Appendix 1). A single-tube, linear phantom was used for data collection based on its simple design and reproducible flow production. The single-tube phantom was constructed using an antibiotic-impregnated catheter with a 1.3-mm inner diameter (Ares model; Medtronic), a common surgically placed distal shunt catheter in patients at our institution. The catheter was secured longitudinally through a rectangular 500-mL biologic media flask filled with water to eliminate air pockets and optimize the magnetic homogeneity of the imaged region.

An Alaris screw-driven syringe pump (Becton, Dickinson and Company) provided modifiable flow rates to drive water through the phantom in a continuous fashion without peristaltic or pulsatile action. The pump was connected to one end of the phantom using small-bore Luer lock tubing (ICU Medical). Materials and a step-by-step procedure for phantom construction are provided in Multimedia Appendix 1.

We calibrated the combination of flow phantom and pump before data collection to ensure accurate and precise volumetric fluid delivery during subsequent testing (Figure 1). An analytical balance (AG204 DeltaRange; Mettler Toledo) was used to collect fluid output from the shunt system at 10 predetermined rates spanning the expected physiological range of CSF flow (0-22 mL per hour). The catheter was oriented flat on the benchtop to allow the liquid to drip freely, avoiding potential siphoning effects from gravitational pressure differences. Each pump setting was tested 3 times against the corresponding volumetric fluid output to evaluate the reproducibility of the pump’s flow.

PC-MRI flow acquisition is suggested with the following base sequence parameters: field of view of 30 × 30 mm, slice thickness of 10 mm, 2 signal averages, matrix size of 92 × 91, acquisition voxel dimensions of 0.20 × 0.20 mm², reconstructed voxel dimensions of 0.10 × 0.10 mm², compressed sensitivity encoding factor of 2, and velocity encoding of 1 cm per second [21].

During image acquisition, the phantom is advised to be positioned parallel to the length of scanner bore and level with the MR bed. The screw-driven pump, located outside the MR room, is attached to the phantom, and connection tubing is carefully positioned to remain free of kinks or air (Figure 2). Triplanar (coronal, axial, and sagittal) T2-weighted images are used for PC-MRI slice localization, and PC-MRI image planes are placed orthogonally (transverse) to the direction of catheter flow such that the catheter lumen appears circular on phase-contrast images. The screw-driven pump is used to establish flow through the phantom at 10 benchtop-tested rates (0, 1, 3, 5, 8, 11, 14, 17, 19, and 22 mL/h) chosen for their overlap with the typical physiological range of CSF production. Four repeat PC-MRI images are obtained at each flow rate. Images are saved as Digital Imaging and Communications in Medicine files and exported to an external processing platform for flow measurement in a typical fashion.

To facilitate multi-institutional implementation, this acquisition and processing protocol is designed to be reproducible across centers with varying MR hardware configurations. Home institution plans for multiscanner calibration include repeating this procedure on 8 MR scanners in which field strength, gradient slew rate, gradient strength, radiofrequency coil type, model, and manufacturer vary (Table 1). Sites participating in multicenter data collection may apply this protocol to calibrate scanners for use in PC-MRI data acquisition.

We established three analytical approaches to measure volumetric flow from each PC-MRI image: (1) mean intensity within a circular region of interest prescribed over the catheter lumen; (2) maximum intensity within a small circular region of interest at the center of the catheter lumen, thereafter converted via a formula to total flow; and (3) full parabolic model of a laminar flow profile within the lumen (Figure 3). A description of each method of analysis and equations for flow calculation are provided in Multimedia Appendix 2.

For each scanner and hardware combination, linear regression will be used to evaluate the relationship between PC-MRI–derived flow rates and true flow outputs from the pump. An ANOVA will compare mean error between datasets, with a significance level set at α=.05. These statistical approaches will be repeated to evaluate the precision and accuracy of each analytical technique for flow measurement. Descriptive statistics will also be used to examine the effect of variables such as gradient strength, field strength, and slew rate on calibration measurements as these variables often covary with MR scanner.

We aim to validate the feasibility of implementing multisite PC-MRI data collection platforms by evaluating PC-MRI image quality and logistical factors such as scheduling feasibility, personnel training requirements, and processes for data transfer. Image quality will be reported as a proportion of analyzable scans within the dataset. Logistical factors surrounding barriers to personnel training or PC-MRI integration into scan protocols will be described individually by each site.

Patients who undergo a PC-MRI flow study will be included in the cohort regardless of age, etiology upon presentation, clinical setting (emergency department, inpatient, or outpatient), or MR hardware used during image acquisition. Patients with distal slit valves, valveless shunts, cystoperitoneal shunts, subdural shunts, and shunts with non–MR-resistant valves will be excluded. To maintain independence among data points, only the first PC-MRI flow measurement for each patient will be included in the analysis of shunt flow over a population of patients. Individuals with repeat PC-MRI measurements during unrelated clinical encounters will be analyzed separately to evaluate longitudinal trends in CSF flow.

Clinical PC-MRI images will be collected across sites following standardized acquisition and processing protocols. A single PC-MRI sequence will be incorporated into routine and emergent full-length structural brain MRI exams for shunted patients at participating institutions. At each site, 2 independent data abstractors will manually review patient charts and clinical records according to predetermined standard operating procedures. Demographic, clinical, and imaging data (Multimedia Appendix 3) will be extracted from the electronic health records and collected using the REDCap (Research Electronic Data Capture; Vanderbilt University) tool [24]. PC-MRI images in which flow data cannot be measured (eg, low resolution, improper shim, and nonorthogonal image plane prescription) will be reported as “missing” and accompanied by documentation of image quality issues. Each institution will use local REDCap collection forms, and periodic queries for data validation will take place at 4-month intervals. For analysis, data will be merged between institutions via secure file transfer with Health Insurance Portability and Accountability Act (HIPAA) compliance. Datasets will be stored for 10 years after the study.

Home institution plans for PC-MRI imaging and data collection have been approved by the Children’s Hospital Los Angeles Institutional Review Board (IRB; CHLA-20-00041), and external sites will rely on local IRB approval along with data sharing agreements. As data for this study will be acquired retrospectively, a waiver of consent will be obtained; patient recruitment and informed consent before imaging will not be necessary. Identifying information will be correlated with unique study identification numbers stored on a local encrypted server.

MRI is generally considered a safe imaging technique, with primary concerns related to implantable devices and projectile risks from ferromagnetic objects [25,26]. All patients will undergo standard MRI safety screening before imaging.

To evaluate the feasibility of implementing a multisite PC-MRI data collection platform, we plan to enroll 100 participants across 3 sites. This target is based on our preliminary work and is expected to be feasible based on site selection and the number of eligible patients over the study period. A formal sample size calculation was not performed as the planned multisite study is focused on feasibility rather than hypothesis testing or estimation of effect sizes. The sample size is intended to ensure adequate experience with clinical data collection and will allow us to identify challenges related to PC-MRI acquisition, image quality, and flow measurement.

PC-MRI can provide additional utility in the diagnosis of shunt malfunction by offering a rapid and noninvasive approach to accurately quantify shunt flow. Our protocol presents a framework to calibrate individual MR scanners for accurate PC-MRI flow measurement in multi-institutional clinical data collection. Standardizing PC-MRI measurements and accounting for error introduced by variable MR hardware will allow individual institutions to perform appropriate calibration corrections. The ability to acquire reliable flow measurements between imaging centers can improve the validity of comparison between patient population flow values even when imaging is performed at different sites. Characterizing shunt flow across a large sample during multicenter data collection may improve the clinical assessment of shunt failure and our understanding of shunted hydrocephalus.

We designed a transportable and easy-to-use phantom that ensures magnetic homogeneity for reliable data acquisition. The single-catheter phantom and screw-driven syringe pump combination enables simple, consistent data collection across a physiological range of flows. To represent human shunt flow conditions, our phantom included a common catheter model used in patients at our institution, and water acted as a surrogate for CSF because both fluids have similar magnetic susceptibility and viscosity [27-29]. While CSF flow is known to vary with cardiac and respiratory cycles [30,31], PC-MRI measurements in this study will not be physiologically gated [21]. The potential influence of cardiac [32] or respiratory gating [33] on shunt flow measurement has not yet been studied and remains a potential area of future investigation.

Understanding the reproducibility of PC-MRI flow measurement across MR scanners is useful both to understand the influence of hardware parameters on measured shunt flow and to implement an appropriate calibration correction as needed. MR scanners can vary in field strength, slew rate, gradient strength, number of radiofrequency channel coils, model, and manufacturer at different institutions. This method for phantom model calibration will allow institutions to confirm the reliability of their MR scanners for accurate PC-MRI flow measurement. The protocol outlined in this work aims to streamline the implementation of PC-MRI by preemptively calibrating the scan sequence on a range of scanners and identifying challenges to multi-institutional data collection.

As MR scanner time is primarily reserved for clinical use at most institutions, obtaining access for scanner calibration via phantom data collection remains a potential barrier. As PC-MRI images must be exported and externally processed by trained study personnel, the study may be limited by staffing constraints and processing delays. Outside of error introduced by MR hardware, it is important to note that human physiology (eg, susceptibility of the surrounding tissue [34] or altered pulsatility [30,35,36]) and movement within the scanner may provide additional potential variability. While this protocol aims to isolate the hardware-introduced variability in flow measurement, future research should investigate real-world sources of variation in flow measurement.

This work details a protocol for multi-institutional clinical PC-MRI data collection. We also describe the construction of an easy-to-use phantom model of shunt flow and an interinstitutional procedure for scanner calibration to enable standardized collection of PC-MRI flow measurements. The tested phantom and pump combination exhibits accurate and precise outputs at both low and high flow rates while minimizing magnetic field inhomogeneities. An appropriate next step in progress at our institution and forthcoming at others is to generate scanner-specific calibration curves by acquiring PC-MRI data using the single-catheter phantom design. Widespread scanner calibration and validation of clinical feasibility will enable institutions to reliably implement PC-MRI, providing a quantitative method to evaluate shunt function in a broader patient population.