Quantitative Non-Invasive Measurement of Cerebrospinal Fluid Flow in Shunted Hydrocephalus

Structured Abstract

Objective:

Standard MRI protocols lack a quantitative sequence that can be used to evaluate shunted patients with history of hydrocephalus. The objective of this study was to investigate the use of phase-contrast MRI (PC MRI), a quantitative MR sequence, to measure CSF flow through the shunt. We report the results from our study, which include: (i) MR flow phantom calibration, (ii) determination of the distal catheter as the optimal location for PC MRI image acquisition, (iii) in vivo validation in a patient with preserved shunt valve and externalized distal shunt, (iv) successful PC MRI measurements in 20 consecutive pediatric patients. The aim of the work is to demonstrate PC MRI as a useful adjunct in the clinical monitoring of shunted patients.

Methods:

The rapid (96-second) PC MRI sequence was calibrated using a flow phantom with known flow rates ranging from 0–24 mL/hr. Following phantom calibration, 20 patients were scanned with the PC MRI sequence. We gathered multiple, successive proximal and distal measurements in five patients to test for measurement error in different portions of the shunt system and to determine intra-patient CSF flow variability. The study also includes the first in vivo validations of PC MRI for CSF shunt flow by comparing PC-measured flow rate to CSF accumulation in a collection burette from patients with externalized distal shunts.

Results:

The PC MRI sequence successfully measured CSF flow rates ranging from 6–19 mL/hr in 20 consecutive pediatric patients. Comparison of PC MRI flow measurement and CSF volume collected in a bedside burette showed good agreement in a patient with externalized distal shunt. Notably, the distal portion of the shunt demonstrated lower measurement error when compared to PC MRI measurements acquired in the proximal catheter.

Conclusions:

The PC MRI sequence provided accurate and reliable clinical measurements of CSF flow in shunted patients. Our work provides the necessary framework to include PC MRI as an immediate addition to the clinical setting in the noninvasive evaluation of shunt function and in future clinical investigations of CSF physiology.

Introduction:

Hydrocephalus is a pathological accumulation of cerebrospinal fluid (CSF) within the brain resulting from a disruption of cerebral fluid production and absorption.^1–3 Over 300,000 patients are newly diagnosed with hydrocephalus annually worldwide,⁴ and treating hydrocephalus leads to approximately 40,000 hospitalizations⁵ and 18,000 shunt surgeries annually in the U.S.⁶ CSF shunts are a common treatment for hydrocephalus, and up to 40% of implanted shunts malfunction within two years after implantation.⁷ Shunt failures are the second most prevalent cause for hospital re-admission within 30 days.⁸ A 20-year retrospective study identified that over 84% of shunted patients require at least one revision surgery.⁹ Therefore, methods that better evaluate shunt flow have a broad clinical interest.^10–14

Patients with hydrocephalus would stand to benefit from techniques that measure shunt flow, because that data might (i) reveal shunt malfunction in the absence of increased ventricular size,^15–17 (ii) guide patient and family counseling or reassurance, (iii) lead to shunt design improvements, and (iv) fill gaps in knowledge about patient-specific or age-related shunt function and CSF physiology. However, we lack methods for safe, non-invasive, repeatable measurement of shunt flow. Addressing this unmet need would improve clinical management and facilitate further investigation in this field of clinically-applied basic human physiology.

Several MRI-based techniques assess physiological fluid flow in vivo,^18–20 and phase contrast magnetic resonance imaging (PC-MRI) is an accurate and precise technique for measuring flow.²¹ Reported uses of PC-MRI include measuring CSF through the intracranial cerebral aqueduct,²² cerebral blood flow through arteries of the neck,²³ and cardiac output through the aorta.²⁴ PC-MRI is rapid, noninvasive, and requires no additional MRI hardware. Prior reports (Table 1) demonstrate PC-MRI shunt flow measurements, however, these studies were relatively small (3–5 patients) and contain no in vivo measurement validation.

Table 1:

Literature Results using PC-MRI to measure Shunt Flow, Compared with this Study

Study (Year)	Calibration	in vivo Measurement			PC Study Parameters
	# Flow Values Measured	Reported CSF Flow; # Subjects (n)	Subject Age Range (years)	Measured Flow Range (mL/hr)	Acquisition Time (s)	Voxel Dimensions (mm)	Field Strength (T)	Venc (cm/s)	Acquisition Location
Martin et al., 198929	7	5	17 – 40	4 – 19	480	0.16 × 0.16 × 10	1.5	Not reported	Unspecified
Drake et al., 199130	9	3	0.17 – 28	3 – 40	480	0.16 × 0.16 × 10	1.5	Not reported	Distal
Kurwale & Agrawal, 201127	N/A	N/A	0.25 – 15	N/A	360	0.31 × 0.31 × —	1.5	10	Proximal
Zhang et al., 201931	N/A	N/A	19 – 64	N/A	360–720	0.06 × 0.08 × 4	3	Not reported	Distal
König et al., 202028	6	N/A	27 – 76	N/A	118	0.3 × 0.3 × 10	3	.3	Proximal
Current Study	20	20	1–19	6–19	105	0.33 × 0.33 × 10	3	1	Distal

Our objective was to evaluate PC-MRI-based shunt flow and address the limitations of prior studies by (i) calibrating our PC-MRI with a CSF shunt flow phantom, (ii) measuring CSF shunt flow data for a large group of patients, (iii) evaluating and resolving potential sources of error that from temporal and spatial variation of CSF flow in the proximal (pre valve) or distal (post valve) catheters, and (iv) validating PC-MRI versus timed collection of CSF, in vivo. Thus, our investigation aims to resolve all obstacles to clinical translation and allow PC-MRI to address a significant unmet clinical and academic need.

Methods:

Design of Phase Contrast MRI Acquisition

We acquired all MRIs on two clinical 3T MR scanners (Philips Achieva, Best, the Netherlands). PC-MRI sequence parameters were: TR=47 ms; TE=30 ms; field of view (FOV)=30×30 mm; slice thickness=10 mm; NSA=2; matrix, 92 × 91; acquisition x-y voxel dimensions=0.20×0.20 mm²; reconstruction x-y voxel dimensions=0.10×0.10 mm²; bandwidth=48 Hz/pixel; scan duration=105 s; and velocity encoding gradient (V_enc)=1 cm/s, selected to improve flow measurements near zero, i.e. in the condition of a possibly blocked shunt. This approach omits cardiac gating and dynamic imaging, and thus measures the average flow during the scan.

Shunt Flow Phantom Construction and Phase Contrast MRI Calibration

We constructed a shunt flow phantom (Figure 1) with the following specifications: (i) a shunt catheter with 1.3 mm inner diameter and 2.5 mm outer diameter (Codman, Integra LifeSciences, New Jersey); (ii) complete immersion of the shunt catheter in a water chamber to reduce potential susceptibility artifact around the tubing, (iii) connection of the flow system to a calibrated continuous flow pump (Masterflex, Gelsenkirchen, Germany), (iv) validation of the pump flow rates using an analytical balance (Mettler, Toledo, Ohio). We calibrated our PC-MRI measurements in the shunt flow phantom at 20 physiologically-applicable flow settings ranging from 0–24 mL/hr.^25,26 We acquired a T2-weighted single shot MRI to identify the shunt catheter, and then placed the PC-MRI image plane orthogonal to the flow in the lumen.

Figure 1.

(a) Schematic of experimental setup for PC-MRI phantom data collection, including flow phantom, connection tubing, and continuous flow pump. (b) Expanded view of flow chamber, with homogeneous water-filled background and traversing shunt tubing (1.3mm ID). Imaging plane is prescribed orthogonal to the shunt catheter. (c) Example T2-weighted axial slice of flow phantom (yellow arrow indicates catheter tubing).

Acquisition of Phase Contrast MRI in Shunted Patients

This study was approved by the Institutional Review Board at Children’s Hospital Los Angeles (CHLA-20–00041). All families and patients provided written informed consent according to institutional guidelines and good clinical research practices. The study team enrolled 22 patients with shunted hydrocephalus and scanned them with PC-MRI (N=11 female, N=11 male; 11 months to 19 years old, median age 6 years). Patients were enrolled sequentially; no patient or data was excluded from the study. All patients were clinically stable and undergoing planned routine T2-weighted limited MRI for screening of ventricular size. No patient required additional anesthesia or sedation to accomplish the investigational protocol. The shunt systems in these patients were 18 non-programmable VP shunts, and 4 programmable MR-resistant shunts, which are reflective of typical hardware present in our patient population. Two inpatients had externalized distal catheters (valve remaining internal), which is a unique shunt configuration allowing for direct accompanying measurement of CSF output. The distal shunt catheter inner diameter was 1.3 mm for all patients.

The PC-MRI imaging plane was prescribed orthogonal to the distal catheter lumen inferior to the shunt valve. In patients with programmable shunts, PC-MRI slices were planned at least 2 cm distal to the valve to avoid artifacts.

A subgroup of these patients had additional imaging collected in other locations to assess the effects of temporal-spatial variations, or the difference in proximal versus distal measurements.

Testing spatial variation and proximal versus distal location

Four patients received additional images to test if flow measurement was dependent on either spatial variation in placement of the imaging plane or on selection of a proximal versus distal catheter location for flow imaging. For these 4 patients, we measured CSF flow at 5 proximal and 4 distal locations.

Testing temporal variation

Two patients underwent imaging to evaluate the effect of temporal variation alone on flow measurement at a single spatial location. In one experiment, we measured CSF flow at 1 proximal and 1 distal location and repeated each measurement 5 times using the standard PC-MRI sequence. In the other experiment, we modified the PC-MRI sequence to achieve higher sampling rates but poorer quality measurements. Our modified PC-MRI sequence used identical parameters, except for the addition of compressed sense (CS) factor=3 and a decrease in the NSA to 1. Using this modified sequence, we acquired 16 individual images (scan duration=11 s) to calculate flow at higher temporal resolution.

Validation of Phase Contrast MRI in Shunted Patients with Externalization of Distal Catheter

We performed in-vivo validation of our PC-MRI measurement of flow to a direct timed measurement of CSF output in two patients with distal shunt catheter externalization at the clavicle. Study personnel prescribed the PC-MRI plane in a long segment of internal subcutaneous catheter in the suboccipital region, timed the duration of the CSF collection epoch, and measured the volume of CSF collected during that time, converting volume output into CSF output rate. CSF collection and PC-MRI scanning were performed simultaneously.

Phase Contrast MRI Image Analysis

One investigator (JH) analyzed the PC-MRI images. Processing workflow included preprocessing of images to (i) remove any aliasing due to CSF flow velocities that exceeded the V_enc setting (ImageJ, needed in N=1 patient) then placing an ROI within the catheter lumen and measuring the velocity in the lumen using Horos (Purview, MD)). The velocity measurement process comprised: (ii) locating the shunt on the magnitude and complex difference images, (iii) placing a circular region of interest on the lumen, (iv) recording the raw mean velocity from the phase image, (v) correcting the raw mean velocity using the calibration data, (vi) computing the flow by multiplying the mean velocity by lumenal area and converting units to mL/hr. Total analysis time per image was approximately 3 minutes.

Results:

Use of PC-MRI to Measure Flow in a Phantom

Figure 2a shows PC-MRI shunt phantom velocity measurements within a circular ROI (1.3 mm diameter), plotted versus actual (weight-measured) bulk flow rate and corresponding calculated actual mean velocity. At the zero-flow condition in the phantom, PC-MRI correctly measured velocity (0.0 ± 0.0 cm/s). At variable (nonzero) flow rates, PC-MRI data were fit with the following function:

$$y=0.56(1-e^{{-11x}^{2.2}})$$ (1)

Figure 2.

Quantitative calibration of PC-MRI using flow phantom. (a) MRI-measured phase-contrast velocity (PC Velocity, y-axis) plotted against pump output fluid flow rate / velocity (dual x-axis) determined by analytical balance. (b) Representative PC-MRI axial sections through the phantom catheter lumen at flow rates of 7, 13, 17, and 19 mL/hr (scale bar 1.3 mm indicates catheter inner luminal diameter). (c) PC MRI velocity profiles across the diameter of the shunt catheter (red line, b) using consecutive 0.1×0.1 mm square ROIs positioned across the catheter lumen (average of 5 measurements per ROI, 4 flow rates). Dashed grey lines correspond to walls of the catheter lumen.

Wherein, x is the scale-measured pump output and y is the PC-MRI measured velocity.

Figure 2b displays representative PC-MRI images from the phantom at 4 unique pump velocities and demonstrates a parabolic flow profile (Figure 2c). Table 2 provides a clinical reference for straightforward conversion between PC-MRI velocity and real expected shunt flow at different catheter luminal diameters, incorporating the phantom-derived velocity correction.

Table 2:

Clinical Reference Table for Conversion of PC-MRI Velocity to CSF Shunt Flow Rate

PC MRI Velocity (cm/s)	Pump Velocity (cm/s)	1.0 mm ID Shunt Flow (mL/hr)	1.3 mm ID Shunt Flow (mL/hr)	1.5 mm ID Shunt Flow (mL/hr)
0.0	0	0	0	0
0.0002	0.01	0.3	0.5	0.6
0.001	0.02	0.6	1.0	1.3
0.003	0.03	0.8	1.4	1.9
0.005	0.04	1.1	1.9	2.5
0.008	0.05	1.4	2.4	3.2
0.012	0.06	1.7	2.9	3.8
0.017	0.07	2.0	3.3	4.5
0.022	0.08	2.3	3.8	5.1
0.03	0.09	2.5	4.3	5.7
0.04	0.10	2.8	4.8	6.4
0.05	0.12	3.4	5.7	7.6
0.07	0.14	4.0	6.7	8.9
0.10	0.16	4.5	7.6	10.2
0.12	0.18	5.1	8.6	11.5
0.15	0.20	5.7	9.6	12.7
0.18	0.22	6.2	10.5	14.0
0.20	0.24	6.8	11.5	15.3
0.23	0.26	7.4	12.4	16.5
0.26	0.28	7.9	13.4	17.8
0.29	0.30	8.5	14.3	19.1
0.32	0.32	9.0	15.3	20.4
0.35	0.34	9.6	16.2	21.6
0.38	0.36	10.2	17.2	22.9
0.40	0.38	10.7	18.2	24.2
0.42	0.40	11.3	19.1	25.4
0.44	0.42	11.9	20.1	26.7
0.46	0.44	12.4	21.0	28.0
0.48	0.46	13.0	22.0	29.3
0.49	0.48	13.6	22.9	30.5
0.50	0.50	14.1	23.9	31.8

Use of PC-MRI to Measure CSF Flow in Patients

Figure 3a shows a schematic of the validation experiment for shunted patients who previously underwent procedures for clavicular externalization of the distal shunt catheter. Their shunt system was equivalent to the native internalized shunt in tissue imaging characteristics at the sites where measurements were performed. The timed collection versus PC-MRI CSF flow rates for the first validated patient were 8 mL/hr versus 8±1 mL/hr (Figure 3a). In a second patient with externalized shunt, we interleaved both proximal and distal measurements along with timed CSF collection as described in the accompanying schematic (Figure 3b). Distal catheter measurements versus known CSF output were 54±2 versus 54 mL/hr (RMSE=1.4). Proximal PC-MRI measurements were not valid (5±4 versus 54 mL/hr; RMSE=48.9, p<0.0001).

Figure 3.

(a) Schematic of patient with externalized distal shunt catheter and location of imaging plane (dashed red line). Calibration-corrected PC-MRI flow measurement is plotted alongside direct measured volume of CSF output in the external collection burette. (b) An additional patient with externalized distal catheter received multiple proximal and distal catheter PC-MRI measurements (dashed red lines), plotted against known output over time in the collection burette. (c) PC-MRI measurements of flow rate obtained at 1 cm increments along the proximal and distal catheters (representative images inset) in 4 separate patients. Data from each patient are normalized to the average distal flow rate as a guide to the eye (dashed line); separate patients are denoted by unique markers. (d) Comparison of proximal and distal PC-MRI flow measurements normalized to average distal flow, holding image slice location constant. (e) PC-MRI velocity measurements at a single proximal catheter location (single patient) using a modified higher temporal resolution (11 s) scan.

Four other patients with functioning internalized shunts received 5 separate PC-MRI scans in the proximal catheter and 4 PC-MRI scans in the distal catheter, with 2-D image slice planes spaced by sequential 1 cm intervals. Figure 3c displays the proximal and distal flow. Distal catheter measurements in individual patients were highly reproducible with one another; slice location did not significantly influence measurement value (Spearman’s rank correlation p=0.74). Proximal catheter measurements had poor reproducibility and poor accuracy (average magnitude of bias 6 mL/hr). Measurements closer to the proximal or distal catheter tip did not systematically result in higher or lower flow compared to those obtained in closer proximity to the valve. We observed no pattern to bias in proximal catheter measurement.

We compared 5 consecutive PC-MRI measurements at the same proximal catheter location and 5 measurements at the same distal catheter location (Figure 3d). The variation in the proximal measurements was high, and the distal catheter measurements displayed low variability. To further investigate the temporal nature of proximal measurement fluctuations, one patient was scanned with accelerated PC-MRI at higher temporal resolution (11 s per measurement) (Figure 3e). These data showed that increased temporal resolution and constant scan location did not resolve the variability in proximal measurement.

To evaluate the clinical feasibility and reliability of distal flow measurement in clinical patients, PC-MRI from all patients who underwent distal flow measurement in this study were collected in aggregate and are presented in Figure 4. A summary of patient clinical and demographic variables is presented in Table 3. No patient was suspected to have shunt obstruction prior to inclusion in this study, and no patient required intervention due to a newly diagnosed obstruction during the 3 months following data acquisition.

Figure 4.

Distal catheter PC-MRI results from 21 pediatric patients. Each patient is included as a separate row, displaying patient-specific representative T2 axial image (distal catheter denoted with red arrow), T2 sagittal image (imaging plane as dashed line), phase-contrast magnitude image, phase-contrast complex difference image, phase image, measured PC-MRI CSF velocity plotted in context of the phantom calibration curve, and final calculated CSF shunt flow. Note, patient 21 was known to have CSF output far excessing the typical physiological range and beyond the calibration range. Data for patient 21 is plotted without calibration correction and is marked with (*).

Table 3:

Demographic and Clinical Characteristics for Each Patient Scanned, Including Measured CSF Flow Rate

Patient ID	Age (years)	Sex	Etiology of Hydrocephalus	Valve	Programmable	Previous Revision(s)	Distal Catheter Site	CSF Flow Rate (mL/hr)
1	1	M	IVH	Medtronic Delta	—	—	Peritoneum	12 ± 1
2	1	F	Schizencephaly	Medtronic Delta	—	—	Peritoneum	15 ± 1
3	1	F	Unknown	Codman Certas	+	—	Peritoneum	12 ± 1
4	1	M	ATRT	Codman Certas	+	—	Peritoneum	8 ± 1
5	1	F	Toxoplasmosis	Medtronic Strata	+	—	Peritoneum	14 ± 1
6	2	F	Arachnoid Cyst	Medtronic Delta	—	1	Peritoneum	12 ± 1
7	2	F	IVH	Medtronic Delta	—	3	Peritoneum	10 ± 1
8	3	M	Choroid Plexus Carcinoma	Medtronic Delta	—	—	Peritoneum	7 ± 1
9	5	F	Dandy-Walker Syndrome	Medtronic Delta	—	—	Peritoneum	10 ± 1
10	6	M	Citrobacter Meningitis	Medtronic Delta	—	3	Peritoneum	10 ± 1
11	6	F	Unknown	Chhabra	—	1	Externalized	8 ± 1
12	6	F	Myelomeningocele	Medtronic Delta	—	5	Peritoneum	8 ± 1
13	6	M	Meningocele	Medtronic Flow Control	—	—	Peritoneum	19 ± 1
14	7	M	Myelomeningocele	Medtronic Delta	—	2	Peritoneum	10 ± 1
15	8	F	IVH	Medtronic Delta	—	2	Atrium	9 ± 1
16	14	M	Arachnoid Cyst	Medtronic Delta	—	1	Peritoneum	11 ± 1
17	15	M	Ependymoma	Medtronic Delta	—	—	Peritoneum	12 ± 1
18	16	F	Tuberculous Meningitis	Medtronic Delta	—	4	Peritoneum	19 ± 1
19	18	M	Pilocytic Astrocytoma	Medtronic Delta	—	—	Peritoneum	6 ± 1
20	19	M	IVH	Medtronic Delta	—	1	Peritoneum	11 ± 1
21	11 months	M	Pfeiffer Syndrome	Codman Certas	+	1	Externalized	54 ± 2

Discussion:

Our results indicate that translation of PC-MRI to routine use is now an appropriate next step: this provides a straightforward option for assessing quantitative shunt flow as a routine, highly relevant variable added to the multiple radiographic and symptomatic parameters used in clinical assessment of shunted patients. Temporal variation of CSF flow within the proximal catheter renders these measurements unreliable, thus distal flow measurements should be used to optimize accuracy of mean flow. Spatial variations in flow do not appear to influence the quality of the distal catheter measurements, although we suggest prescribing the imaging plane at least 2 cm away from programmable shunt valves to avoid image artifacts. Phantom calibrations indicate subtle nonlinear effects at the highest and lowest CSF velocities. These are likely relevant in patients with very low flow, and the translation of these findings to a dedicated investigation of shunt failure will be the topic of a subsequent study. Validation of PC-MRI for CSF flow in 2 patients with externalized catheters demonstrate excellent agreement between PC-MRI flow measurement and volumetric CSF output in the collection burette. PC-MRI yielded statistically accurate measurements in both patients, including in one with 4 sequential distal measurements. The distal technique was applied to 21 patients in total, with successful measurements obtained in all cases, and with measurements that fell within the known normal range of physiological CSF production. We provided our PC-MRI parameters and our velocity-to-flow conversion table for use by clinicians (Table 2). PC-MRI is ready for translation and can offer a major clinical advancement for patients with shunted hydrocephalus. The method can also serve as a fast, non-invasive, repeatable research tool for improved academic understanding of CSF physiology in shunted hydrocephalus, and in the long-term individualized assessment of flow changes over time − a contribution to patient-directed care.

Our approach has key differences from the 4 extant reports of CSF flow imaging. Two reports measured CSF flow in the proximal catheter; one displayed the pulsatile behavior of CSF flow in the proximal catheter²⁷ and the other attempted to use high temporal resolution imaging to resolve this effect.²⁸ The 3 prior studies of CSF flow in the distal portion of the shunt or using a mix of proximal/distal measurements did not compare results between proximal and distal locations, and did not identify proximal variability as the source of inaccuracy for the technique (Table 1).^29–31 In contrast, our study evaluated both proximal and distal measurement locations, and assessed the influence of both temporal and spatial variation on flow measurements. The choice to obtain measurements in both the proximal and distal catheters was motivated by our hypothesis that intracranial pulsatility may disproportionately affect CSF flow in the proximal catheter. Based on these data, the high CSF pulsatility of the proximal side is dampened by the polymer chambers of the valve, and thus becomes continuous flow on the distal side.

Prior studies used in vitro calibration and linear fits of PC-MRI data in reference to known phantom flow.^28–30 However, owing to the large phantom dataset collected in this work with multiple repeat measurements, our data highlight a mild sigmoidal relationship between velocity measured by PC-MRI versus the pump within the phantom-tested range. Definition of this association is critical because the sigmoidal and linear models are most dissimilar at very low flows, which are the most important for the eventual evaluation of shunt malfunction.

Certain prior studies report making a plug-flow assumption during in vitro calibration.²⁸ Our data demonstrate a parabolic flow profile at all measurement conditions (Figure 2c). Thus, the empirical data provided here are consistent with laminar flow predicted by the low Reynold’s number (median R_e=5 based on CSF viscosity, fluid density, flow velocity and catheter diameter).^32,33

No prior studies demonstrate an in vivo validation of the PC-MRI method for assessing shunt flow. We had the opportunity to study 2 patients whose shunts were externalized at the clavicle. The patients had very different CSF flow rates (one patient was a known high producer of CSF, with daily output >1 L), which enabled us to validate a wide range of flows. Clavicular externalization is relatively uncommon at our institution because most patients who could require externalization (e.g., due to shunt infection) would be likely to receive a complete hardware removal, with placement of temporary valveless external ventricular drain (a system that would not be expected to dampen pulsatility due to the lack of valve). The MR tissue characteristics surrounding the valve and distal catheter in our patient model with valve in place were identical compared to a fully internalized shunt.

Alternative approaches to obtain shunt flow or correlates of flow.

The drive to create or implement any clinically available shunt flow assessment (even qualitative, semi-quantitative, or invasive) indicate that shunt flow measurement can be a vital aspect of patient management in the correct clinical scenario. Radiopharmaceutical shunt flow studies use burdensome radioisotopes, require invasive percutaneous puncture of the shunt system with possibility of contamination, provide only correlates of flow rather than quantitative flow, and can be prone to erroneous interpretation.¹⁵ Temperature-based flow techniques use a cold or heat source and thermosensors to obtain an indirect measure of flow using Fick’s principle. This approach might be appropriate in settings with limited access to imaging, or as a method to corroborate a preexisting low clinical suspicion of shunt malfunction.^34–36 In contrast, our approach integrates a quantitative, non-invasive method into the standard-of-care periodic rapid imaging of ventricular size, which is already performed routinely at academic pediatric medical centers as a component of outpatient follow-up or inpatient management. Thus, the ease of use, simplicity, and repeatability of PC-MRI provide a solution to an unmet need for the neurosurgical and medical communities at large.

Limitations.

One limitation of our study was the design of our flow phantom. Our phantom calibration measured at 0, 1, and 4–24 ml/hr, which were selected to overlap with the expected physiologic range of CSF production. Our flow phantom only evaluated a 1.3 mm diameter lumen. Future flow phantom studies should use new pump hardware to explore the range from 1–4, and above 24 ml/hr, and include multiple lumen diameters Our in vivo validation of CSF measured 8 and 54 ml/hr. Future validation studies in patients with externalized shunts producing CSF flow less than 8 or greater than 54 ml/hr can extend the range of validation. Different encoding velocities may be appropriate in those ranges. Until new calibration and validation are completed, we recommend caution for the use of quantitative PC-MRI shunt measurements within the low flow or very high flow range.

A secondary limitation of our study is the limited number (N=6) of valve types that we evaluated. While we speculate that other valve types will have similar effects on MR imaging and fluid dynamics, we have no data to support the speculation. Programmable valves that do not maintain a constant setting in the MR environment (e.g. Strata II^® valve; Codman^® Hakim^® Programmable valve) pose a unique issue with regard to interpreting shunt flow, as the valve setting during scanning would be unknown. While a diminishing number of individuals in our patient population have this style of valve, the proportion of patients with non-MR protected valves would be expected to vary regionally with local/historical clinical practice.

A tertiary limitation of our study is the use of a single site, scanner model, vendor, and field strength. Thus, the parameters in calibration equation 1 should be independently measured on other systems.

Conclusions:

PC-MRI is a practical technique that is ready for noninvasive clinical measurement of CSF flow within shunted patients. Future work using PC-MRI should assess CSF flow in patients without suspected shunt malfunction to determine if shunt failure can be predicted. Other studies should investigate CSF production in patients with and without hydrocephalus.