Measurement of CSF flow and brain motion in Chiari malformation type I subjects undergoing posterior fossa decompression surgery

Abstract

OBJECTIVE

Radiologically, Chiari malformation type I (CM-I) is characterized by cerebellar tonsil herniation of at least 5 mm through the foramen magnum. In symptomatic cases, posterior fossa decompression (PFD) surgery is often performed and improves symptoms in approximately 75% of patients. However, the surgery involves risks, and identifying which candidates will benefit from surgery is important. It has previously been shown that the amount of tonsillar descent does not correlate with symptom severity or surgical outcomes. The authors hypothesized that using advanced neuroimaging methods to directly measure CSF flow and brain motion will give insights regarding which patients have the greatest likelihood of cerebral dynamic improvements from surgery.

METHODS

Here, the authors evaluated 108 CM-I patients (age 19–70 years), 61 of whom underwent PFD surgery. The authors used phase-contrast MRI to measure CSF flow/stroke volume and cine displacement encoding with stimulated echoes (DENSE) imaging to measure brain motion, with a goal to predict postsurgical cerebral dynamic improvements from presurgical images.

RESULTS

The authors found that CSF stroke volume increased after PFD surgery by 28.9% (p = 0.014), brainstem motion decreased after surgery by 17.3% (p = 0.002), and cerebellum motion decreased 45.2% (p < 0.001). Notably, the amount of CSF flow increase after surgery had no relationship to tonsillar descent (R = 0.059, p = 0.767) but did relate to the amount of presurgical CSF flow (R = −0.518, p = 0.005). Likewise, improvements to brain motion were better predicted by the amount of presurgical motion (brainstem, R = −0.638, p < 0.001; cerebellum, R = −0.878, p < 0.001) than by tonsillar descent (brainstem, R = −0.312, p = 0.093; cerebellum, R = −0.620, p < 0.001).

CONCLUSIONS

Here, the authors found that presurgical measures of cerebral dynamics were more descriptive of improvements to CSF flow and brain motion after PFD surgery than the conventional measure of presurgical tonsillar descent. These expanded quantitative assessments to determine which patients may benefit from surgery could improve the overall quality of patient care.

The radiological definition of Chiari malformation type I (CM-I) is herniation of the cerebellar tonsils through the foramen magnum of at least 5 mm.¹ This metric is not considered an absolute diagnosis as it is unreliable for describing the severity of symptoms.^2,3 The main symptom of CM-I is headaches induced by activities that transiently increase intracranial pressure such as coughing, sneezing, or straining, but other signs/symptoms include paresthesia, nausea, clonus, dysphagia, nystagmus, and sleep apnea.⁴ Signs/symptoms can vary widely, and some individuals have > 5 mm tonsillar descent present with no clinical symptoms.⁵ Therefore, the criteria for surgical intervention are ill defined, and a decision to intervene is formed from a combination of symptoms, imaging results, and the patient’s willingness or suitability to undergo surgery.^6,7 CM-I is typically diagnosed after onset of symptoms using traditional T1- and T2-weighted anatomical MRI. MRI is also used to assess subjects for the presence of syringomyelia, a secondary condition that can form with CM-I, that involves the formation of one or more spinal cord cysts. These cysts can enlarge over time and cause progressive myelopathy, beginning with sensory dysfunction due to the crossing sensory fibers within the cord and only later motor dysfunction as the syrinx enlarges.

CSF is housed in the subarachnoid space and ventricular system, and its biochemical makeup, volume, and pressure play vital roles in cerebral homeostasis including maintenance of normal cerebral perfusion.⁸ In healthy individuals, CSF passively flows through the ventricles into the subarachnoid space of the brain and spinal cord. This exchange is driven largely by the cardiac and respiratory synchronized pulsatile motion of the brain. As the arterial pulse leads to systolic expansion of the brain, it causes the propagation of inferiorly directed motion of the CSF.⁹ However, in people with CM-I, normal CSF flow is obstructed as the herniated tonsils acts as a stopple within the foramen magnum. Changes in CSF flow can cause pressure variations in and around the intracranial space.

The surgical intervention recommended for symptomatic individuals with radiological evidence of CM-I is posterior fossa decompression (PFD), which involves using a suboccipital craniectomy, C1 laminectomy, and duroplasty procedure to enlarge the foramen magnum and cervicomedullary junction to establish normal CSF flow.¹⁰ Surgery improves symptoms in 70%–80% of patients and is often accompanied by complete or partial resolution of the syrinx.^11,12 However, the surgery is not without risks, with a surgical morbidity/mortality rate as high as 1.3%.¹³ Therefore, identifying metrics that can better characterize a patient’s potential to improve with surgical intervention would aid decision-making.

Recently, interest has grown in directly and quantitatively measuring the secondary neurophysiological effects resulting from tonsillar descent, as these altered cerebral dynamics presumably cause the clinical symptoms.¹³ Although CSF flow is commonly assessed in the evaluation of CM-I patients, it is almost always done using either qualitative assessments or phase-contrast magnetic resonance (PCMR) imaging acquired in the sagittal plane, both of which are inadequate to characterize total CSF flow. Sagittal PCMR in CM-I is often used to measure single-voxel peak CSF velocity in areas where the CSF channel has narrowed; this technique can suffer from noise, aliasing, and the presence of blood vessels with high velocity that can corrupt single-pixel measurements. We expect that quantitative measures of CSF, stroke volume, or the total amount of CSF that flows into or out of the area of interest in a single heartbeat, can better describe the severity of the blockage from the tonsillar descent.

A recently discovered secondary effect of narrowing of the CSF space in the foramen magnum is altered cardiac-induced brain motion. The Monro-Kellie doctrine states that to maintain normal intracranial pressure, the cranial cavity must maintain a constant volume; therefore, when the arterial pulse leads to systolic increase in cerebral blood volume and an expansion of brain volume, CSF is passively pushed out of the skull.¹⁴ However, in CM-I, the passive flow of CSF is obstructed by the cerebellar tonsils, and to compensate, the brain tissue must displace more to maintain cerebral homeostasis.^15,16 Brain motion has only recently been measured in vivo, and therefore work in quantitatively measuring brain motion in CM-I is limited. Brain motion can be measured using cine displacement encoding with stimulated echoes (DENSE). DENSE can quantify submillimeter displacements associated with brain tissue motion. Early work using DENSE has demonstrated strong evidence of elevated brain pulsatile motion in CM-I, but these findings are not comprehensive.^14–16

In this study, we sought to use advanced neuroimaging methods to quantify how cardiac-induced pulsatile brain motion and CSF flow change in people with CM-I after PFD surgery. We expected that CSF flow would be reduced before surgery and that brain motion would be elevated before surgery, and we hypothesized that measuring these effects would give better insights regarding which patients will have the greatest likelihood of cerebral dynamic improvements from surgery than the standard measure of tonsillar descent.

Methods

Participants

A total of 108 participants (mean [range] age 37.1 [19–70] years; 19 males, 88 females, and 1 sex not disclosed) with a clinically suspected diagnosis of CM-I were enrolled in this study. Participants included individuals with symptoms severe enough that they sought diagnosis and treatment, and all had undergone a previous imaging study that suggested Chiari malformation. All subjects had symptoms present for at least 1 year prior to enrolling in this study. The study was approved by the IRB, and all subjects provided informed written consent.

Prior to the first MRI, 12 patients were removed due to MRI contraindications, including claustrophobia, pregnancy, and scheduling challenges. Two patients were excluded after their first research MRI scan for not meeting the radiological standard for diagnosis of CM-I (cerebral tonsils > 5 mm below the basion-opisthion line). After exclusions, the study sample size included 96 people (mean [range] age 37.4 [19–70] years; 16 males, 79 females, and 1 sex not disclosed).

Of these 96 individuals, 61 underwent PFD surgery at a single university-based hospital by a single board-certified neurosurgeon with 39 years of post-fellowship experience. Forty-eight of these surgical patients participated in a follow-up research MRI scan using the same protocol as the presurgical time point. The follow-up scan was scheduled for no sooner than 3 months after surgery, with all subjects completing the follow-up scan within 6.5 months of surgery (mean duration 119 ± 34 days after surgery).

Imaging included an axial 2D PCMR scan to examine CSF flow and a high-resolution DENSE scan to examine tissue displacement, as seen in Fig. 1. Not every participant had useable PCMR and DENSE imaging at both time points, due to factors including time constraints, challenges with sequence execution, poor quality/corrupted images, or statistical outliers. Before surgery, 70 of the 96 patients had an analyzable DENSE scan and 67 subjects had an analyzable PCMR scan, with 45 subjects having both; after surgery, 31 subjects had an analyzable DENSE scan and 28 subjects had an analyzable flow scan, with 16 subjects having both (Fig. 2). DENSE images from a subset of 23 of these 96 participants have been previous published.¹⁴

FIG. 1.

Panels on the left show an individual with Chiari malformation type 1, showing a descent of the cerebellar tonsils of at least 5 mm through the foramen magnum and the presence of a syrinx at C3–4 before surgery. Axial PCMR images show CSF flow around the brainstem during peak diastole and DENSE images show brain motion at peak systole. After PFD surgery, syrinx resolved, CSF flow increased, and brain motion decreased. Panels on the right show methods for calculating CSF stroke volume and cerebellum motion from imaging data.

FIG. 2.

Patient enrollment numbers before and after intervention, with subsets of patients who had PCMR and DENSE scans. Numbers associated with each image type are shown after the exclusion of patients who did not come back for follow-up evaluations, statistical outliers, and exclusions due to poor image quality.

On the day of follow-up imaging, each participant completed the Chicago Chiari Outcome Scale (CCOS), a 16-point assessment of pain symptoms, nonpain symptoms, functionality, and complications.

Surgical Procedure

Surgery was conducted under general anesthesia and included standard suboccipital craniectomy, C1 laminectomy, and expansile duroplasty. The cerebellar tonsils were shrunk using bipolar cautery at low power under constant saline irrigation. Dural closure was completed utilizing Durepair collagen matrix (Medtronic) sewn into place in a watertight fashion using running 4–0 Nurolon sutures. Adherus dural sealant (Stryker) was used to reinforce the dural closure.

Imaging Protocol

Participants were scanned on a Siemens 3.0-T Prisma Fit MRI with a 20-channel head/neck coil. Sagittal turbo-spin-echo T2-weighted high-resolution anatomical scans were collected for all participants and at a resolution of 0.6 × 0.6 × 3.0 mm³ and TE/TR 106/4790 msec.

Retrospectively peripherally pulse unit (PPU)–gated, 2D axial PCMR images were acquired at image planes that were placed perpendicular to the spinal cord in the C-spine at the inferior aspect of the C2 (n = 49) or C6 (n = 18) vertebra. Imaging parameters included field of view 220 × 220, resolution 1.2 × 1.2 × 5 mm, TE/TR 6/21 msec, VENC (velocity encoding value) 15 cm/sec, and flip angle 20°. For each cardiac cycle, 25 temporal images with 2 segments and 2 averages were collected. Image acquisition time varied depending on heart rate, but for a heart rate of 60 bpm the scan time was 1 minute 48 seconds.

DENSE scans were acquired at the midline in the sagittal orientation using a 2D prospectively PPU-gated spiral cine technique.^17,18 In DENSE, a series of images are acquired where the signal in the phase image is directly proportional to the motion of the tissues within the imaging plane. Because the motion is proportional to the phase shift, displacements as small as 20 μm can be reliably measured in each pixel.¹⁹ We acquired images in the sagittal plane and across the cardiac cycle. Two sets of DENSE data were acquired, one with anterior-posterior displacement encoding and one with cranial-caudal displacement encoding. In total, 13–29 DENSE images were acquired over the cardiac cycle depending on the subject’s heart rate,^20,21 and the scan time was approximately 3 minutes to collect both directions for a heart rate of 60 bpm. Other DENSE parameters included the following: flip angle 15°, temporal resolution 30–40 msec, encoding frequency 0.6 cycles/mm, spiral interleaves per heartbeat 2, total spiral interleaves per image 192, field of view 256 × 256, pixel size 0.9 × 0.9 mm, and slice thickness 8 mm.

Image Analysis

Tonsillar descent was measured on sagittal T2-weighted imaging from the bottom tip of the cerebellum to the basion-opisthion line; measurements were made twice and averaged to minimize error. The intraclass correlation coefficient between measurements was 0.914, indicating excellent measurement repeatability.

Cine PCMR images were used to calculate CSF flow as a function of time using Segment (research version 12067, Medviso). PCMR images capture time-resolved velocity-encoded data representing the directional flow of CSF over the cardiac cycle. Regions of interest (ROIs) were manually drawn around the subdural space around the CSF fluid surrounding the brain stem and spinal cord, taking care to avoid vascular structures. The pixel-by-pixel velocity values (cm/sec) in the ROI were integrated to determine the flow (ml/sec) at each time point across the cardiac cycle. Because CSF pulses back and forth over the cardiac cycle, the net flow is assumed to be 0, as there is no measurable production of CSF volume over the scan duration. CSF flow curves were adjusted to reflect this zero net flow value. The CSF stroke volume (ml) is determined by calculating half of the absolute value of the positive (cranial) and negative (caudal) flow over the cardiac cycle. The resulting stroke volume quantifies the amount of CSF displaced either cranially or caudally during each heartbeat.

The phase displacement data from DENSE imaging are processed to generate quantitative maps of tissue displacement across the cardiac cycle, revealing patterns of cyclic motion driven by cardiac pulsations. DENSE images were analyzed in 1) the cerebellum and 2) the brainstem superior to the foramen magnum. Region displacement was calculated from the phase information using an internally developed program (DENSEpro) on MATLAB (R2020a MathWorks).²² DENSEpro was used for manual segmentation of the ROI and calculation of pixel-by-pixel displacements. Only the first two-thirds of the timepoints were used, as T1-weighted decay caused a progressive reduction in the signal-to-noise ratio. The maximum tissue displacement over the cardiac cycle was calculated by subtracting the highest and lowest displacement values for each pixel, then these maximum values were averaged over the ROIs to get values for displacement. For comparison, we used data that were previously acquired and published by our group about the brainstem and cerebellum motion in 25 control subjects.²³

Statistical Analysis

Statistical analysis was performed using JMP Pro 16.0.0 (SAS). Outliers were eliminated using multivariate robust outlier analysis. In this test, we first determined the 10% tail quantile (indicating the middle 80% of data as most robust); then, we excluded any data points that existed more than 3 times away from this 80% interquartile range. Any subject who had outliers in their presurgical data was removed from analysis; subjects who had outliers in their postsurgical data were included in the presurgical analysis but excluded from pairwise analysis. We excluded presurgical CSF flow data from 2 patients. We conducted the Shapiro-Wilk goodness-of-fit test on all metrics to evaluate normality of distribution. For normally distributed data, the 2-tailed nonpaired t-test was performed between presurgical/no-intervention data and postsurgical data for both CSF flow and brain motion in the brainstem and cerebellum. The paired t-test was additionally conducted for only subjects who had undergone surgical intervention; this was done for CSF flow and brain motion in the brainstem and cerebellum. For nonnormally distributed data, we used the nonparametric Wilcoxon signed-rank test at the group level and pairwise basis. For evaluations where 1 ROI or time point was not normally distributed, we applied the same nonparametric testing to all ROIs. For tests where multiple ROIs were assessed, Bonferroni correction was applied. Correlations between tonsillar descent before surgery, CSF flow before surgery, brain motion before surgery, and changes in CSF flow and motion after surgery were conducted using Pearson’s correlation analysis. R values were reported to assess the strength of the correlations, with significance at p < 0.05.

Results

CSF Flow

Figure 3 shows CSF stroke volume in individuals with CM-I. Across all subjects, including both those who did and did not undergo surgery, stroke volume was 0.49 ± 0.24 ml per cardiac cycle. In subjects who underwent surgical intervention, stroke volume was significantly greater (28.9%) after surgery (0.63 ± 0.28 ml) (p = 0.014). Taken on a pre/postsurgical pairwise basis of only those subjects who underwent intervention, subjects had an average 22.6% increase of CSF stroke volume (p = 0.057). This effect of increased stroke volume with surgery was consistent, with only 1 of 20 subjects showing a decrease in flow of greater than 1 standard deviation from the mean. There was no significant difference in presurgical stroke volume between subjects who did and did not undergo surgery (p = 0.421). We did not observe a significant change in peak maximum velocity after surgery, with paired data showing an average presurgical velocity of 5.81 cm/sec and average postsurgical peak velocity of 5.51 cm/sec (p = 0.178). We evaluated the differences in CSF dynamics between people who did/did not have a syrinx initially and found no significant groups differences in peak velocity (p = 0.243), stroke volume (p = 0.425), or CSF flow (p = 0.143).

FIG. 3.

CSF stroke volume across 1 cardiac cycle pre- and post-PFD surgery in people with CM-I. Box plots include all CM-I subjects regardless of whether they underwent intervention. Median (middle line), interquartile range (box), and minimum and maximum (whiskers) are shown. Line graphs show before and after surgery displacements in just people who underwent surgery. CSF flow is greater after surgery. *p < 0.05.

We found that the amount of tonsillar herniation below the foramen magnum before surgery was not related to the improvements in CSF flow after surgery (R = 0.059, p = 0.767) (Fig. 4). However, the amount of presurgical CSF flow was related to the amount of improvement in CSF flow after surgery (R = −0.518, p = 0.005); the less flow before surgical intervention, the more CSF flow increase after surgery.

FIG. 4.

Relationship between change in CSF flow with PFD surgery and the amount of tonsillar descent before surgery (left) and the amount of CSF flow before surgery (right), respectively. Surgical improvements in CSF flow were better predicted by the amount of presurgical CSF flow than by the amount of tonsillar descent. *p < 0.05.

Figure 5 shows representative examples of 2 CM-I patients. Patients had similar tonsillar descent (17.2 mm in patient A vs 17.5 mm in patient B). However, patient A had restricted CSF flow before surgery (0.31 ml) while patient B did not (0.72 ml). With surgery, patient A saw an increase in CSF flow (to 0.98 ml), whereas patient B did not show an increase in CSF flow. CSF flow data were normally distributed on both pre- (p = 0.140) and postsurgical (p = 0.053) imaging.

FIG. 5.

Two representative Chiari patients showing that presurgical CSF stroke volume better predicts improvements to CSF flow than the standard measure of tonsillar descent. Each patient has a similar amount of tonsillar descent, but only patient A had restricted flow before surgery and therefore was the patient who showed meaningful improvements to flow with surgery.

Brain Motion

Average brain displacement in individuals with CM-I decreased after PFD in both the brainstem and the cerebellum (Fig. 6). Across all CM-I subjects, brainstem motion was 187 ± 64 μm and motion was 162 ± 53 μm (p = 0.097) after decompression surgery. This change in motion was driven by changes in the cranial-caudal motion, which went from 180 ± 61 μm in patients before decompression surgery to 154 ± 45 μm with decompression surgery (p = 0.038), compared to 86 ± 36 μm and 78 ± 37 μm (p = 0.356) in the anterior-posterior direction, respectively. On a pairwise basis, subjects showed a 17.3% decrease in brainstem motion before and after decompression surgery (p = 0.002), with a maximum change in motion of any subject being 55.9%.

FIG. 6.

Average brain motion in the brainstem (A) and cerebellum (B) across the cardiac cycle before and after PFD surgery in people with CM-I. Box plots include all CM-I subjects regardless of whether they underwent intervention. Median (middle line), interquartile range (box), and minimum and maximum (whiskers) are shown. Gray indicators show average brain motion in healthy controls determined by previously published work, with stars showing maximum brain motion in those healthy controls. Line graphs show before and after surgery displacements in just people with decompression surgery. After Bonferroni correction for multiple comparisons, brain motion was found to decrease significantly after surgery in both the brainstem and cerebellum. *p < 0.05.

Cerebellum motion changed more than brainstem motion with surgery. Presurgical cerebellar motion was 142 ± 69 μm, compared to 91 ± 36 μm after surgery (p < 0.001). Pre- to post-motion differences again were larger in the cranial-caudal direction (65 μm) than the anterior-posterior direction (33 μm), though both directions were significant (p < 0.001). On a pairwise basis, subjects showed 45.2% average decrease in cerebellar motion before and after surgery (p < 0.001), with the largest change in any subject being 78.8%.

In both the brainstem and cerebellum, decreases to brain motion with surgery were better predicted by the amount of presurgical motion than by the amount of tonsillar descent (Fig. 7). In the brainstem, tonsillar descent was not significantly correlated with motion changes (R = −0.312, p = 0.093), while presurgical motion and change in motion were significantly related, with those who had larger presurgical motion also having greater changes in motion with surgery (R = −0.638, p < 0.001). In the cerebellum, change in motion was significantly associated with the amount of tonsillar descent before surgery (R = −0.620, p < 0.001), with more individuals who had larger descent having larger effects from surgery. However, the amount of presurgical motion in the cerebellum had a much higher correlation coefficient than presurgical tonsillar descent (R = −0.878, p < 0.001), and therefore presurgical motion better predicted surgical effects to brain motion.

FIG. 7.

The amount of change in brain motion in both the brainstem (A) and cerebellum (B) with surgery was better predicted by the amount of presurgical brain motion than by the presurgical amount of tonsillar descent. *p < 0.05.

Note that motion data were normally distributed in the brainstem both before surgery (p = 0.452) and after surgery (p = 0.248) but was skewed for cerebellum motion both before (p = 0.002) and after (p = 0.013) surgery.

Relationship Between CSF Flow and Brain Motion

We found that CSF stroke volume and motion in the brainstem were related in subjects with CM-I before surgery (R = 0.44, p = 0.002) but not in the cerebellum (R = 0.23, p = 0.129). After surgery, CSF stroke volume and motion were related in both the brainstem (R = 0.52, p = 0.048) and cerebellum (R = 0.56, p = 0.024). There was no significant relationship between change in CSF motion with surgery and either change in brainstem motion (R = 0.179, p = 0.505) or change in cerebellar motion (R = 0.187, p = 0.488) with surgery (Fig. 8).

FIG. 8.

Correlations between stroke volume and motion. Significant effects are determined at p < 0.05.

Cerebral Dynamics and Patient Outcomes

Finally, when assessing brain motion and CSF flow compared to change in symptoms as measured by the CCOS assessment, we found that the amount of motion in the cerebellum before intervention was positively related to CCOS score; this outcome trended toward but did not cross the threshold of significance (R = 0.308, p = 0.053) (Supplemental Figure S1). Improvements in symptoms had an inverse relationship with change in brain motion in the brainstem (R = −0.286, p = 0.133) and cerebellum (R = −0.348, p = 0.064), though these changes were not significant (Supplemental Figure S2).

Discussion

We investigated the effects of PFD surgery on CSF flow and brain motion in individuals with CM-I. A major objective of Chiari surgery is to improve CSF flow by alleviating the obstruction caused by the herniated cerebellar tonsils at the foramen magnum. We found that CSF stroke volume increased after surgery, and brain motion, which is known to be elevated in people with CM-1, decreased with surgery.

A notable finding was that for both CSF stroke volume and brain motion, the amount of tonsillar descent, which is a commonly used clinical metric, was not the best predictor of cerebral dynamics improvements from surgery. Instead, the best predictors of improved CSF flow and brain motion after surgery was the amount of presurgical CSF flow and presurgical brain motion, respectively. This is to say that an individual is only likely to improve in terms of CSF flow with surgery if their presurgical flow is restricted, regardless of amount of tonsillar descent. Likewise, an individual is only likely to show a return to more normal brain motion with surgery if their presurgical motion is considerably above normal. We conclude that tonsillar descent alone is not a sufficient indicator for predicting which patients will benefit most from improved neural dynamics. Tonsillar descent has been questioned as a reliable measure of CM-I severity in the past, with other studies showing that the amount of tonsillar descent is not related to severity of CM-I symptoms.² Because CM-I is not categorized as a life-threatening condition, the decision to undergo surgery can be challenging for individuals. Successful surgical procedures improve symptoms in more than 85% of patients, but as with any surgical procedure, complications can arise. Therefore, identifying presurgical metrics that can identify which patients have the most to gain from surgery can significantly improve an individual’s abilities to make informed decisions about their care.

We used 2D axial PCMR to measure CSF flow and found a significant increase in CSF stroke volume of 28.9% after surgery. Our results agree with those of a small past study by Panigrahi et al. who reported that PFD surgery resolved CSF blockages in the dorsal portion of the foramen magnum in all subjects other than those who had surgical complicaitons.²⁴ While CSF space changes have previously been observed in CM-I, nearly all observations have used nondynamic evaluations. Shaffer et al. describes the inadequacy of using static sagittal geometrical images to assess CM-I presence and severity and surgical results and highlights a major gap in the literature regarding quantitative measures of cerebral dynamics.²⁵ Past work has used sagittally oriented PCMR to measure maximum CSF velocity, but maximum velocity is not spatially uniform, with the diastole and anterior subarachnoid space showing particularly increased CSF flows.^26,27 This effect is compounded by the heightened importance of individual pixels, which can be subject to noise, aliasing, and the presence of blood vessels with high velocity that can corrupt single-pixel measurements. Furthermore, CSF velocity results are inconsistent in the literature, with some studies reporting increased maximum velocity that is attributed to a more forceful fluid movement through a restricted opening, while other studies report decreased maximum velocity that is attributed to a decrease in overall fluid movement.²⁷ Both options are plausible, but this lack of consistency makes maximum velocity a nonviable clinical metric. Here, we saw no significant change to maximum velocity with surgery, though we were not necessarily collecting images at the point of greatest narrowing. The physiological and imaging measurement limitations underscore the importance of considering total CSF stroke volume as a more reliably measure to be used clinically.

The Monro-Kellie doctrine describes how the intracranial compartment is a closed system composed of brain tissue, CSF, and blood. When blood flows into the brain from the cardiac cycle, the total brain volume expands. This motion can be significant, on the order of 100–200 μm at the brainstem in a healthy adult. In addition, the pulsatile blood flow expanding the brain volume is a primary driving force of active CSF flow in and out of the skull. This phenomenon is critical to ensure optimal cerebral perfusion and intracranial pressure.^28,29

Using high-resolution DENSE, we found that displacement of the brainstem and cerebellum decreased significantly after PFD surgery. Brain tissue displacement comes from brain volumetric change over the cardiac cycle due to blood volume changes in the microvasculature.³⁰ The motion reduction at both ROIs after surgery was more pronounced in the cranial-caudal direction, which corresponds to the direction of largest motion from cardiac pulsation. In past work, we observed that healthy volunteers had a brainstem motion of 117 μm,²³ compared to brainstem motion in CM-I of 187 μm. Likewise, past work shows the healthy cerebellum had an average motion of 67 μm.²³ In this study, we observed an average cerebellum motion of 142 μm in Chiari subjects. Our work agrees with past DENSE imaging studies, which consistently showed greater than normal brain motion in people with CM-I that was attributed to increased time-varying pressure across the cardiac cycle.^31,32 Past work has shown y-direction motion elevations that were 163% greater in CM-I patients than healthy controls preoperatively, with CM-I patients having a 48% reduction of y-direction motion from surgery; however, these past studies, as well as our work, show that even postsurgically CM-I brain motion remained higher than that of healthy controls.¹⁵ Other work also observed the spatial distribution of motion effects within the brain and found that the most increased motion occurs in the lower regions of the cerebellar tonsils and brainstem. After surgery, both regions demonstrated spatially consistent motion patterns across the entire ROI. Spatially heterogenous motion patterns may be attributable to the lower cerebellum being closer to the changes in the anatomical structure of the CSF spaces.¹⁴ Past work has not found differences in brain motion in people with CM-I with and without syringomyelia.¹⁵

Our study had limitations. All surgical procedures were done at a single center by a single surgeon. Not every subject had both pre- and postsurgical DENSE and CSF flow data. Every surgical experience, even within a single site, has noncontrolled variations, including complications such as CSF leaks. Additionally, although our study suggests a relationship between brain motion and CSF flow alterations, further research is needed to uncover the mechanisms driving this correlation. We know that physiological phenomena beyond cardiac motion affect CSF flow, and this includes respiration and potentially neural activity. Future studies should examine the relationship between brain motion and/or CSF flow and long-term outcomes, patient-specific anatomy, and postsurgical recovery trajectories. Finally, while these data are an important contribution to the field, they do not yet establish a clinical standard for deciding surgery.

Conclusions

Our study offers valuable insights into the complex neural dynamics that occur pre- and postoperatively in people with CM-I. We found that the level of CSF flow and brain motion before surgery were better predictors of the surgical changes to cerebral dynamics than tonsillar descent. We quantitively showed that brain motion significantly decreases after surgery and CSF flow increases after surgery. These dynamic measures may enlighten additional aspects of CM-I that were previously not understandable through conventional geometrical images or measures of maximum CSF velocity. Notably, this research contributes to our ability to determine which surgical candidates are the most likely to have cerebral dynamic benefit from surgical intervention.

Supplementary Material

Supplemental Information

Online-Only Content

Supplemental material is available with the online version of the article.

Supplemental Figs. S1 and S2. https://thejns.org/doi/suppl/10.3171/2024.11.JNS241509.

ABBREVIATIONS

CCOS

Chicago Chiari Outcome Scale

CM-I

Chiari malformation type I

DENSE

displacement encoding with stimulated echoes

PCMR

phase-contrast magnetic resonance

PFD

posterior fossa decompression

PPU

peripherally pulse unit

ROI

region of interest