The Relationship Between Imbalance Symptom and Cardiac Pulsation Induced Mechanical Strain in the Brainstem and Cerebellum for Chiari Malformation Type I

Abstract

Chiari malformation Type I (CMI) is known to have an altered biomechanical environment for the brainstem and cerebellum; however, it is unclear whether these altered biomechanics play a role in the development of CMI symptoms. We hypothesized that CMI subjects have a higher cardiac-induced strain in specific neurological tracts pertaining to balance, and postural control. We measured displacement over the cardiac cycle using displacement encoding with stimulated echoes magnetic resonance imaging in the cerebellum, brainstem, and spinal cord in 37 CMI subjects and 25 controls. Based on these measurements, we computed strain, translation, and rotation in tracts related to balance. The global strain on all tracts was small (<1%) for CMI subject and controls. Strain was found to be nearly doubled in three tracts for CMI subjects compared to controls (p < 0.03). The maximum translation and rotation were ∼150 μm and ∼1 deg, respectively and 1.5–2 times greater in CMI compared to controls in four tracts (p < 0.005). There was no significant difference between strain, translation, and rotation on the analyzed tracts in CMI subjects with imbalance compared to those without imbalance. A moderate correlation was found between cerebellar tonsillar position and strain on three tracts. The lack of statistically significant difference between strain in CMI subjects with and without imbalance could imply that the magnitude of the observed cardiac-induced strain was too small to cause substantial damage to the tissue (<1%). Activities such as coughing, or Valsalva may produce a greater strain.

Introduction

Chiari malformation Type I (CMI) is characterized by cerebellar tonsillar descent and crowding of neural structures at the foramen magnum (FM). Researchers have studied the biomechanical environment in subjects with and without CMI to understand the cause of symptomatology [1–5]. While researchers have previously examined the relationship between fiber tract connectivity and CMI symptoms, the association is not fully explored [6]. Imbalance is a symptom that affects nearly half of CMI patients and is thought to be caused by damage to neural tracts in the brainstem and cerebellum [7,8]. Paul et al. reported sensory loss related to the dorsal column in 19 of 71 CMI patients (∼27%) through physical examination and 29 (∼40%) of these patients reported loss of balance [7]. Milhorat et al. reported 189 of 364 CMI patients (∼52%) experienced poor equilibrium [8]. Similar findings have also been reported by other researchers [9,10]. These studies suggest possible damage to the neural circuitry pertaining to balance, equilibrium, and postural controls.

In this study, we aim to examine the relationship between self-reported imbalance symptoms and cardiac-induced strain in specific neurological circuits that are known to affect balance, equilibrium, and postural control. Herein, we have selected several neural circuits that are important for balance, equilibrium, and postural controls within the brainstem and cerebellum. In particular,

• The cerebellar unconscious proprioception (i.e., the ability to sense movement, action, and location) input derived from muscle spindles, tendons, and joint capsules reaches the cerebellum through the superior and inferior cerebellar peduncles (SCP, ICP) [11,12]. Vestibular information related to the head's position and motion reaches the cerebellum through the ICP [13,14].

• Reflexive postural adjustment of the head and body results from the influence of the vestibular system on skeletal muscular tone ensuring coordination between different muscle groups through two tracts: the lateral and medial vestibulospinal tract (lVST, mVST) [13].

• The dorsal column medial lemniscus (DCML) pathway is a tract responsible for tactile discrimination (fine touch, pressure, vibration, and two-point discrimination) and conscious proprioception sensation [11,13]. The DCML ascends along the posterior aspect of the spinal cord until it arrives to caudal medulla (the closed part of the medulla) where it traverses anteromedially decussating to the opposite side [13].

As such, we chose the axonal tract subsections and cerebellar peduncles of interest to be SCP, ICP, lVST, mVST, and DCML. A higher strain on a tract/peduncle would be expected to impact its function. For example, strain exceeding unknown limits on lVST and mVST would be expected to present as weakened reflexive postural control.

Altered morphometric and cranio-spinal pressure dynamics during the cardiac cycle have been shown to cause greater brain tissue motion in CMI subjects compared to healthy controls [2–5,15–17]. Numerous research groups have noninvasively quantified cardiac-induced brain tissue motion using various magnetic resonance imaging (MRI) sequences [3–5,15–23], and the reported displacements are listed in Table 1. It has been established that the caudal portion of the cerebellum moves caudally during systole and returns cranially during diastole [2,3,18–20,24,25]. A number of researchers have demonstrated increased motion for cerebellar tonsils and spinal cord in CMI subjects compared to healthy controls (∼2–3 times larger) [2–5,16]. In addition to cardiac-induced brain tissue displacement, researchers have also investigated the associated tissue strain, a biomechanical mechanism by which nerve tissue damage may occur [17,19,20,22,26,27]. Results from these studies are summarized in Table 1, and we can conclude that cardiac-induced brain tissue strain is greater in CMI subjects compared to controls. Recently, there has been an interest in examining the relationship between CMI symptomatology and strain/motion magnitude during the cardiac cycle [3,15,17–19,22]. The symptoms and their findings are shown in Table 2. Results are mixed with some studies showing no relationship to biomechanical parameters [18,19,22] and others showing a modest relationship [3,15,17]. Further research is necessary to determine if a clear relationship exists between CMI symptomatology and cardiac-induced brain tissue displacement/strain.

Table 1.

Literature review of strain and displacement comparisons between CMI (pre/post posterior fossa decompression) and controls

Strain (%)
Structure	Location	CMI (Pre/Post)	Controls	Method	References
Cerebellum	FT	Mean = 1.25 ± 0.25 / 0.8 ± 0.2	Mean = 0.6 ± 0.2	bFFE	Leung et al.
Cerebellum		Tensile: mean = 0.66 ± 0.34, maximum = 1.59 ± 1.11, Compressive: mean = 0.64 ± 0.28, maximum = 1.61 ± 1.02	Tensile: mean = 0.44 ± 0.10, maximum = 0.89 ± 0.26, Compressive: mean = 0.42 ± 0.11, maximum = 0.80 ± 0.21	DENSE	Nwotchouang et al.
Cerebellum	FT	median = 1.48 / 0.51	—	bFFE	Dawes et al.
Cerebellum		—	Tensile: mean = 0.23 ± 0.06, Compressive: mean = 0.30 ± 0.07	DENSE	Pahlavian et al.
Brainstem		—	Tensile: mean = 0.26 ± 0.07, Compressive: mean = 0.38 ± 0.09	DENSE	Pahlavian et al.
Brainstem		Tensile: mean = 0.84 ± 0.30, maximum = 1.63 ± 1.08, Compressive: mean = 0.72 ± 0.19, maximum = 1.27 ± 0.53	Tensile: mean = 0.60 ± 0.17, maximum = 1.07 ± 0.41, Compressive: mean = 0.69 ± 0.15, maximum = 1.09 ± 0.34	DENSE	Nwotchouang et al.
Displacement (μm)
Location		CMI (Pre/Post)	Controls	Method	Reference
Cerebellum		Fastigium: 680 ± 380 / 450 ± 160, tonsil: 990 ± 540 / 630 ± 360	Fastigium: 380 ± 140, Tonsil: 390 ± 140	bFFE	Leung et al.
Cerebellum	FT	Median = 560 / 180	—	bFFE	Dawes et al.
Cerebellum		Mean = 138 ± 55, maximum = 369 ± 185	mean = 67 ± 24, maximum = 120 ± 42	DENSE	Nwotchouang et al.
Cerebellum		Mean = 171 ± 63 / 92 ± 30, maximum = 427 ± 200/153 ± 43	—	DENSE	Eppelheimer et al.
Cerebellum	Tonsil	Mean = 46 ± 25 index	mean = 16 ± 7 index	PCMRI	Pujol et al.
Cerebellum		—	maximum = 105.2 ± 38.2	DENSE	Pahlavian et al.
Cerebellum			70 ± 30	DENSE	Zhong et al.
Brainstem		Obex: 810 ± 380/570 ± 230, PMJ: 810 ± 420/550 ± 190, CMJ: 740 ± 390/480 ± 220	Obex: 440 ± 170, PMJ: 460 ± 160, CMJ: 440 ± 200	bFFE	Leung et al.
Brainstem		Mean = 192 ± 60, maximum = 300 ± 125	Mean = 117 ± 38, maximum = 148 ± 43	DENSE	Nwotchouang et al.
Brainstem		Mean = 210 ± 64/164 ± 54, maximum = 300 ± 95 / 200 ± 56	—	DENSE	Eppelheimer et al.
Brainstem		—	Maximum = 186.7 ± 50.2	DENSE	Pahlavian et al.
Medulla			Mean = 210 ± 60	DENSE	Zhong et al.
Spinal cord	FM	Mean = 530 ± 30/380 ± 20	mean = 230 ± 80	PCMRI	Lawrence et al.
Spinal cord	C2	Systole: mean = 0.9 ± 0.6 ml/s, Diastole: mean = 0.4 ± 0.3 ml/s	Systole: mean = 0.5 ± 0.2 ml/s, Diastole: mean = 0.2 ± 0.1 ml/s	PCMRI	Hofmann et al.
Spinal cord	C2	Mean = 373 ± 198	Mean = 174 ± 47	PCMRI	Alperin et al.

bFFE, balanced fast-field echo; CMJ, cervicomedullary junction; DENSE, displacement encoding with stimulated echoes; FM, foramen magnum; FT, tissue connecting the fastigium of the fourth ventricle to the cerebellar tonsil; PMJ, pontomedullary junction; PCMRI, phase contrast magnetic resonance imaging.

Statistically significant differences are in bold.

Table 2.

Literature review of CMI symptomatology related to strain or displacement

Structure	Symptoms	Parameters	Findings	Reference
Cerebellum FT	Valsalva headache	Strain	strain = 1.5 ± 0.7% (with), 0.9 ± 0.6% (without)	Leung et al.
Cerebellum FT	headache, Valsalva headache	Strain and displacement	not significantly different	Dawes et al.
Cerebellar tonsils	Headache	Displacement and reduction in CSF space at FM	correlation with displacement (r = 0.76) and with a combination of displacement and reduction of CSF space (r = 0.85)	Pujol et al.
Cerebellar tonsils	A list of ten symptoms	Displacement	Correlation for snoring (θ = 0.57), and trending for ataxia (θ = 0.35)	Collins et al.
Cerebellum, brainstem	A list of ten symptoms	Strain and displacement	Not significantly different	Nwotchouang et al.
Cerebellum, brainstem	A list of 23 symptoms	Displacement	Not significantly different	Eppelheimer et al.

CSF: Cerebrospinal fluid, FM: foramen magnum, FT: tissue connecting the fastigium of the fourth ventricle to the cerebellar tonsil.

Statistically significant differences are in bold.

We hypothesize that CMI subjects have a higher cardiac-induced strain on the SCP, ICP, lVST, mVST, and DCML than healthy controls, and that elevated strain negatively impacts CMI patients' balance and postural control as reported clinically. In the present study, we utilized the displacement field over the cardiac cycle measured by DENSE MRI to compute strain during the cardiac cycle on specific axonal tract subsections in the cerebellum, brainstem, and spinal cord. Mechanical damage to these axonal tracts due to strain, may alter function. While the strain from cardiac-induced motion is small, for the first time, we have examined whether the continuous repetition of these small strains results in loss or altered function which could explain imbalance in CMI subjects.

Materials and Methods

Participants.

In this study, 37 CMI subjects consisting of 6 males and 31 females with mean age of 36 ± 11 years, body mass index (BMI) of 29.3 ± 6.0 kg/m², and cerebellar tonsillar position (CTP) of 13.1 ± 6.0 mm. In addition, 25 healthy controls consisting of 8 males and 17 females with mean age of 25 ± 4 years and with BMI of 22 ± 2 kg/m² [19] were included in the study. All subjects were scanned at the Center for Systems Imaging at Emory University School of Medicine between January 2017 and March 2022 under a protocol approved by the institutional review board (IRB00008711). CMI subjects had no additional neurological disorders. All participants signed a written consent before their enrollment in the study. Along with the MRI scans, the symptomatology was collected by interviewing CMI subjects at the time of their presurgical scan using approved questionnaires [19]. Based on reported symptoms in the questionnaires regarding balance, CMI subjects were divided into two subsets, CMI with imbalance (n = 16), and CMI without imbalance (n = 21). Out of the 37 CMI subjects, 11 subjects had a syrinx. Two CMI subjects (one with, and one without imbalance) were excluded from all VST analyses because of the extension of the syrinx into the region of interest (ROI).

Imaging Protocol.

A Prisma Fit 3T MRI scanner (Siemens Medical Solutions, Erlangen, Germany) with 20-channel head coil was used to acquire sagittal T1- and T2-weighted scans of the entire brain. All DENSE MRI scans were acquired in the midsagittal plane using two-dimensional (2D) spiral cine technique [21,28]. The image acquisition used in this paper followed the same methodology as the previous works by our group [18–20]. Each DENSE scan generates a set of magnitude images and two sets of phase images with the displacement field encoded in two orthogonal directions: anteroposterior (AP) or the x-direction; and cephalocaudal (CC) or the y-direction. To encode displacement over the cardiac cycle, a peripheral pulse unit-gating was employed to capture peak systolic displacement. Due to heart rate variation and using a constant temporal resolution per frame (34 ms), the number of frames in the cardiac cycle varies among patients. Subsequently, the scan time varies based on the subject's heart rate (1.5 min per direction for a heart rate of 60 bpm). DENSE imaging parameters were as follows: the flip angle = 15 deg, encoding frequency = 0.6 cycle/mm, spiral interleaves per heartbeat = 2, total spiral interleaves per image = 192, field of view = 256 × 256, reconstruction matrix = 256 × 256, pixel size = 0.86–0.94 × 0.86–0.94 mm, and slice thickness = 8 mm.

Pre-Processing.

An in-house developed matlab code (MathWorks, Natick, MA) was used to process the 2D DENSE images in the midsagittal plane following the procedure previously described by our group [18–20]. The DENSE images were inspected for aliasing, and a manual phase unwrapping was performed when necessary. The magnitude image was used to apply a mask over the ROI utilizing the paintbrush tool in the developed code. The 2D displacement encoded in DENSE was used to calculate lengthening, strain, translation, and rotation.

Regions of Interest.

Regions of Interest (ROI) selection was done visually and based on the location of tracts relative to well-defined anatomical landmarks using anatomy books/publications as our guide [11,13,14,29]. The strain, translation, and rotation of each tract were analyzed using two ROI boxes placed at two locations on the structure in the midsagittal plane (Fig. 1) in the following manner:

Fig. 1.

Location of the ROI boxes on an anatomical MRI of a CMI patient at the midsagittal plane: (a) ICP, (b) ICP2, (c) SCP, (d) VST, and (e) FT. The route of DCML was depicted in (d).

Inferior cerebellar peduncle (ICP and ICP2).

The anterior ICP (aICP) box was placed in the medulla at the level of the pontomedullary junction just ventral to the 4th ventricle (Fig. 1(a)). The posterior ICP (pICP) box was placed at the fastigium of the fourth ventricle in the cerebellum. To investigate the strain on the anterior half of the ICP, the location of pICP was moved halfway between aICP and the fastigium (Fig. 1(b)) which coincides with the nodulus part of the flocculonodular lobe of the cerebellum (pICP2). This line was called ICP2 and provides a more localized strain measurement of the ICP tract.

Superior cerebellar peduncle.

The anterior SCP (aSCP) box was placed in the midbrain at the level of the inferior colliculus just ventral to the cerebral aqueduct (Fig. 1(c)). The posterior SCP (pSCP) box was placed in the cerebellum at the fastigium of the fourth ventricle.

Vestibulospinal tract. Due to the proximity of lVST and mVST at vestibular nuclei and in the cervical spinal cord, we employed two ROI boxes for both tracts and referred to them as the vestibulospinal tract (VST) (Fig. 1(d)). The rostral VST (rVST) box was placed at the start of the tract (∼3–4 mm anterior to the posterior aspect of the pontomedullary junction to ensure ROI placement on tissue and not fluid), and the caudal VST (cVST) was placed at the center of the spinal cord at the level of C1/C2 to ensure a fluid-free margin of ∼ 2 mm.

Fastigium-tonsil. The FT was investigated by placing the rostral box (rFT) at the fastigium of the fourth ventricle, and a caudal box (cFT) at the tip of the cerebellar tonsil (Fig. 1(e)). There is no tract pertaining to the balance function in the FT. However, we included it because motion and strain of the tonsils are often mentioned in the literature as important in CMI subjects [17,22].

Dorsal column medial lemniscus. The rostral DCML box (rDCML) is located at the posterior aspect of the spinal cord at the level of McRae line about two voxels away from the CSF fluid. The caudal DCML box (cDCML) is located at the posterior aspect of the spinal cord at the level of C1/C2 about two voxels away from the CSF fluid. Because of the close proximity to the VST, the DCML was not investigated and results from the VST were considered representative to the DCML.

Signal-to-noise ratio (SNR) in CMI subjects was found to be 37.5 ± 13.1 and 14.1 ± 2.9 at the start and end of the cardiac cycle, respectively. Similarly for controls, SNR was found to be 33.4 ± 16.2 and 15.2 ± 3.7 at the start and end of the cardiac cycle, respectively. SNR was calculated as the ratio of the average of the signal (10 × 10 voxels ROI in the brainstem) over the standard deviation of the noise (10 × 10 voxels outside of the head) throughout the cardiac cycle on magnitude images.

Post-Processing.

The 2D displacement field image was analyzed in the midsagittal plane. Each ROI box was a binary mask of 3 × 3 voxels that measured approximately 2.7 × 2.7 × 8 mm in the CC, AP, and mediolateral (through-plane), respectively. The position vector is defined by the coordinates of the ROI box and changes during the cardiac cycle based on the tissue displacement measurement. The length of the line connecting the two ROI boxes is also a function of time and was calculated using Eq. (1) below. Maximum lengthening (L_max) and global maximum tensile strain (ε_max) were calculated using Eqs. (2), and (3), respectively. This definition of the lengthening and strain considers the tissue in tension only (no compression) as the undeformed configuration is unknown. Note: we called the strain quantified using two ROIs 3 × 3 voxels global strain to differentiate it from local strain (strain distribution)

$$L\left(t\right)=\hspace{0.17em}\sqrt{{\left({X\left(t\right)}_{{\mathrm{Box}}_1}-{X\left(t\right)}_{{\mathrm{Box}}_2}\right)}^2+{\left({Y\left(t\right)}_{{\mathrm{Box}}_1}-{Y\left(t\right)}_{{\mathrm{Box}}_2}\right)}^2}$$ (1)

$${\mathit{Len}}_{\max }=L_{\max }-\hspace{0.17em}{L}_{\min }$$ (2)

$${\epsilon }_{\max }=\left(\frac{{\mathrm{Len}}_{\max }}{L_{\min }}\right)\mathrm{*}100$$ (3)

The local strain distribution along the line connecting the two ROI boxes was computed as follow: a set of boxes along each line comprised of the average of one voxel on the line, and the two voxels adjacent to it (1 × 3) were determined. Subsequently, the strain was computed using Eq. (3) for two boxes spaced seven voxels apart to obtain strain distribution at each time point in the cardiac cycle while minimizing noise from spurious signals. The maximum strain during the cardiac cycle was reported at specific locations along the line. Since the length of the lines among subjects was different, the one-to-one comparison of each calculated localized strain and the subsequent averaging were deemed impossible. As such as averaging of the localized strains were made not based on the actual distance on the line but based on the relative distance from the caudal end. In particular, the relative distance was chosen in 5% increments from the caudal end, and the corresponding localized strain data were averaged for each increment (location) for each participant. Subsequently, the mean distributions were calculated based on the aligned localized strains. The average of a given increment was not computed for cases with missing subject data. The strain distribution was only calculated for cases when the line connecting the two ROI boxes was continuous within the soft tissue and did not cross over cerebrospinal fluid (VST and FT).

The translation was quantified as the motion of the line midpoint throughout the cardiac cycle. The maximum translation value during the cardiac cycle was used for the comparison. The rotation of the line was quantified as the maximum angle range during the cardiac cycle. Figure 2 illustrates an example of the tissue motion of rVST, and cVST of one CMI participant.

Fig. 2.

The uncertainty in the path trajectory due to an estimated error of 15 μm in the x and y directions (horizontal and vertical error bars, respectively) of rVST and cVST of a CMI patient. The green (lower) box represents the start of the cardiac cycle. The red (upper) box represents the last acquisition time point of the DENSE sequence. The difference in x and y locations between the green and red boxes is due, in part, to the error in the x- and y-direction displacement and because the scanner only captures 90% of cardiac cycle in prospective gating which is required for DENSE MRI.

Statistical Analysis.

The analysis of variance (single-factor, between-subjects ANOVA) and posthoc adjustment using Tukey tests were used to compare strain, translation, and rotation of each tract in controls, CMI with, and without imbalance. Independent group t-tests were performed to compare strain, translation, and rotation of each tract in controls and all CMI subjects, and also to compare CMI subjects with and without imbalance. Local strain distribution for controls, CMI with and without imbalance was compared using independent group t-tests. Correlations were used to examine the relationship between global tract strain in CMI subjects and four parameters: CTP, age, BMI, and the presence of syrinx. Results were reported as mean ± standard deviation. A p-value less than 0.05 was considered significant.

Results

Neural Tract Strain.

While the global strain on all tracts was small (<1%) for CMI and controls, we found statistically significant differences for groups (CMI with and without imbalance and controls) and tracks. A series of ANOVAs showed significant strain comparisons between CMI with imbalance and controls for ICP, ICP2, and VST (Table 3). However, this series did not show significant strain comparison between CMI without imbalance and controls for all tracts. The global strain on the ICP, ICP2, and VST in CMI with imbalance was about twice that of the controls (p < 0.023).

Table 3.

A series of ANOVAs showing comparisons between CMI with and without imbalance as compared to controls for strain, translation, and rotation for each tracts

Maximum strain (%)
	CMI without			CMI with			Controls
Tract	n	Mean	STD	n	Mean	STD	n	Mean	STD	F	p
ICP	21	0.45	0.20	16	0.75∗	0.72	25	0.30	0.16	6.43	0.003
ICP2	21	0.80	0.55	16	1.14∗	0.92	25	0.59	0.33	4.03	0.023
VST	20	0.85	0.47	15	1.03∗	0.72	25	0.51	0.24	6.17	0.004
SCP	21	0.35	0.16	16	0.34	0.16	25	0.30	0.12	0.78	0.462
FT	21	0.45	0.21	16	0.48	0.23	25	0.38	0.19	1.26	0.290
Maximum translation (micron)
	CMI without			CMI with			Controls
Tract	n	Mean	STD	n	Mean	STD	n	Mean	STD	F	p
ICP	21	66∗∗	27	16	71∗	36	25	40	21	8.19	0.0007
ICP2	21	51	30	16	64∗	39	25	34	15	5.80	0.005
VST	20	131	56	15	140∗	69	25	93	45	4.29	0.018
SCP	21	37	16	16	38	19	25	32	8	1.13	0.330
FT	21	100∗∗	93	16	108∗	53	25	53	28	4.82	0.012
Maximum rotation (deg)
	CMI Without			CMI With			Controls
Tract	n	Mean	STD	n	Mean	STD	n	Mean	STD	F	p
ICP	21	0.4∗∗	0.1	16	0.4∗	0.2	25	0.3	0.1	11.2	0.0001
ICP2	21	0.5∗∗	0.2	16	0.7∗	0.4	25	0.4	0.2	8.6	0.0005
VST	20	0.4∗∗	0.2	15	0.5∗	0.3	25	0.2	0.1	9.8	0.0002
SCP	21	0.2	0.1	16	0.2	0.1	25	0.2	0.0	2.1	0.132
FT	21	0.4∗∗	0.3	16	0.5∗	0.2	25	0.2	0.1	7.4	0.001

Superscript (∗) refers to significance difference between CMI with imbalance and controls, and (∗∗) between CMI without imbalance and controls.

Note: FT is not a tract pertaining to balance, rather it was added due to the importance of the tonsil in Chiari literature. p-values are in bold when significant (< 0.05).

We also found differences in global strain using paired comparisons (t-test). Global strain was found to be nearly two times higher in three tracts (ICP, ICP2, and VST) for CMI subjects compared to controls (p < 0.03) as shown in Fig. 3 and Table 4. The highest strain observed in CMI subjects was on the ICP2 where the strain was (0.95 ± 0.75%) 1.6 times greater than that of the controls (0.59 ± 0.33%, p = 0.029). The second highest strain in CMI was on the VST with a magnitude of 0.93 ± 0.59%, 1.8 times greater than that of the controls (0.51 ± 0.24%, p = 0.002). Finally, the strain on the ICP was 0.58 ± 0.51%, which is about twice as high as that of the controls (0.30 ± 0.16%, p = 0.011). There was no significant difference between the strain on SCP, or FT in CMI subjects compared to the controls. There was no significant difference between strain on the analyzed tracts in CMI subjects with imbalance compared to those without imbalance (Table 5 and Fig. 3).

Fig. 3.

Strain in each tract for controls (green, 38-62) and CMI subjects with (red, 22-37) and without imbalance (blue, 1-21). Strain in tracts was sorted by ICP strain magnitude, such that subject numbers are consistent for each plot. The mean and standard deviation of each group are shown as dashed lines and error bars, respectively. X-axis represents the subject number. Two CMI subjects (one with and one without imbalance) were excluded from the VST analyses due to the presence of a syrinx. *p < 0.05, **p < 0.01.

Table 4.

Strain, translation, and rotation of the tracts in CMI participants compared to controls reported as mean, standard deviation

Maximum strain (%)
	CMI			Controls
Tract	n	Mean	STD	n	Mean	STD	p
ICP	37	0.58	0.51	25	0.30	0.16	0.011
ICP2	37	0.95	0.75	25	0.59	0.33	0.029
VST	35	0.93	0.59	25	0.51	0.24	0.002
SCP	37	0.34	0.16	25	0.30	0.12	0.223
FT	37	0.46	0.22	25	0.38	0.19	0.131
Maximum Translation (micron)
	CMI			Controls
Tract	n	Mean	STD	n	Mean	STD	p
ICP	37	68	31	25	40	21	0.0002
ICP2	37	57	35	25	34	15	0.003
VST	35	135	61	25	93	45	0.005
SCP	37	37	17	25	32	8	0.142
FT	37	104	78	25	53	28	0.003
Maximum rotation (degree)
	CMI			Controls
Tract	n	Mean	STD	n	Mean	STD	p
ICP	37	0.4	0.2	25	0.3	0.1	0.00001
ICP2	37	0.6	0.3	25	0.4	0.2	0.0003
VST	35	0.4	0.2	25	0.2	0.1	0.0002
SCP	37	0.2	0.1	25	0.2	0.0	0.460
FT	37	0.4	0.3	25	0.2	0.1	0.001

Note: p-values are in bold when statistically significant (<0.05).

Table 5.

Strain, translation, and rotation of the tracts in CMI participants with and without imbalance reported as mean, standard deviation

Maximum strain (%)
	CMI with imbalance			CMI without imbalance
Tract	n	Mean	STD	n	Mean	STD	p
ICP	16	0.75	0.72	21	0.45	0.20	0.072
ICP2	16	1.14	0.92	21	0.80	0.55	0.169
VST	15	1.03	0.72	20	0.85	0.47	0.368
SCP	16	0.34	0.16	21	0.34	0.16	0.899
FT	16	0.48	0.23	21	0.45	0.22	0.637
Maximum translation (micron)
	CMI with imbalance			CMI without imbalance
Tract	n	Mean	STD	n	Mean	STD	p
ICP	16	71	36	21	66	27	0.628
ICP2	16	64	39	21	51	30	0.278
VST	15	140	69	20	131	56	0.645
SCP	16	38	19	21	37	16	0.814
FT	16	108	53	21	100	93	0.750
Maximum rotation (deg)
	CMI with imbalance			CMI without imbalance
Tract	n	Mean	STD	n	Mean	STD	p
ICP	16	0.4	0.2	21	0.4	0.1	0.811
ICP2	16	0.7	0.4	21	0.5	0.2	0.258
VST	15	0.5	0.3	20	0.4	0.2	0.183
SCP	16	0.2	0.1	21	0.2	0.1	0.105
FT	16	0.5	0.2	21	0.4	0.3	0.333

Note: FT is not a tract related to balance. p-values are in bold when statistically significant (< 0.05).

The local strain distribution was obtained for the VST and FT (Fig. 4 and Table 6) where the strain distribution values were typically 1–3% while global strain values were 0.5–1%. The mean strain distribution showed higher values caudally that decreased toward the cranium. The strain distribution along FT in controls remained constant and ∼1.6 times larger than the global strain however, strain values were small (∼0.6%). The mean strain along FT in CMI (1.38 ± 0.43%) more than doubled that of the controls (0.62 ± 0.05%, p < 10⁻⁶). In contrast, the mean strain along VST in CMI (1.97 ± 0.64%) was not significantly different from that of the controls (1.63 ± 0.61%). The average strain along FT and VST was not significantly different between CMI with and without imbalance.

Fig. 4.

Strain distribution along FT (cerebellum) and VST (brainstem) of each participant using a span of (seven voxels) for controls and Chiari subjects with and without imbalance. Positions along the tracts were normalized by their length starting from the caudal end and percentage locations were averaged for each 5% increment. The caudal and rostral ends of the tract are marked. The solid black lines and the shaded areas represent the mean and standard deviation, respectively. The dashed lines represent the global mean of the strain using the classical (3 × 3 voxels) technique.

Table 6.

Strain values (strain distribution) along the tract. Top: comparisons between CMI and controls. Bottom: comparison between CMI with and without imbalance.

Strain values along the tract (%)
	CMI			Controls
Tract	n	Mean	Std	n	Mean	Std	p
FT	37	1.38	0.43	25	0.62	0.05	3.5 × 10−7
VST	35	1.97	0.64	25	1.63	0.61	0.128
Strain values along the tract (%)
	CMI with imbalance			CMI without imbalance
Tract	n	Mean	Std	n	Mean	Std	p
FT	16	1.51	0.55	21	1.28	0.37	0.170
VST	15	1.97	0.71	20	1.97	0.60	0.996

Note: FT is not a tract related to balance. p-values are in bold when statistically significant (< 0.05).

CTP was found to have a moderate correlation with the global strain on ICP2, VST, and SCP (p < 0.03). The correlation coefficients for ICP2, VST, and SCP were 0.50, 0.43, and 0.35, respectively (Fig. 5). Age, BMI, and the presence of syrinx did not correlate with the strain on any of the tracts.

Fig. 5.

Strain in ICP2, SCP, and VST tracts were plotted against cerebellar tonsillar position (CTP). Dashed lines represented linearly fitted data.

Neural Tract Translation.

While the translation for all tracts was small (<150 μm) for CMI and controls, we found statistically significant differences for groups and tracks. A series of ANOVAs showed significant translation comparisons between CMI with imbalance and controls for ICP, ICP2, VST and FT (Table 3) with only ICP and FT for CMI without imbalance and controls.

We also found differences in translation using paired comparisons (t-test). However, translation was 1.5–2 times greater in CMI compared to controls in ICP, ICP2, VST, and FT (p < 0.005) as shown in Table 4, and Fig. 6. The highest translation was seen on the VST in CMI (135 ± 61 μm), which was 1.5 times greater than that of the controls (93 ± 45 μm, p = 0.005). The second highest observed translation in the CMI group was seen on the FT with a magnitude of 104 ± 78 μm that doubles that of the controls (53 ± 28 μm, p = 0.03). For ICP, and ICP2, the translation in CMI was 1.7 times greater than that of the controls (p < 0.003). The translation on SCP in CMI was not significantly different from that of the controls. There was no significant difference between the translation of the analyzed tracts in CMI subjects with imbalance compared to those without imbalance (Table 5 and Fig. 6).

Fig. 6.

Translation in each tract for controls (green, 38-62) and CMI subjects with (red, 22-37) and without imbalance (blue, 1-21). Translation in tracts was sorted by the ICP strain magnitude, such that subject numbers are consistent for each plot. The mean and standard deviation of each group were shown as dashed lines and error bars, respectively. X-axis represents the subject number. Two CMI subjects (one with and one without imbalance) were excluded from the VST analyses due to the presence of a syrinx. *p < 0.05, ^**p < 0.01, ^***p < 0.001.

Neural Tract Rotation.

While the rotation for all tracts was small (<1 deg) for CMI and controls, we found statistically significant differences for groups and tracks. A series of ANOVAs showed significant rotation comparisons between CMI with and without imbalance versus controls for ICP, ICP2, VST, and FT (Table 3).

We also found differences in rotation using paired comparisons (t-test). The rotation in CMI was 1.5–2 times greater than that of the controls for ICP, ICP2, VST, and FT (p < 0.001) as shown in Table 4, and Fig. 7. The highest rotation was observed for ICP2 with a value of (0.6 ± 0.3 deg) which was 1.5 times greater than that of the controls (0.4 ± 0.2 deg, p = 0.0003). The rotation for ICP in CMI was 0.4 ± 0.2 deg which was 1.3 times greater than that of the controls (0.3 ± 0.1 deg, p = 0.00001). The rotation for VST and FT in CMI was 0.4 deg, doubled that of the controls (0.2 ± 0.1 deg, p < 0.001). The rotation of the SCP in CMI was not significantly different from that of controls. There was no significant difference between the rotation of the analyzed tracts in CMI subjects with imbalance compared to those without imbalance (Table 5 and Fig. 7).

Fig. 7.

Rotation in each tract for controls (green, 38-62) and CMI subjects with (red, 22-37) and without imbalance (blue, 1-21). Rotation in tracts was sorted by the ICP strain magnitude, such that subject numbers are consistent for each plot. The mean and standard deviation of each group were shown as dashed lines and error bars, respectively. X-axis represents the subject number. Two CMI subjects (one with and one without imbalance) were excluded from the VST analyses due to the presence of a syrinx. *p < 0.05, ^**p < 0.01, ^***p < 0.001.

Discussion

In this study, we quantified cardiac-induced strain, translation, and rotation on tracts/peduncles pertaining to balance in controls; and CMI subjects with and without imbalance. The results demonstrate that CMI subjects have nearly double the strain on the VST, ICP, and ICP2, although strains were generally small (<1%). The maximum value along the FT for CMI was more than double that of the controls. The translation was 1.5–2 times greater in CMI subjects compared to controls for ICP, ICP2, VST, and FT, although distances traveled were generally small (<150 micron). Similarly, rotation was 1.5–2 times greater in CMI subjects compared to controls for ICP, ICP2, VST, and FT although angles were again generally small (<1 deg). A moderate correlation was found between CTP and the strain on ICP, ICP2, and VST tracts in CMI subjects. Finally, in the present study, we did not find a significant difference between CMI subjects with and without self-reported subjective imbalance for the three biomechanical markers.

It is worth noting that some of the analyzed tracts, like SCP, VST, and FT, have different lengths between CMI and controls. The tissues may have similar displacement values in CMI and controls. As such, the calculated linear global strain was a reasonable metric to differentiate between the deformation of these tracts in CMI and controls. This difference in tracts length is because CMI subjects have a stretched cerebellum, settling brainstem, shorter fastigium height, and herniated tonsils [30]. Statistical comparison of the lengths of the tracts in CMI and controls is tabulated in Table 7. The length of FT was determined to be 7.8 mm (28%) greater in CMI as compared to controls (p < 0.0001). The difference between FT length in CMI and that of controls is expected as the tonsil is located at least 5 mm below McRae's line in CMI while it is above the line in controls. In contrast, the VST tract was 6.4 mm (16%) shorter in CMI as compared to controls (p < 0.0001). The shorter VST tract length in CMI is also expected due to the brainstem settling, measured by the pons height, in CMI literature [31]. Finally, the ICP, ICP2, and SCP tract lengths were not significantly different in CMI as compared to those of the controls. In short, using mechanical strain made the comparison of the CMI and controls tract deformation independent of differences in the observed initial lengths in these groups.

Table 7.

Tracts lengths in CMI participants compared to the controls

	CMI—tract length (mm)					Controls—tract length (mm)
Tract	n	Mean	Std	Max	Min	n	Mean	Std	Max	Min	p
ICP	37	19.3	2.0	23.1	15.0	25	20.0	1.8	24.3	16.5	0.143
ICP2	37	10.8	1.3	13.8	7.9	25	11.3	1.4	14.6	9.1	0.146
VST	35	34.3	3.9	41.6	25.8	25	40.7	2.5	44.0	33.7	1.6 × 10−9
SCP	37	21.5	4.4	31.0	15.7	25	19.6	2.2	24.3	16.5	0.053
FT	37	35.3	4.9	50.8	27.5	25	27.5	3.2	32.8	21.8	3.2 × 10−9

Note: p-values are in bold when statistically significant (<0.05).

The lack of statistically significant difference in the present study between strain in CMI subjects with and without imbalance could imply that the magnitude of the observed cardiac-induced strain was too small to cause substantial damage to the tissue (<1%). The low strain magnitude could also be the reason previous researchers found only limited or no correlation with cardiac-induced strain. It is challenging to measure brain tissue strain during activity, but it may be necessary to determine a strong association with CMI symptomatology. It is unclear if the current state of MR technology is sufficient to measure tissue strain during activities such as Valsalva maneuver or coughing. We believe that tissue strain analysis during a physiological activity is the future direction of this research.

One interesting aspect of the strain results in both controls and CMI subjects is that the brainstem (VST) has greater strain in general than the cerebellum (FT). The mean strain distribution along the VST was 2.6 times greater than that of the FT in controls (p < 10⁻⁶). In CMI, the mean strain distribution along VST was 1.5 times higher than that of FT (p = 0.004). The above-mentioned comparisons demonstrate that the brainstem (VST) has a generally higher strain than the cerebellum (FT) in both physiological and pathological states. However, the cerebellum experiences a greater difference in strain magnitude between CMI and controls than the brainstem. This difference between the cerebellum and brainstem is evidence that the impact of crowding is greater on cerebellar tissues than the brainstem.

Similarly, the strain results of the current study are similar to those reported by several other researchers [17,19,20,22,26,27,32]. For example, it has been established that the cardiac-induced brain tissue strain is higher caudally than rostrally [19,20,22]. Our localized strain along VST and FT confirms this finding as illustrated in Fig. 4. The mean localized strain for FT in the present study for CMI and controls were 1.38 and 0.62% compared to 1.25 and 0.6% from Leung et al. [17]. However, the present study showed only 0.46 and 0.38% for CMI and controls, respectively for the global strain value. This strain comparison indicates that the global strain values may underestimate the strain values for those structures compared to our strain distribution method or the pixel-by-pixel method of Leung et al.

The precise mechanism that causes greater brain tissue strain during the cardiac cycle in CMI is unclear. However, this mechanism could be simply a pressure alteration due to an increased cerebrospinal fluid velocity with area reduction through the Bernoulli effect. The correlation observed between strain and CTP in three tracts is evidence that the level of crowding plays an essential role in the magnitude of brain tissue strain during the cardiac cycle. To the authors' knowledge, this is the first report of a correlation between brain tissue strain and CTP. Note that age, BMI, and the presence of syrinx did not correlate with the strain on any of the tracts. Lawrence et al. found a moderate positive correlation (r = 0.49) between CTP and spinal cord motion (p = 0.0003) [16]. Eppelheimer et al. found no correlation between CTP and neural tissue displacement [18]. In addition, crowding, as measured by the reduction of subarachnoid space at the FM, was shown to improve the correlation that Pujol et al. found between tonsillar motion and cough-associated headaches [3]. Researchers have shown tissue strain/motion reduction after posterior fossa decompression surgery, further implicating the importance of crowding in tissue strain [17,18,22].

This study has several limitations. First, the locations and dimensions of the subject's neurological tracts are unknown. Hence anatomical landmarks were used to estimate their location. The author who conducted these measurements (MAS) is a trained physician and has experience identifying anatomical landmarks. Second, we considered strain in tension because the undeformed configuration (zero lengthening/strain) was unknown. Third, our displacement analysis was performed in the midsagittal plane, neglecting the mediolateral displacement (i.e., z-direction). However, our group showed that the mediolateral displacement is smaller than in the midplane [20], especially with a large slice thickness (8 mm). Fourth, there is reduction in SNR throughout the cardiac cycle caused by a decay in the T1 signal [19], and this may impact the accuracy of the strain measurements. Signal fading over the cardiac cycle is due to T1 effects on the stored magnetization after the DENSE tagging pulse [28]. To test the effect of the last third of the cardiac cycle on the magnitude of biomechanical markers, we repeated the analysis using only the first two-thirds of the cardiac cycle. We found that the significance of the statistical comparisons of biomechanical markers for CMI and controls and also for CMI with and without imbalance remained the same. All biomechanical markers for each group decreased in magnitude but the trends remained the same. Fifth, although DENSE MRI is capable of quantifying small brain tissue strain noninvasively, strain values have uncertainties ranging between 0.07 and 0.28% due to an error in the displacement measurement. The uncertainty in the strain was dependent on the length of the tract with shorter tracts having greater uncertainty. The analysis assumed a peak displacement error of 15 μm based on previously reported error analyses of DENSE MRI [33]. We considered the error in the displacement to be the same in the x- and y-directions and constant throughout the cardiac cycle since the scanner computes the displacements of each phase with respect to the start of the cycle. The difference in x and y locations between the first and last time points of the cardiac cycle (Fig. 2) is due, in part, to the error in the x- and y-direction displacement and because the scanner only captures 90% of cardiac cycle in prospective gating which is required for DENSE MRI. Sixth, while clinical data were collected using questionnaires administered by a nurse, assessment of balance could be improved with an objective and quantitative methodology [34,35]. In addition, other factors that may impact balance and should be considered such as inner ear and/or vision problems, peripheral neuropathy, and medications. Seventh, the assumption that DCML pathway experiences similar strain as the VST due to their proximity may not be true for all subjects (Fig. 1(d)). The reason for adopting such an assumption was that we did not expect an abrupt change in the displacement gradient within a few voxels, and, as such, we assumed that the DCML strain should be similar to that of the VST. Finally, the results presented here do not account for the respiratory effect on displacement or strain analysis. Sloots et al. conducted a study on healthy subjects using DENSE MRI to quantify volumetric strain in the cerebrum and showed that cardiac-induced strain was three times greater than that of respiration [32]. Also, Yildiz et al. examined the CSF motion of healthy individuals using real-time PCMRI and found that the cardiac component was associated with a larger CSF velocity compared to the respiratory component [36]. Thus, it is reasonable to expect larger motion and strain during respiration for neural tissue in general. However, it is unclear whether respiratory-induced strain would be different between CMI subjects with and without imbalance. Further research is needed to better understand the effect of respiration on brain tissue motion and strain in CMI and controls. The BMI was different for the CMI subjects (29.3) and the control group (22.0 kg/m²); however, we do not expect that this difference would impact the results as previous research has shown that BMI is not correlated with brain tissue displacement or strain [19].

Conclusion

While greater strain (by a factor of 2) was observed between CMI and controls on many of the tracts examined herein, we did not find a difference between strain in CMI subjects with and without imbalance. This result raises the question whether cardiac-induced strain is sufficient in magnitude to cause tissue damage with respect to imbalance. Brain tissue strain during various activities such as coughing; or Valsalva may be better indicators of tissue damage in CMI subjects than cardiac-induced strain. However, measurement of activity-related strain may be difficult using MRI methods. In addition, the results that strain in the cerebellum and brainstem are correlated with CTP is evidence that the crowding near the FM is an important factor in brain tissue strain and may help to understand the pathophysiology of CMI.

Nomenclature

All abbreviations used in this paper are listed in alphabetical order.

2D =

two-dimensional

aICP =

anterior end of the inferior cerebellar peduncle line

AP =

anteroposterior

aSCP =

anterior end of the superior cerebellar peduncle line

bFFE =

balanced fast-field echo

BMI =

body mass index

CC =

cephalocaudal

cDCML =

caudal end of dorsal column medial lemniscus

cFT =

caudal end of the fastigium-tonsil line

CMI =

Chiari malformation Type I

CMJ =

cervicomedullary junction

CSF =

cerebrospinal fluid

CTP =

cerebellar tonsillar position

cVST =

caudal end of the vestibulospinal tract

DCML =

dorsal column medial lemniscus

DENSE =

displacement encoding with stimulated echoes

FM =

foramen magnum

FT =

fastigium-tonsil

ICP =

inferior cerebellar peduncle

ICP2 =

anterior half of the inferior cerebellar peduncle line

lVST =

lateral vestibulospinal tract

MRI =

magnetic resonance imaging

mVST =

medial vestibulospinal tract

PCMRI =

phase contrast magnetic resonance imaging

pICP =

posterior end of the inferior cerebellar peduncle line

pICP2 =

posterior end of the anterior half of the inferior cerebellar peduncle line

PMJ =

pontomedullary junction

pSCP =

posterior end of the superior cerebellar peduncle line

rDCML =

rostral end of dorsal column medial lemniscus

rFT =

rostral end of the fastigium-tonsil line

ROI =

region of interest

rVST =

rostral end of the vestibulospinal tract

SCP =

superior cerebellar peduncle

VST =

vestibulospinal tract