Diagnostic utility of parasagittal measurements of tonsillar herniation in Chiari I malformation

Abstract

Background and purpose

Although the cerebellar tonsils are parasagittal structures, the extent of tonsillar herniation (ETH) in Chiari I malformation (CMI) is currently measured in the midsagittal plane. We measured the ETH of each cerebellar tonsil in the parasagittal plane and assessed their diagnostic utility by comparing them to the midsagittal ETH measurements in predicting cough-associated headache (CAH), an indicator of clinically significant disease in CMI.

Methods

Eighty-five CMI patients with 3D-MPRAGE images were included. Neurosurgeons determined the presence of CAH. Sagittal images were used to measure ETH in the midsagittal (MS_ETH) and parasagittal planes (by locating tonsillar tips on each side on reformatted coronal images). Given the parasagittal ETH (PS_ETH) asymmetry in the majority of cases, they were considered Smaller_PS_ETH or Larger_PS_ETH. The accuracy of ETH measurements was assessed by the receiver operating characteristic (ROC) curve.

Results

Of 85 patients, 46 reported CAH. ROC analysis showed an area under the curve (AUC) of 0.78 for Smaller_PS_ETH significantly better than 0.65 for MS-ETH in predicting CAH (p = 0.001). An AUC of 0.68 for Larger_PS_ETH was not significantly different from MS_ETH. The sensitivity and specificity of predicting CAH were 87% and 28% for MS_ETH >6 mm versus 90% and 46% for Smaller_PS_ETH >6 mm, and 52% and 67% for MS_ETH >9 mm versus 48% and 87% for Smaller_PS_ETH >9 mm. At ETH >15 mm, no differences were seen between the measurements.

Conclusions

Diagnostic utility of ETH measurements in detecting clinically significant CMI can be improved by parasagittal measurements of the cerebellar tonsillar herniation.

Introduction

Chiari I malformations (CMI), characterized by caudal displacement of the hindbrain into the upper cervical spinal canal, is radiologically diagnosed on imaging by measuring the extent of cerebellar tonsillar herniation of ≥5 mm beyond the plane of the foramen magnum. The extent of herniation is classically determined in the midsagittal plane by measuring the distance between the cerebellar tonsil’s tip from the line connecting basion to opisthion (McRae line).^1–6 However, this characterization is anything but perfect, as many patients fulfilling this criterion are asymptomatic, and others have typical CMI symptoms despite having <5 mm tonsillar herniation.^7,8

Cough-associated headache (CAH) is the most distinctive symptom of CMI, characterized by headache exacerbated by coughing or Valsalva-like maneuvers and produced by cerebrospinal fluid (CSF) flow obstruction at the foramen magnum by tonsillar herniation.^8–12 CAH is believed to be indicative of clinically significant disease and predictive of a favorable outcome after decompression surgery. ¹³ In a recent study, it has been shown that ETH is the only anatomical measurement that differs between patients with and without CAH. ¹⁴ However, it was shown that the ETH measurement only weakly predicted the presence of CAH, with an area under the curve (AUC) of 0.66 on the receiver operating characteristic (ROC) curve. ¹⁴ Given the parasagittal location of the cerebellar tonsils, it is possible that the measurement of ETH performed in the midsagittal plane may not fully assess the neural crowding in CMI and thereby is weakly correlated with CAH. ¹⁴ Based on this observation, we hypothesized that parasagittal measurement of ETH in CMI might better depict neural crowding and resistance to CSF motion in the spinal canal and improve the prediction of symptoms such as CAH.

Asymmetry of ETH in CMI has previously been well recognized, and some studies have correlated them with symptomatology or the location of syringomyelia. ¹⁵ However, these studies measured individual tonsils on coronal images, which we found difficult to adopt due to the variability of the foramen magnum landmarks on magnetic resonance imaging (MRI) scans. Given the availability of high-resolution thin-section 3D MRI images, we postulated that each cerebellar tonsil could also be accurately measured in the parasagittal plane using anterior and posterior lips of the foramen magnum as the landmarks. Based on these observations, our purpose was to measure ETH of each cerebellar tonsil in the parasagittal plane and compare their specificity and sensitivity to measurements made in the midsagittal plane in predicting CAH in CMI patients.

Methods

Patient population

The Institutional Review Board approved this retrospective HIPAA-compliant study with a waiver for informed consent. We reviewed our institution’s imaging and clinical database to search for all consecutive CMI patients seen by the neurosurgery department between 2012 and 2020. The inclusion criteria were: (a) cerebellar tonsillar herniation ≥5 mm measured in the midsagittal plane (currently accepted definition of CMI), and (b) availability of high-resolution T1-weighted 3D MPRAGE images. The exclusion criteria were: (a) prior suboccipital decompression surgery, (b) bony anomaly at the craniocervical junction such as atlanto-occipital assimilation, which can obscure landmarks for measurements, (c) mass or mass-like lesion in the posterior fossa causing tonsillar herniation, (d) papilledema, and (e) clinical symptom of intracranial hypotension (headache that is worse when standing up and better when lying down). Using these criteria, a total of 85 patients were identified and included in the study.

Clinical data

Online medical records of all patients were reviewed to determine if, at the time of neurosurgical consultation, the presence or absence of CAH was recorded by the neurosurgeon. CAH was defined as a headache associated with Valsalva-like activities such as coughing, sneezing, and laughing. ¹⁶ Other CMI clinical features were also recorded.

MRI

All the MRI scans were performed on a 1.5 Tesla or 3 Tesla GE Signa HDx scanner (GE Healthcare, Milwaukee, WI) or a 1.5 Tesla Magnetom Espree scanner (Siemens, Erlangen, Germany). All MRI scans were performed with a standard protocol, including sagittal and axial spin-echo T1-weighted and axial T2-weighted, FLAIR, GRE, and diffusion images. The protocol also included axial T1-weighted spin-echo and sagittal T1-weighted post-contrast 3D MPRAGE images. The slice thickness for MPRAGE images was 1 mm with zero overlaps, and they were routinely reconstructed in axial and coronal planes. Other parameters for MPRAGE sequence were TR 6.5–7.7 ms, TE 3.3 ms, flip angle 10–12°, and matrix 240 × 240.

MR measurement

Two radiologists with 5 and 25 years of experience and blinded to the clinical history used 3D MPRAGE images to perform the following measurements independently:

1. Midsagittal measurement of the extent of tonsillar herniation (MS_ETH): The perpendicular distance measured between the McRae line (the line connecting basion and opisthion) and the tip of the cerebellar tonsil was defined as MS_ETH.^1,3,6,8,17 The midsagittal plane was defined as one in which the cerebral aqueduct and fourth ventricle were best visualized (Figure 1).

2. Parasagittal measurement of the extent of tonsillar herniation (PS_ETH): The perpendicular distance measured between a line connecting the anterior and posterior lips of the foramen magnum and tip of each tonsil was defined as PS_ETH. The tip of each cerebellar tonsil was identified using a reference line of the coronal reformats (Figure 2). For this purpose, the most anterior coronal section that contains the tonsils was selected.

Figure 1.

The white line shows the measurement of the extent of tonsillar herniation (ETH) in the midsagittal plane (MS_ETH). The black line is the McRae line connecting basion and opisthion. Note that both the aqueduct and fourth ventricle are visualized for MS_ETH measurement.

Figure 2.

(a)–(d) The ETH measured in the right and left parasagittal planes and their corresponding coronal planes are displayed. Left: Measurements of ETH on the parasagittal planes are shown. The black line connects the anterior and posterior lips of the foramen magnum on the left (a) and right (c) parasagittal plane. The white line indicated tonsillar position measurement. Right: Coronal plane with the vertical white line indicating the location of the left (b) and right (d) cerebellar tonsillar tip is shown.

Note that the line connecting the anterior and posterior lips of the foramen magnum is similar to but not precisely in the same position as the McRae line, which is defined at the midsagittal plane.

Before performing the measurements, the method described above was reviewed by the two radiologists using a few random examples of CMI patients. Measurements made during this practice session were not included in the analyses.

Determination of the presence or absence of syringomyelia was performed in all patients using cervical and thoracic spine MRI scans. Syringomyelia was defined as a cystic cavity in the cervical or thoracic spine measuring >2 mm in width.

Data analysis

Demographic data and all measurements were recorded. The intra-class correlation was used to test the interrater agreement. For each parameter, the average measurement between the two readers was used for analysis and comparisons between patients. The majority of the PS_ETH measurements were asymmetric (>1 mm difference) and were defined as Smaller_PS_ETH and Larger_PS_ETH. In a very small number of cases where both PS_ETH measurements were identical, one was assigned as Smaller_PS_ETH and the other as Larger_PS_ETH for statistical analysis.

Pearson’s r correlation coefficient was used to determine the level of interrater agreement. The Mann–Whitney U-test was used to compare MS_ETH and PS_ETH measurements between CMI patients with and without CAH. ROC curve analysis was performed to find out the predictive accuracy of each measurement. Comparison of ROC curves was made using Delong et al. methodology. ¹⁸ These analyses were done using IBM SPSS Statistics for Windows v26 (IBM Corp., Armonk, NY) and MedCalc software v9.6 (MedCalc Software bvba, Ostend, Belgium). A p-value of <0.05 was considered significant.

Results

The study population consisted of 15 men and 70 women (N = 85 patients) with a mean age of 40.6 years (standard deviation = 14.4 years). Headache of any type was present in 73/85 (85.9%) and CAH in 46/85 (54.1%) patients. Other clinical symptoms reported by patients were paresthesia 40/85 (47.1%), neck pain 35/85 (41.2%), dizziness 29/85 (34.1%), scoliosis 18/85 (21.2%), nausea and vomiting 15/85 (17.6%), ataxia 14/85 (16.5%), tinnitus 12/85 (14.1%), dysphagia 8/85 (9.4%), weakness 7/85 (8.2%), and nystagmus 3/85 (3.5%).

Tonsillar herniation measurements and CAH

The ETH was larger on the right in 53 (62.4%), larger on the left in 29 (34.1%), and equal in 3 (3.5%) cases. Compared to the Smaller_PS_ETH, MS_ETH was larger in 57/85 (67.1%), smaller in 19/85 (22.4%), and equal in 9/85 (10.6%) of cases. Compared to the Larger_PS_ETH, MS_ETH was smaller in 64/85 (75.3%), larger in 7/85 (8.2%), and equal in 14/85 (16.5%) of cases. There was excellent interrater agreement in measuring ETH between the two readers: MS_ETH (0.97), Smaller_PS_ETH (0.96), and Larger_PS_ETH (0.98). Table 1 displays the difference of measurements between CMI patients with and without CAH. All three ETH measurements were significantly different between patients with and without CAH.

Table 1.

Midsagittal and parasagittal measurements of tonsillar herniation in CMI patients with and without CAH.

	CAH present (N = 46)	CAH absent (N = 39)	p-Value
MS_ETH	10.0 (4.4)	8.0 (3.1)	0.02
Smaller_PS_ETH	9.3 (3.8)	6.2 (2.5)	<0.001
Larger_PS_ETH	12.0 (4.6)	9.4 (3.4)	0.005

All values are shown as means. p-Values were assessed with the Mann–Whitney U-test.

CMI: Chiari I malformation; CAH: cough-associated headache; MS_ETH: midsagittal measurement of the extent of tonsillar herniation; PS_ETH: parasagittal measurement of the extent of tonsillar herniation.

Table 2 shows the AUC for each ETH measurement in predicting CAH. The AUC for MS_ETH (midsagittal measurement) was 0.65. The AUC for Smaller_PS_ETH (smaller parasagittal ETH) was 0.78, which is a significant improvement over MS_ETH (p = 0.001). The AUC for Larger_PS_ETH (larger parasagittal ETH) was 0.68, which was not significantly different from MS_ETH. The ROC curves for MS_ETH and Smaller_PS_ETH are shown in Figure 3.

Table 2.

The AUC for different measurements of the tonsillar herniation in predicting CAH in CMI

	AUC (95% CI)	P Value
MS_ETH	0.65 (0.54–0.75)
Smaller_PS_ETH	0.78 (0.67–0.86)	0.001
Larger_PS_ETH	0.68 (0.57–0.77)	0.378

AUC: area under the curve; CI: confidence interval.

Figure 3.

Receiver operating characteristic curve for MS_ETH and Smaller_PS_ETH.

The sensitivity and specificity of four different cutoff points (6, 9, 12, and 15 mm) of MS_ETH, Smaller_PS_ETH, and Larger_PS_ETH for the presence of CAH were calculated (Table 3). The specificity of ETHs improved for all measurements as the magnitude of ETHs increased until at 15 mm, when both MS_ETH and Smaller_PS_ETH showed 100% specificity. However, for ETH measurements of 6, 9, and 12 mm, Smaller_PS_ETH showed much better specificity in detecting CAH in a CMI patient than MS_ETH and Larger_PS_ETH. As expected, by increasing the cutoff point, the sensitivity decreased, and the specificity increased.

Table 3.

Sensitivity and specificity of four different cutoff points (6, 9, 12, and 15 mm) for tonsillar herniation in predicting CAH in CMI.

	Sensitivity				Specificity
Measurements	6 mm	9 mm	12 mm	15 mm	6 mm	9 mm	12 mm	15 mm
MS_ETH	87%	52%	33%	11%	28%	67%	85%	100%
Smaller_PS_ETH	90%	48%	20%	7%	46%	87%	97%	100%
Larger_PS_ETH	98%	72%	48%	22%	5%	46%	74%	87%

Syringomyelia and ETH measurements

Syringomyelia was present in 12/85 (14.1%) patients, and of those 12 patients, six had CAH. MS_ETH, Smaller_PS_ETH, and Larger_PS_ETH were not significantly different between patients with and without syringomyelia. In the subgroup of 12/85 CMI patients with syringomyelia, AUC for Smaller_PS_ETH (0.74) was significantly greater than AUC for MS_ETH (0.56) with a p-value of 0.03. Similarly, in the subgroup of 73/85 CMI patients without syringomyelia, AUC for Smaller_PS_ETH (0.80) was greater than AUC for MS_ETH (0.71) with a p-value of 0.04. There was no significant difference in AUC between Larger_PS_ETH and MS_ETH in either subgroup.

Discussion

CMI patients may suffer from various types of headaches such as migraine or tension varieties similar to the general population. However, transient activity-associated (coughing, sneezing, laughing, Valsalva, or Valsalva-like maneuver) headaches are considered to be typically associated with this condition and can be collectively known as CAH. CAH is the most common presentation of CMI. Presence of CAH in a CMI patient is generally an indication for decompressive surgery and a predictor of a better outcome. ¹⁹ Therefore, CAH is the potential indicator of clinically significant disease in a CMI patient who needs to be further evaluated. ¹⁴

CAH has been defined by the International Classification of Headache Disorders to have a specific duration, location (suboccipital), and relation to precipitating factors. ¹⁶ However, in the clinical setting, a probable CAH may not perfectly fulfill such strictly defined criteria for the classical pattern. Consequently, in many cases, clinicians may wish for additional confirmation from imaging to help them in their decision making. ¹⁴

It is an accepted fact that CAH in CMI is produced by obstruction to the free flow of CSF between the head and spine due to tonsillar herniation. ⁹ Therefore, unless the tonsillar herniation achieves sufficient neural crowding in the upper cervical spine canal, a CMI patient may not show CAH. It has been shown that tonsillar measurements made in the midsagittal plane do not fully estimate the neural crowing and are inadequate indicators of the presence of CAH in a CMI patient.^5,13,19–21 Prior studies comparing the midsagittal ETH between those with and without CAH showed conflicting results. While Alperine et al. ²⁰ and Sansur et al. ²¹ failed to find any differences, Huang et al. ¹⁴ showed significant differences in the midsagittal ETH between the two groups. However, they conclude that these differences can be used in clinical decision making if and only if the midsagittal ETH is much larger than 5 mm. Our results reinforced this notion.

In addition, this study has shown that the specificity of prediction of the presence of CAH and consequently clinically significant disease in CMI patients can be improved over the currently used midsagittal measurement technique without compromise on sensitivity by individually measuring each cerebellar tonsil from parasagittal images and using the smaller of the two measurements. This improvement was shown to be irrespective of the presence or absence of the syringomyelia. Therefore, if a radiologist is counseling a clinician about the neurosurgical referral, the limitation of midsagittal ETH can be significantly decreased by using the smaller parasagittal ETH.

The fact that the improvement in specificity was seen only with the smaller rather than the larger herniation of the two tonsils can be explained by the pathophysiology of CAH. We believe the reason for that is unless both of the two herniated tonsils are large enough, sufficient neural crowding in the upper cervical spinal canal and CSF flow obstruction may not be achieved to produce CAH. In other words, the magnitude of resistance to CSF motion is determined by the smallest of the two herniated tonsils because it determines how much crowding occurs. This matches the previous findings of unsteady resistance to CSF motion in the spinal canal using computational fluid dynamics,^11,12 in which ETH at the midsagittal plane did not always indicate high resistance to CSF motion depending on the three-dimensional morphology of the cerebellar tonsils.

Asymmetry of tonsillar herniation in CMI patients has been recognized previously. In 2002, Tubbs et al. ¹⁵ were the first to report the significance of tonsillar asymmetry in CMI using coronal images. In their study population, the tonsillar herniation was observed to be asymmetric in 77.3% of patients (larger ETH on the right in 64% and on the left in 13.3%). They observed the larger right tonsillar herniation to be more likely to be associated with syringomyelia. In addition, unilateral symptoms/physical findings were also reported to be more likely on the side of larger tonsillar herniation. In 2014, Deng et al. ²² studied 104 CMI patients with syringomyelia and used coronal images to observe tonsillar asymmetry in 91.3% of patients (larger ETH on the right in 47.1% and on the left in 44.2%). They also observed a significant correlation between the larger tonsillar herniation and the side (and deviation ratio) of syringomyelia. Furthermore, they found a significant relationship between the side of the larger tonsil and the convexity of scoliosis. Although we used sagittal images for mid- and parasagittal measurements of ETH, we also found tonsillar asymmetry in the majority (96.5%) of our CMI patients, with right tonsillar herniation more likely to be larger than the left (larger ETH on the right in 62.4% and on the left in 34.1%). However, unlike the studies by Tubs and Deng, we used a different clinical finding (presence of CAH) and found the measurement of smaller rather than the larger ETH to be significantly related, suggesting that it is essential to measure both cerebellar tonsils individually.

In 2016, Tubbs et al. ²³ measured ETH on coronal images in 50 pediatric patients using the inferior margin of the foramen magnum as a reference line. They observed that measurements of both individual tonsils were generally smaller than midsagittal tonsillar measurements and concluded that midsagittal measurements overestimate ETH and can thus falsely diagnose some patients with tonsillar ectopia (ETH <5 mm) as CMI. We found the accurate measurement of the extent of tonsillar herniation on the coronal plane to be challenging due to the presence of undulations and foramina on the skull base lateral to the foramen magnum. These variations made it nearly impossible to define a replicable reference line on coronal images in the same reference plane as the McRae line used on sagittal images. To avoid this issue, we only used coronal images to find the tonsillar tips and identify the ideal parasagittal plane. By connecting the anterior and the posterior edge of the foramen magnum at that parasagittal plane, we made a reference line for measuring tonsillar herniation depth. This definition for the reference line was straightforward, replicable, and reliable, and it was reflected in our high correlation coefficient for interrater agreement.

Our findings highlight some important issues in diagnosing clinically significant CMI, using the current definition of ETH measurement of ≥5 mm in the midsagittal plane. First, as our results and those of others (mentioned above) show, midsagittal measurements may not be enough to capture anatomic severity (crowding of neural structures) in CMI.^15,22 Second, routine sagittal images of the brain or cervical spine with section thickness of 4–5 mm have the potential for over- or underestimating ETH, depending on whether the larger or smaller tonsillar herniation is included in the measurement due to the partial volume averaging effect. Therefore, we suggest routine use of high-resolution thin-section 3D images in the diagnostic workup of CMI. The high-resolution 3D images can accurately define the midsagittal plane without partial volume averaging of surrounding tissues. They can be further used to reformat and locate individual tonsillar tips for parasagittal measurements. Moreover, better defining the midsagittal plane can improve interrater reliability, as seen in our study compared to the previous reports.^24,25

There are some limitations of our study. First, due to the retrospective nature of our study, we had to depend on the history of CAH in the electronic records, which were recorded by different neurosurgeons. However, since our neurosurgeons are experienced in taking care of CMI patients, they are aware of the importance of CAH history. Therefore, any misclassification of CAH, if it happened, was likely minimal. Second, it is likely that our parasagittal measurements include some differences in inclination of the foramen magnum with parasagittal location slightly higher than the midsagittal location. However, we believe that such a difference did not affect our results, since we did not directly compare the measurements to each other but compared them to a clinical symptom. Third, this study only focuses on the utility of the PS_ETH in evaluating the presence of CAH in CMI. Further studies are needed to investigate the correlation of other morphometric parameters of the posterior fossa and the possible effacement of the CSF spaces to PS_ETH and clinical findings.

Conclusions

In conclusion, we described the feasibility, technique, and reproducibility of parasagittal measurement of tonsillar herniation in CMI. Furthermore, we were able to show that the diagnostic utility of detecting clinically significant CMI by ETH measurements can be improved by using smaller of the two parasagittal measurements of tonsillar herniation, and this improvement is irrespective of the presence or absence of syringomyelia.