Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2025 Oct 1.
Published in final edited form as: Otol Neurotol. 2024 Sep 5;45(9):e647–e654. doi: 10.1097/MAO.0000000000004310

Facial nerve tractography using diffusion MRI: A comparison of acquisition b-values and single- and multi-fiber tracking strategies

Lorenz Epprecht 1,2, Leo Zekelman 3,4, Katherine L Reinshagen 5, Guoqiang Xie 6, Isaiah Norton 7, Ron Kikinis 7, Nikos Makris 8,9, Marco Piccirelli 10, Alexander Huber 1,2, Daniel J Lee 11, Fan Zhang 7, Lauren J O’Donnell 7
PMCID: PMC11458140  NIHMSID: NIHMS2009137  PMID: 39234825

Abstract

Hypothesis:

This study investigates the impact of different diffusion magnetic imaging (dMRI) acquisition settings and mathematical fiber models on tractography performance for depicting cranial nerve (CN) VII in healthy young adults.

Background:

The aim of this study is to optimize visualization of CN VII for preoperative assessment in surgeries near the nerve in the cerebellopontine angle, reducing surgery-associated complications. The study analyzes 100 CN VII in dMRI images from the Human Connectome Project, using three separate sets with different b-values (b=1000 s/mm2, b=2000 s/mm2, b=3000 s/mm2) and four different tractography methods, resulting in 1200 tractographies analyzed.

Results:

The results show that multi-fiber and free water (FW) compartment models produce significantly more streamlines than single-fiber tractography. The addition of a FW compartment significantly increases the mean streamline fractional anisotropy (FA). Expert quality ratings showed that the highest rated tractography was the 1 tensor (1T) method without FW at b-values of 1000.

Conclusions:

In this young and healthy cohort, best tractography results are obtained by using a 1T model without a FW compartment in b=1000 diffusion MR images. The FW compartment increased the contrast between streamlines and CSF (higher mean streamline FA). This finding may help ongoing research to improve CN VII tractography results in tumor cases where the nerve is often stretched and thinned by the tumor.

Keywords: Facial nerve, vestibular schwannoma, diffusion MRI, diffusion tensor imaging, single-tensor tractography, multi-tensor tractography

1. Introduction

Mapping of the the facial nerve (CN VII) is clinically important in the management of neurosurgical and neuro-otological patients, particularly in those with tumors affecting the cerebellopontine angle (CPA) and the internal acoustic canal (IAC) such as vestibular schwannomas (VS). Due to deviation of the nerve by the tumor, surgical removal necessitates careful handling in order to avoid facial nerve palsy1,2. A preoperative evaluation of the CN VII aims at the reduction of such surgery-associated complications.

Diffusion magnetic resonance imaging (dMRI) has emerged as an advanced imaging technique to visualize CN VII. It uniquely enables in-vivo mapping of the 3D trajectory of the CNs in a non-invasive way by a computational process called tractography. Many studies have shown the utility of dMRI tractography for CN VII mapping (Table 1). The remaining challenge in dMRI and tractography for CN VII delineation, however, is that the hitherto employed imaging and tractography techniques are heterogeneous and have varying success rates (Table 1).

Table 1:

Previous studies performing tractography for the facial nerve

Study Year b-value (s/mm2) Gradients Type of tractography used Tractography success rate (%) Intraoperative accordance
Blanch Pujol et. al. 4 2022 800 32 DTI - deterministic 100 100
Castellaro et al. 5 2020 300–2000 12–32 (Advanced) - deterministic / probabilistic 60–80 n/r
Szmuda et al. 6 2020 1500 64 DTI - deterministic n/r 82
Zolal et al. 7 2017 800 20 DTI - deterministic / probabilistic 100 81
Zhang et al 8 2017 1000 64 DTI - deterministic 100 97
Huan et al 9 2017 1000 32 DTI - deterministic 95 95
Hilly et al 10 2016 nr 32 DTI - deterministic 78 71
Borkar et al 11 2016 800 15 DTI - deterministic 95 89
Yoshino et al 12 2016 384–5000 101 DSI - deterministic 90 n/r
Yoshino et al 13 2015 1000 30 DTI - deterministic 82 17
Wei et al 14 2015 nr 30 DTI - deterministic 91 100
Choi et al 15 2014 1000 32 DTI - deterministic 100 100
Zhang et al 16 2013 1000 n/r DTI - deterministic 88 100
Roundy et al 17 2012 1000 32 DTI - deterministic 100 100
Gerganov et al 18 2011 1000 12 DTI - deterministic 100 91
Chen et al 19 2011 1000 25 DTI - deterministic 100 n/r
Taoka et al 20 2006 1000 n/r DTI - deterministic 88 71
Open in a new tab

Many studies have shown the utility of dMRI tractography for CN VII mapping. Most previous studies report a b-value around 1000 s/mm2 since this is the clinically most available b-value. The reviewed papers defined success rate as the rate of plausible depiction of the facial nerve by tractography, irrespective of anatomical accuracy. Anatomical accuracy, i.e. intraoperative accordance of tractography and actual facial nerve, was not assessed in all studies. Abbreviations: not reported (n/r), diffusion (single) tensor imaging (DTI), diffusion spectrum imaging (DSI).

The choice of parameters for the dMRI acquisition and tractography are crucial to achieve a robust depiction of the CN VII and therefore influences the correct preoperative understanding of the anatomy. The strength and duration of the magnetic diffusion gradient used for dMRI (i.e. b-value), has a significant influence on the quality of brain white matter tractography21. The higher the b-value, the higher the diffusion contrast22. Recent studies assessing the CN VII (Table 1) typically used the clinically most widely available b-value of around 1000 s/mm2.

Moreover, the choice of the tractography algorithm to interpret dMRI signals is important to correctly trace nerve fibers. Different tractography methods are known to yield different results in the brain28 and for the cranial nerves24.

The main contribution of this study is to investigate the performance of different dMRI acquisitions (b-values) and different tractography algorithms (mathematical fiber models) on dMRI tractography performance for depicting CN VII.

2. Material and Methods

2.1. Evaluation Dataset

100 CN VII of 50 subjects (left and right nerves for each subject) were analyzed. Tractography was performed using dMRI data from the young adult Human Connectome Project database25. All subjects were between 22 and 35 years old. For the purpose of this study, only de-identified data from the HCP were used and therefore no IRB approval was necessary. The HCP data acquisition and processing are described in Glasser et al. 201326. dMRI were acquired in a 3 Tesla MRI scanner using a multi-band (3) spin-echo EPI sequence (repetition time (TR) 5520 ms, echo time (TE) 89.5 ms), flip angle 78 degrees, field of view (FOV) 210×210×210 mm3, and a resulting isotropic voxel size of 1.25 mm. 3 shells of b=1000, 2000, and 3000 s/mm2 were acquired with 90 diffusion weighted images and 6 interspersed baseline image images (b=5 s/mm2) for each shell. We used preprocessed HCP dMRI data as described in27 with motion correction and distortion correction as in28–30. In order to assess performance of CN VII tracking at different b-values, we separated the original multi-shell data into single-shells with b-values of 1000 s/mm2, 2000 s/mm2, and 3000 s/mm2. We then performed dMRI tractography, analysis and visualization in 3D Slicer (www.slicer.org) via the SlicerDMRI project (http://dmri.slicer.org)31–33). We also used T2-weighted images in order to facilitate region of interest (ROI) selection. T2-weighted data were acquired with TE = 565 ms, TR = 3200 ms, and voxel size = 0.7 × 0.7 × 0.7 mm3.

2.2. CN VII Tractography

For CN VII tracking, we used the unscented Kalman filter (UKF)1 31,32. We chose to study UKF tractography34, 35 because it has been shown to be successful for tracking multiple cranial nerve24,36 and has demonstrated consistent results in various independently acquired dMRI data as well as in test-retest datasets37. The UKF method is more sensitive than standard single-tensor tractography, especially in the presence of crossing fibers and peritumoral edema38. The UKF package provides both single-tensor (1T) and two-tensor (2T) fiber tracking methods, as well as the option to include a free water (FW) compartment. This enables a comparison between a single-fiber model and a higher-order model, with or without FW, using the same underlying mathematical framework. We were interested in the performance of the FW compartment in particular, because of its theoretical ability to reduce partial volume effects in the context of CN VII passing through isotropic CSF. We compared the performance of these four tractography approaches, i.e. methods, in each single-shell dataset (b-values of 1000 s/mm2, 2000 s/mm2, and 3000 s/mm2). The four tractography methods that we investigated were the following: Single-tensor (1T) UKF tractography, 1T UKF tractography with free water (FW) estimation, two-tensor (2T) UKF tractography, and 2T UKF tractography with FW estimation. We performed UKF tractography using parameters optimized for cranial nerve tracking in HCP data (supplementary Table 1)24.

2.2.1. Seeding CN VII tractography in all datasets

1T, 1T FW, 2T and 2T FW UKF tractography were seeded in each of the three single-shell dMRI datasets (b-values = 1000 s/mm2, 2000 s/mm2, and 3000 s/mm2). An expert (LE, consultant otolaryngologist with 10+ years of clinical experience, including three years training in neurosurgery) manually created a spherical or oval ROI in the region of the cerebellopontine angle (CPA) where the CNs VII run from their root exit zones (REZ) to the internal acoustic canal (IAC) using the Editor module in 3D Slicer. One separate seeding mask for each side, left and right CN VII, was drawn. Tractography was seeded from all voxels within these masks with two seeds per voxel. Consequently, a total of 12 different tractography datasets were generated per subject (1T-1000, 1T-2000, 1T-3000, 1T-FW-1000, 1T-FW-2000, 1T-FW-3000, 2T-1000, 2T-2000, 2T-3000, 2T-FW-1000, 2T-FW-2000, 2T-FW-3000). These datasets were further processed using defined ROIs for selecting CN VII (see following section).

2.2.2. Selection of the CN VII by ROIs

Two ROIs for each side (left and right CN VII) were chosen for streamline selection in the region of the CN VII REZ and the IAC. The ROIs were drawn by the same expert mentioned above (LE) using the Editor module of 3D Slicer (Figure 1). The ROIs for the REZ were drawn with a spherical volume of 20 voxels on the baseline (b=5s/mm2) dMRI images. The ROIs at the level of the IAC were drawn on the T2-weighted images, since the IAC was not clearly identifiable on the baseline (b=5 s/mm2) dMRI images. To avoid false negative results (no streamlines) because of potential small geometric distortions between the T2-weighted image and the dMRI image, the IAC ROI was slightly enlarged by a dilate effect of four adjacent voxels in the Editor module of 3D Slicer. Streamlines were selected by a logical AND operator of the two nerve specific ROIs (REZ and IAC) using the Tractography ROI selection module in 3D Slicer.

Figure 1: Examples of regions of interest for streamline selection.

Figure 1:

Open in a new tab

In (A), an axial section of a baseline image (b=5 s/mm2) at the level of the pontomedullary junction shows the result of tractography of the facial nerve (cranial nerve (CN) VII) in blue on the left and red on the right patient side. In (B), also an axial section, examples of regions of interest (ROIs) are shown for seeding (blue and red circles) and selecting the streamlines (REZ and IAC) representing the facial nerves (left (blue) and right (red)). (C) and (D) display sagittal and coronal sections, respectively, of a baseline image on which the ROIs used for selecting the nerves are marked.

2.3. Outcome measures

2.3.1. Quantitative descriptors of tractography results

We recorded a) the number of streamlines along the anatomical course of CN VII, b) the mean FA of the streamlines and c) success rate. The number of streamlines is a surrogate marker for the robustness of the connectivity visualized by tractography 39,40 and the streamline FA relates to the contrast between CN VII depiction (anisotropic diffusion) and CSF (almost no diffusion restriction)41,. Success rate was defined as the percentage of cases with streamlines present.

2.3.2. Expert rating and Quality Rankings of CN VII tractography

We also assessed the quality of the CN VII tractography (bundle of tractography streamlines) by expert ratings. For the assessment of quality, we determined three quality criteria: a) All streamlines are oriented in the same direction (run parallel to CN VII), b) there is only one streamline bundle representing CN VII tractography visible (as opposed to multiple separate streamline bundles), and c) there are no streamline loops present. Loops were defined as visually assessed streamlines taking off from the main streamline bundle, forming a loop, and coming back to the main streamline bundle. All CN VII tractographies were displayed using the analysis software provided by the whitematteranalysis package2. Two raters LE and KR (consultant radiologist (9 years with specialization in head and neck radiology, including 2 years training in neuroradiology)) independently scored all CN VII tractographies (100 CN VII x 3 dMRI shells x 4 UKF tractography methods = 1200 CN VII tractographies). Figure 2 gives a visualization of the CN VII of one example subject.

Figure 2:

Figure 2:

Open in a new tab

Example of tractography display of the right facial nerve (red streamlines) with a co-registered highly T2-weighted anatomical MRI as background across tractography methods and b-values for expert quality rating. The CN VII tractographies have overall good quality, but e.g. the streamlines traced by single-tensor tractography with free water compartment (1T FW) at a b-value of 1000 s/mm2 are not all parallel to each other (white arrow). The same tractography methods (1T FW) at a b-value of b=2000 show two streamline bundles representing CN VII (red and yellow streamlines white numbers “1” and “2”) and unwanted streamline loops (yellow streamlines, white arrow head).

Scoring was performed as follows: Each of the three quality criteria could receive a score of 0 or 1. 0 points were scored if a criterion did not apply, and 1 point was scored if the criterion statement applied to the tractography assessed. By doing this, each CN VII tractography obtained a sum score from the three quality criteria between 0 and 3 from each reviewer and a combined sum score between 0 and 6 for the ratings from the two reviewers combined. A CN VII tractography with 0 points would have the worst possible quality, whereas a CN VII tractography with 6 points would have the best possible quality. For a better representation of the CN VII tractography quality, we categorized the CN VII tractography according to the achieved combined sum scores into three quality ranking categories: “excellent quality” (6 points), “good quality” (3–5 points), “poor quality” (0–2 points). Cases in which no streamlines could be traced at all were categorized as “poor quality.”

2.4. Statistical analysis

Statistical analysis was performed in R (version 4.2.3, R Foundation for Statistical Computing, Vienna, Austria)42.

2.4.1. Quantitative descriptors of tractography results

One-way repeated measures ANOVAs were performed to assess the effect of tractography methods on FA and number of streamlines. When significant differences in means were detected (alpha = 0.05), post-hoc paired t-tests were conducted to further examine pairwise differences. P-values from pairwise comparisons were corrected using the Benjamini & Hochberg false discovery rate (FDR)43. Four CN VII tractographies from four participants were excluded from FA analyses for lack of data (i.e., FA cannot be calculated when number of streamlines is 0). Statistical analyses and plots were conducted and generated using R statistical software, using the “ggplot2”, “tidyverse”, “rstatix” and “ggpubr” packages44–48.

2.4.2. Expert inter-rater agreement of perceived quality of CN VII tractography

For each of the 1200 CN VII tractographies, it was determined whether the two expert raters agreed on the predefined three quality criteria described above. We assessed if the expert raters gave the same points (0 or 1) for each criterion in each of the tractographies. We then calculated the percentage of agreement of the two raters across all tractographies and criteria (overall agreement) and the inter-rater reliability by Cohen’s Kappa49.

2.4.3. Analysis of Perceived Quality of CN VII Tractography across Tractography Methods

To detect if perceived CN VII tractography quality differed across the 12 investigated tractography methods, the Freidman test was used50. This omnibus test was selected because it is a nonparametric statistical test that can be used to detect differences in treatments across multiple attempts and can accommodate categorical ordinal data. Quality ratings from 100 CN VII collected from 50 participants were interpreted as a categorical variable with three levels (“excellent” > “good” > “poor”). This data was grouped according to the 12 tractography methods investigated. To account for the hemispheric variability in the diffusion data collected from each participant, participant hemisphere (100 hemispheres from 50 participants) was used as a blocking variable. The effect of the tractography method on CN tractography quality was computed using Kendall’s W51. Post-hoc pairwise comparisons were made across all tractography methods using the Eisinga, Heskes, Pelzer & Te Grotenhuis all-pairs test with exact p-values for a two-way balanced complete block design52. P-values were corrected using Benjamini & Hochberg’s false discovery rate (FDR) (n = 66 comparisons)43. Overall, this analysis followed the methodology outlined by Mangiafico53. Statistical assessments were conducted using R statistical software, using the “stats”, “PMCMRplus”, and “rcompanion” packages54,55.

3. Results

3.1. Quantitative Results

Figure 3 shows that the number of streamlines produced is significantly affected by the tractography methods, and is not strongly influenced by the choice of b-value of the input dataset. The number of streamlines is considered a proxy for the robustness of fiber tracking 39,40. In general, the number of streamlines produced increased with higher-order fiber models (more than one tensor and/or free water (FW) compartment). Mean streamline FA was significantly affected by the choice of input b-value and the tractography method (supp. Figure 1). Overall, it could be demonstrated that for all input data and tractography methods, anisotropic diffusion was beneficial to enable the tracking of CN VII. The measured free water fraction was highest at b=1000 (Table 2). Finally, in concordance with the quantitative results in Figures 3 and supplementary Figure 1, there was success of CN VII fiber tracking in 97–99% of subjects, across all compared input b-values and tractography methods. For specific number of streamlines and FA as well as success rates across all tractography methods and b-values see supplementary Tables 2–4.

Figure 3: Number of streamlines across different tractography methods and b-values.

Figure 3:

Open in a new tab

First row of violin plots: b-value has little influence on number of streamlines. Second row of violin plots: Number of streamlines generally increases with the addition of a free water (FW) compartment and/or more tensors. The effect is more pronounced at lower b-values and not always significant at b=3000. The addition of a free water (FW) compartment increases the number of streamlines in all cases (e.g. T1 vs. T1 FW, or T2 vs. T2 FW). Violin plots show the density distribution of the number of streamlines. Red middle dot represents the mean number of streamlines and extending error bars depict +/− one standard deviation of the mean. Abbreviations: single tensor (1T), two tensor (2T), free water (FW), not significant (ns), (*) p < 0.05; (**) p < 0.01; (***) p < 0.001, p < 0.0001 (****).

Table 2:

Mean free water (FW) fraction in 1T and 2T tractography methods with FW compartment

b 1000 b 2000 b 3000
1T FW 0.70 ± 0.10 0.27 ± 0.08 0.23 ± 0.08
2T FW 0.46 ± 0.10 0.25 ± 0.07 0.23 ± 0.07
Open in a new tab

Abbreviations: single-tensor (1T), two-tensor (2T), free water compartment (FW).

3.2. Expert Rating Results

There was a statistically significant difference in the perceived quality of tractography depending on which type of tractography was applied to the CN VII, χ2(11) = 494.33, p < 2.2e-16 (n = 100). The tractography method had a substantially large effect on perceived quality, W = 0.449, CI (0.396, 0.513). Overall, the 1T b=1000 method produced the highest percentage of “excellent” ratings (Figure 4, shown in green). The three best performing tractographies were 1T tractography methods at b-values of 1000, 2000, and 3000. These 1T tractographies produced “excellent” ratings for 78% (b=1000), 64% (b=2000), and 45% (b=3000) of the investigated CNs VII (Figure 4). FDR corrected p-value pairwise comparisons for the top-performing 1T tractography methods with b-values of 1000, 2000, and 3000 demonstrate that the perceived quality of tractography was significantly different between the 1T b=1000 and b=3000 methods, p = 0.003, and between the 1T b=2000 and 1T b=3000 methods, p = 0.0456. However, no significant difference in the tractography quality was found between 1T b=1000 and b=2000 tractography methods, p = 0.334.

Figure 4: Quality of CN VII Tractographies across different methods.

Figure 4:

Open in a new tab

Each column represents one set of tractography methods, single tensor (1T), single tensor with free water compartment (1T FW), two-tensor ( 2T), and two tensor-tensor with free water compartment (2T FW) as well as different b-values (b=1000 s/mm2, b=2000 s/mm2, b=3000 s/mm2). Each column contains a total of 100 CN VII tractographies (50 subjects x 2 facial nerve tractographies) which are split into the different quality categories (excellent / good / poor) across the column.

See Supplementary Table 5 for pairwise comparisons between all tractography methods investigated.

3.3. Expert inter-rater validation

Overall agreement across all tractographies and quality criteria between the two expert raters was 91%. Cohen’s kappa for inter-rater agreement was 0.803 (considered to be substantial agreement following the literature,54). Inter-rater agreement for the different quality criteria was as follows: a) All streamlines are oriented in the same direction (run parallel to CN VII (81%), b) There is only one streamline bundle representing CN VII visible (96%), and c) there are no streamline loops present (97%).

4. Discussion

CN VII paralysis is a surgical-complication that can have a long-lasting negative impact on an individual’s psychosocial health2. The goal of this study was to better understand the influence of different b-values and tractography methods on the overall quality of CN VII tractography in healthy young adults. To judge the tractography quality of CN VII, we used expert raters. The expert rating process was designed to highly rate true positive streamlines, while penalizing false positive errors. Overall, highest quality ratings were produced with a single-tensor model at low b-values (b=1000 and b=2000), suggesting that this method produced more positive streamlines and fewer false positive errors than other investigated methods.

Additionally, we found that higher-order tractography methods (more than one tensor or addition of a FW compartment), resulted in significant increases in the number of streamlines. However, higher-order tractography methods also produced lower quality ratings. This finding suggests the additional streamlines produced by the higher-order tractography methods are affected by false positive tracking, resulting in more erroneous streamlines. This is in line with findings in the cerebrum, where higher-order models are known to increase sensitivity at the expense of increased false positive fiber tracking56. Increasing the b-value did not impact the number of streamlines.

Furthermore, we inspected the effect of tractography methods and b-values on fractional anisotropy (FA). We found that higher-order tractography methods generally resulted in greater measures of FA, whereas higher b values produced lower mean streamline FA. As expected, the addition of a FW compartment to the tractography algorithm (1T FW, 2T FW) led to an increase of the FA, i.e. an apparent improvement of the contrast between CSF and CN VII. However, the addition of a FW compartment did not result in higher quality ratings.

Compared to previous studies, the success rate reported in this study, meaning the percentage of cases in which a fiber tract was detected, was not different (compare Tables 1, second last column “tractography success” of previous studies, and supplementary Table 4 “success rates’’ of this study). This means that higher-order tractography methods do not provide a significant advantage in this respect. The use of higher-order tractography methods did increase the number of streamlines, which is good in principle because it increases the reliability of tractography, but, as has been shown in other previous studies24, these tracking strategies also increased false positive streamlines (loops, multiple streamline bundles, etc.). It appears that tractography algorithms with more than one tensor provide an advantage especially in the area of branching or crossing fibers, as has been shown in other cranial nerves whose courses have also been traced in the brainstem, a region in the brain well known for its crossing and branching fibers 24, 57. In our experiment, we were only interested in the course of CN VII in its cisternal (i.e. CPA) segment, because this is the clinically relevant segment in tumor surgery in this anatomical region. There, CN VII runs more or less straight or with a slight curve when it is displaced by a tumor. We think, based on the results presented here, that therefore multi-tensor tractography algorithms have no advantage for this application. It would be interesting to see if higher-order tractography methods can show an improvement in terms of accuracy compared to the intraoperative course and anatomical extent of the nerves (ground truth) in patients with cerebellopontine angle tumors (i.e. VS), because the conventional algorithms show large differences in quality in this regard, as can be seen in Table 1, last column (“intraoperative accordance”).

4.1. Clinical considerations

Clinical aspects also emerge from this study. The study shows that regular clinical scans most likely are sufficient to visualize CN VII preoperatively in patients with VS. It is perhaps more important that the scan protocols are designed to minimize distortion artifacts, as was done in the HCP protocol of the diffusion-weighted images used here, than that higher-order tractography algorithms or higher b-values are used. Surgically, the preoperative assessment of whether the nerve is thinned by the tumor or not is also an important aspect. It would therefore be interesting to see whether the degree of thinning influences the parameters investigated here, such as FA or the free water compartment. This study provides normative values in healthy, non-thinned nerves, which could be compared with pathological situations.

4.2. Limitations

There are limitations to this study. The images used were taken under optimal conditions in healthy volunteers without tumors. If we had used a different testing dataset with alternative protocols, it is possible that results would have been different. A study with clinical data would show whether the results can be reproduced there as well. We used optimized tractography parameters from a previous study on the trigeminal nerve24. It is possible that the choice of these parameters, especially the very low FA values used with high b-values, influenced the final results. We used UKF tractography due to its good performance and because it allowed us to compare 1T and 2T tractography strategies with the same mathematical model. Future studies could incorporate additional tractography methods such as the clinically popular conventional streamline single-tensor tractography, constraint spherical deconvolution or tractography based on higher-order fiber models.

5. Conclusion

In summary, in the high-resolution images we used for this study, higher b-values and more than one tensor provide no advantage in the tractography of CN VII. Overall, UKF tractography performed best in the constellation of b-1000 and 1T without FW. The addition of a FW compartment, however, allowed for a better contrast between CN VII tractography and surrounding CSF, expressed by a higher streamline FA. In clinical applications, this finding opens up possibilities for future research to enhance CN VII tractography in the complex anatomical context of vestibular schwannomas.

Supplementary Material

Supplemental Digital Content, Fig 1
Supplemental Digital Content, Table 1
Supplemental Digital Content, Table 2
Supplemental Digital Content, Table 3
Supplemental Digital Content, Table 4
Supplemental Digital Content, Table 5

Funding

We acknowledge the following NIH grants: P41EB015902, P41EB015898, R01MH125860, R01MH119222, and R01NS125781. F.Z. acknowledges a BWH Radiology Research Pilot Grant Award. RK acknowledges R01 CA235589, NCI Task Order HHSN26110071

Abbreviations:

CN VII

cranial nerve VII (facial nerve)

CPA

cerebellopontine angle

CSF

cerebrospinal fluid

dMRI

diffusion magnetic resonance imaging

DTI

diffusion tensor imaging

DWI

diffusion weighted imaging

IAC

internal acoustic canal

REZ

root exit zone

ROI

region of interest

VS

vestibular schwannoma

Footnotes

Declarations of interest: none

References

  • [1].Samii M, Gerganov V, Samii A. Improved preservation of hearing and facial nerve function in vestibular schwannoma surgery via the retrosigmoid approach in a series of 200 patients. J Neurosurg 2006;105(4):527–535. [DOI] [PubMed] [Google Scholar]
  • [2].Cuenca-Martínez F, Zapardiel-Sánchez E, Carrasco-González E, La Touche R, Suso-Martí L. Assessing anxiety, depression and quality of life in patients with peripheral facial palsy: a systematic review. PeerJ 2020;8:e10449. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [3].Basser PJ, Pajevic S, Pierpaoli C, Duda J, Aldroubi A. In vivo fiber tractography using DT-MRI data. Magn Reson Med 2000;44(4):625–632. [DOI] [PubMed] [Google Scholar]
  • [4].Blanch Pujol G, Sanmillan JL, Sánchez-Fernandez JJ, Fernandez-Conejero I, Cifre Serra P, Torres A, et al. Anticipating facial nerve position using three-dimensional tractography during the preoperative assessment of cerebellopontine angle tumors. World Neurosurg 2022;168:e317–e327. [DOI] [PubMed] [Google Scholar]
  • [5].Castellaro M, Moretto M, Baro V, Brigadoi S, Zanoletti E, Anglani M, et al. Multishell Diffusion MRI-Based Tractography of the Facial Nerve in Vestibular Schwannoma. AJNR Am J Neuroradiol 2020;41(8):1480–1486. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [6].Szmuda T, Słoniewski P, Ali S, Pereira PMG, Pacholski M, Timemy F, et al. Reliability of diffusion tensor tractography of facial nerve in cerebello-pontine angle tumours. Neurol Neurochir Pol 2020;54(1):73–82. [DOI] [PubMed] [Google Scholar]
  • [7].Zolal A, Sobottka SB, Podlesek D, Linn J, Rieger B, Juratli TA, et al. Comparison of probabilistic and deterministic fiber tracking of cranial nerves. J Neurosurg 2017;127(3):613–621. [DOI] [PubMed] [Google Scholar]
  • [8].Zhang Y, Mao Z, Wei P, et al. Preoperative Prediction of Location and Shape of Facial Nerve in Patients with Large Vestibular Schwannomas Using Diffusion Tensor Imaging-Based Fiber Tracking. World Neurosurg 2017;99:70–78. doi: 10.1016/j.wneu.2016.11.110 [DOI] [PubMed] [Google Scholar]
  • [9].Li Huan, Wang Liang, Hao Shuyu, et al. Identification of the Facial Nerve in Relation to Vestibular Schwannoma Using Preoperative Diffusion Tensor Tractography and Intraoperative Tractography-Integrated Neuronavigation System. World Neurosurg 2017;107:669–677. [DOI] [PubMed] [Google Scholar]
  • [10].Hilly O, Chen JM, Birch J, et al. Tensor Imaging Tractography of the Facial Nerve in Patients With Cerebellopontine Angle Tumors. Otol Neurotol 2016;37(4):388–393. [DOI] [PubMed] [Google Scholar]
  • [11].Borkar SA, Garg A, Mankotia DS, Joseph SL, Suri A, Kumar R, et al. Prediction of facial nerve position in large vestibular schwannomas using diffusion tensor imaging tractography and its intraoperative correlation. Neurol India. 2016;64(5):965–970. [DOI] [PubMed] [Google Scholar]
  • [12].Yoshino M, Abhinav K, Yeh FC, Panesar S, Fernandes D, Pathak S, et al. Visualization of Cranial Nerves Using High-Definition Fiber Tractography. Neurosurgery. 2016;79(1):146–165. [DOI] [PubMed] [Google Scholar]
  • [13].Yoshino M, Kin T, Ito A, Saito T, Nakagawa D, Ino K, et al. Combined use of diffusion tensor tractography and multifused contrast-enhanced FIESTA for predicting facial and cochlear nerve positions in relation to vestibular schwannoma. J Neurosurg 2015;123(25670618):1–9. [DOI] [PubMed] [Google Scholar]
  • [14].Wei PH, Qi ZG, Chen G, Hu P, Li MC, Liang JT, et al. Identification of cranial nerves near large vestibular schwannomas using superselective diffusion tensor tractography: experience with 23 cases. Acta Neurochir 2015;157(7):1239–1249. [DOI] [PubMed] [Google Scholar]
  • [15].Choi KS, Kim MS, Kwon HG, Jang SH, Kim OL. Preoperative identification of Facial Nerve in Vestibular Schwannomas Surgery Using Diffusion Tensor Tractography. J Korean Neurosurg Soc 2014;56(1):11–15. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [16].Zhang Y, Chen Y, Zou Y, Zhang W, Zhang R, Liu X, et al. Facial nerve preservation with preoperative identification and intraoperative monitoring in large vestibular schwannoma surgery. Acta Neurochir 2013;155(10):1857–1862. [DOI] [PubMed] [Google Scholar]
  • [17].Roundy N, Delashaw JB, Cetas JS. Preoperative identification of the facial nerve in patients with large cerebellopontine angle tumors using high-density diffusion tensor imaging. J Neurosurg 2012;116(4):697–702. [DOI] [PubMed] [Google Scholar]
  • [18].Gerganov VM, Giordano M, Samii M, Samii A. Diffusion tensor imaging-based fiber tracking for prediction of the position of the facial nerve in relation to large vestibular schwannomas. J Neurosurg 2011;115(6):1087–1093. [DOI] [PubMed] [Google Scholar]
  • [19].Chen DQ, Quan J, Guha A, Tymianski M, Mikulis D, Hodaie M. Three-dimensional in vivo modeling of vestibular schwannomas and surrounding cranial nerves with diffusion imaging tractography. Neurosurgery. 2011;68(4):1077–1083. [DOI] [PubMed] [Google Scholar]
  • [20].Taoka T, Hirabayashi H, Nakagawa H, Sakamoto M, Myochin K, Hirohashi S, et al. Displacement of the facial nerve course by vestibular schwannoma: Preoperative visualization using diffusion tensor tractography. J Magn Reson Imaging. 2006;24(5):1005–1010. [DOI] [PubMed] [Google Scholar]
  • [21].Ning L, Laun F, Gur Y, DiBella EVR, Deslauriers-Gauthier S, Megherbi T, et al. Sparse Reconstruction Challenge for diffusion MRI: Validation on a physical phantom to determine which acquisition scheme and analysis method to use? Med Image Anal 2015;26(1):316–331. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [22].Pereira RS, Harris AD, Sevick RJ, Frayne R. Effect of b value on contrast during diffusion-weighted magnetic resonance imaging assessment of acute ischemic stroke. J Magn Reson Imaging. 2002;15(5):591–596. [DOI] [PubMed] [Google Scholar]
  • [23].Schilling KG, Rheault F, Petit L, Hansen CB, Nath V, Yeh FC, et al. Tractography dissection variability: What happens when 42 groups dissect 14 white matter bundles on the same dataset? Neuroimage. 2021;243:118502. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [24].Xie G, Zhang F, Leung L, Mooney MA, Epprecht L, Norton I, et al. Anatomical assessment of trigeminal nerve tractography using diffusion MRI: A comparison of acquisition b-values and single- and multi-fiber tracking strategies. Neuroimage Clin 2020;25:102160. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [25].Van Essen DC, Smith SM, Barch DM, Behrens TEJ, Yacoub E, Ugurbil K, et al. The WU-Minn Human Connectome Project: an overview. Neuroimage. 2013;80:62–79. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [26].Glasser MF, Sotiropoulos SN, Wilson JA, Coalson TS, Fischl B, Andersson JL, et al. The minimal preprocessing pipelines for the Human Connectome Project. Neuroimage. 2013;80:105–124. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [27].Sotiropoulos SN, Moeller S, Jbabdi S, Xu J, Andersson JL, Auerbach EJ, et al. Effects of image reconstruction on fiber orientation mapping from multichannel diffusion MRI: Reducing the noise floor using SENSE: Effects of dMRI Image Reconstruction on Fiber Orientations. Magn Reson Med 2013;70(6):1682–1689. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [28].Andersson JLR, Sotiropoulos SN. Non-parametric representation and prediction of single- and multi-shell diffusion-weighted MRI data using Gaussian processes. Neuroimage. 2015;122:166–176. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [29].Andersson JLR, Skare S, Ashburner J. How to correct susceptibility distortions in spin-echo echo-planar images: Application to diffusion tensor imaging. Neuroimage. 2003;20(2):870–888. [DOI] [PubMed] [Google Scholar]
  • [30].Andersson JLR, Sotiropoulos SN. An integrated approach to correction for off-resonance effects and subject movement in diffusion MR imaging. Neuroimage. 2016;125:1063–1078. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [31].Malcolm JG, Shenton ME, Rathi Y. Filtered multi-tensor tractography. Psychiatry: Interpersonal and Biological Processes. 2011;29(9):1664–1675. [Google Scholar]
  • [32].Reddy CP, Rathi Y. Joint Multi-Fiber NODDI Parameter Estimation and Tractography Using the Unscented Information Filter. Front Neurosci 2016;10:166. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [33].Zhang F, Noh T, Juvekar P, Frisken SF, Rigolo L, Norton I, et al. SlicerDMRI: Diffusion MRI and Tractography Research Software for Brain Cancer Surgery Planning and Visualization. JCO Clin Cancer Inform 2020;4:299–309. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [34].Malcolm JG, Shenton ME, Rathi Y. Neural tractography using an unscented Kalman filter. Inf Process Med Imaging. 2009;21:126–138. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [35].Reddy CP, Rathi Y. Joint Multi-Fiber NODDI Parameter Estimation and Tractography Using the Unscented Information Filter. Front Neurosci 2016;10:166. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [36].Zhang F, Xie G, Leung L, Mooney MA, Epprecht L, Norton I, et al. Creation of a novel trigeminal tractography atlas for automated trigeminal nerve identification. Neuroimage. 2020;220:117063. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [37].Zhang F, Wu Y, Norton I, Rathi Y, Golby AJ, O’Donnell LJ. Test-retest reproducibility of white matter parcellation using diffusion MRI tractography fiber clustering. Hum Brain Mapp 2019;40(10):3041–3057. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [38].Liao R, Ning L, Chen Z, Rigolo L, Gong S, Pasternak O, et al. Performance of unscented Kalman filter tractography in edema: Analysis of the two-tensor model. Neuroimage Clin 2017;15:819–831. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [39].Robert MT, Gutterman J, Ferre CL, Chin K, Brandao MB, Gordon AM, et al. Corpus Callosum Integrity Relates to Improvement of Upper-Extremity Function Following Intensive Rehabilitation in Children With Unilateral Spastic Cerebral Palsy. Neurorehabil Neural Repair. 2021;35(6):534–544. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [40].Taghvaei M, Cook P, Sadaghiani S, Shakibajahromi B, Tackett W, Dolui S, et al. Young versus older subject diffusion magnetic resonance imaging data for virtual white matter lesion tractography. Hum Brain Mapp Published online May 6, 2023. doi: 10.1002/hbm.26326 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [41].Shapey J, Vos SB, Vercauteren T, Bradford R, Saeed SR, Bisdas S, et al. Clinical Applications for Diffusion MRI and Tractography of Cranial Nerves Within the Posterior Fossa: A Systematic Review. Front Neurosci 2019;13:23. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [42].R Core Team. R: A Language and Environment for Statistical Computing. Published online 2023. https://www.R-project.org/
  • [43].Benjamini Y, Hochberg Y. Controlling the false discovery rate: A practical and powerful approach to multiple testing. J R Stat Soc 1995;57(1):289–300. [Google Scholar]
  • [44].R Core Team. R: A Language and Environment for Statistical Computing. Published online 2020. https://www.R-project.org/
  • [45].Wickham H ggplot2: Elegant Graphics for Data Analysis. Published online 2016. https://ggplot2.tidyverse.org
  • [46].Wickham H, Averick M, Bryan J, Chang W, McGowan LD, François R, et al. Welcome to the tidyverse. Journal of Open Source Software. 2019;4(43):1686. doi: 10.21105/joss.01686 [DOI] [Google Scholar]
  • [47].Kassambara A rstatix: Pipe-Friendly Framework for Basic Statistical Tests. Published online 2023. https://CRAN.R-project.org/package=rstatix
  • [48].Kassambara A ggpubr: “ggplot2” Based Publication Ready Plots. Published online 2023. https://CRAN.R-project.org/package=ggpubr
  • [49].McHugh ML. Interrater reliability: the kappa statistic. Biochem Med 2012;22(3):276–282. [PMC free article] [PubMed] [Google Scholar]
  • [50].Friedman M The Use of Ranks to Avoid the Assumption of Normality Implicit in the Analysis of Variance. J Am Stat Assoc 1937;32(200):675–701. [Google Scholar]
  • [51].Kendall MG. Rank correlation methods. Published online 1948. https://psycnet.apa.org/fulltext/1948-15040-000.pdf
  • [52].Eisinga R, Heskes T, Pelzer B, Te Grotenhuis M. Exact p-values for pairwise comparison of Friedman rank sums, with application to comparing classifiers. BMC Bioinformatics. 2017;18(1):68. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [53].Mangiafico SS. Summary and Analysis of Extension Program Evaluation in R. 1.20.01 ed. Rutgers Cooperative Extension; 2016. [Google Scholar]
  • [54].Mangiafico SS. rcompanion: Functions to Support Extension Education Program Evaluation. Published online 2023. https://CRAN.R-project.org/package=rcompanion/
  • [55].Pohlert T PMCMRplus: Calculate Pairwise Multiple Comparisons of Mean Rank Sums Extended. R Package. Published online 2021. https://CRAN.R-project.org/package=PMCMRplus
  • [56].Thomas C, Ye FQ, Irfanoglu MO, Modi P, Saleem KS, Leopold DA, et al. Anatomical accuracy of brain connections derived from diffusion MRI tractography is inherently limited. Proc Natl Acad Sci U S A 2014;111(46):16574–16579. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [57].Behan B, Chen DQ, Sammartino F, DeSouza DD, Wharton-Shukster E, Hodaie M. Comparison of Diffusion-Weighted MRI Reconstruction Methods for Visualization of Cranial Nerves in Posterior Fossa Surgery. Front Neurosci 2017;11:554. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplemental Digital Content, Fig 1
NIHMS2009137-supplement-Supplemental_Digital_Content__Fig_1.docx (423.4KB, docx)
Supplemental Digital Content, Table 1
NIHMS2009137-supplement-Supplemental_Digital_Content__Table_1.docx (14.8KB, docx)
Supplemental Digital Content, Table 2
NIHMS2009137-supplement-Supplemental_Digital_Content__Table_2.docx (14.2KB, docx)
Supplemental Digital Content, Table 3
NIHMS2009137-supplement-Supplemental_Digital_Content__Table_3.docx (14.1KB, docx)
Supplemental Digital Content, Table 4
NIHMS2009137-supplement-Supplemental_Digital_Content__Table_4.docx (14.2KB, docx)
Supplemental Digital Content, Table 5
NIHMS2009137-supplement-Supplemental_Digital_Content__Table_5.docx (16.5KB, docx)

RESOURCES