Feasibility of peripheral nerve MR neurography using diffusion tensor imaging adapted to skeletal muscle disease

Abstract

Background

Diffusion tensor imaging (DTI) of peripheral nerves may provide additional information about nerve involvement in muscular disorders, but is considered difficult due to different optimal scan parameters tailored to magnetic resonance (MR) signal properties of muscle and neural tissues.

Purpose

To assess the feasibility of sciatic nerve DTI using two different approaches of region of interest (ROI)-localization in DTI scans with b-values 500 s/mm², in participants with muscular disorders and in controls.

Material and Methods

DTI of the thigh was conducted on a 3T system in ten patients (6 men, 4 women; mean age =54 ± 15 years) with neuromuscular disorders and ten controls. T1-weighted (T1W) images were co-registered to fractional anisotropy (FA) color-encoded images. The apparent diffusion coefficient (ADC), FA, and fiber track length (FTL) were analyzed by two operators using a freehand ROI and a single-point ROI covering the sciatic nerve. Interclass correlation coefficient (ICC) and Bland–Altman analysis were used for evaluation of inter-operator and inter-technical agreement, respectively.

Results

Three-dimensional visualization of sciatic nerve fiber was achievable using both techniques. The ICC of DTI metrics showed excellent inter-operator agreement both in patients and controls. Bland–Altman analysis revealed good agreement of both techniques. A maximum FTL was achieved using the single-point ROI technique, but with a lower inter-operator agreement (ICC = 0.99 vs. 0.83). The ADC and maximum FTL were significantly decreased in patients compared to controls.

Conclusion

Both ROI localization techniques are feasible to analyze the sciatic nerve in the setting of muscular disease. A maximum FTL is reached using the single-point ROI, however, at the cost of lower inter-operator agreement.

Keywords

Diffusion tensor imaging magnetic resonance imaging (MRI)MR neurography sciatic nerve apparent diffusion coefficient fractional anisotropy muscular dystrophy

Introduction

Diffusion tensor imaging (DTI), which is already well-established for evaluation of the central nervous system, is also an evolving technique for the assessment of diffusion and fiber coherence in neuromuscular diseases (1). Diffusion weighted imaging (DWI) and DTI use the principle of Brownian motion, the phenomenon of water diffusion within a certain medium or environment. Molecular motion is affected by intrinsic properties of biological tissues, reflecting both the tissue architecture and microstructure. Anisotropic diffusion occurs in densely packed tissue structures, reflecting a greater diffusion in certain directions as seen along axons or in skeletal muscle, whereas isotropic diffusion is characterized by equal diffusion in all directions, e.g. as seen in cerebrospinal fluid. Taking advantage of the characteristic diffusion properties of different tissues, DTI allows a spatial description of the medium studied (2). Recent studies evaluating DTI of skeletal muscle in chronic muscular dystrophies found significant alterations of apparent diffusion coefficient (ADC) and fractional anisotropy (FA) in the affected muscles, which seemed to be influenced by the grade of muscular fatty infiltration and degeneration (3,4). Magnetic resonance (MR) neurography allows further insights into the axonal integrity of neuronal tissues (5 –7). A reduction of FA is frequently observed in conditions of axonal degeneration due to trauma or inflammation, probably reflecting an increase of the isotropic water diffusion due to a loss of nerve tissue integrity (8 –14). The structural integrity of the nerve fiber as well as the coherence of the fiber is more sensitively reflected by the FA than the apparent diffusion coefficient (10,15). Previous studies evaluating histopathological and DTI analysis on peripheral nerves found a stronger correlation of axonal density and axonal diameter with FA values than with myelin density and myelin thickness. Furthermore, previous studies on chronic inflammatory demyelinating polyradiculoneuropathies (CIDP) found strong correlations between anisotropy and electrophysiological action potential amplitudes, which is thought to correlate with axonal damage (16). The joint application of peripheral nerve DTI and skeletal muscle DTI is considered challenging as the two techniques require different imaging parameters for optimal visualization. Whereas low b-values in the range of 400–500 s/mm² and a minimum number of diffusion gradient directions (NDGD) have been proposed for accurate DTI of the skeletal muscle (17), optimal DTI acquisitions use b-values 800–1200 s/mm² for peripheral nerve MR neurography. Sequences of higher b-values are considered to be more diffusion-weighted than sequences with lower b-values. Despite a supposed longer fiber track length (FTL) and higher image quality of tractography, higher b-values have also been shown to be more susceptible to artifacts due to a lower signal-to-noise ratio (SNR) of acquired images (18 –23).

The aim of this study was to assess the feasibility sciatic nerve DTI adapted for skeletal muscle in various muscular dystrophies and myopathies in comparison to controls, by using two different approaches of region of interest (ROI) localization: a freehand ROI and a single-point ROI technique.

Material and Methods

Study population

The local ethics committee approved the study and written consent was obtained from each volunteer and patient. Ten healthy volunteers (7 men, 3 women; age =41 ± 15 years) served as a control group. None of the controls reported recent trauma or discomfort of the limb muscles. There was no history of hereditary neuromuscular disorders.

The study group consisted of ten patients (6 men, 4 women; age =54 ± 15 years) with myotonic dystrophy type 1 (n = 2), myotonic dystrophy type 2 (n = 1), limb girdle muscular dystrophy (n = 2), Becker muscular dystrophy (n = 1), facioscapulohumeral muscular dystrophy (n = 2), central core disease (n = 1), and inclusion-body myositis (n = 1). Patient details are listed in Table 1.

Table 1.

Physiognomic data of patient study group including data of muscle fatty infiltration averaged on representative thigh muscles.

Diagnosis	Genetics	Gender	Age (years)	BMI (kg/m²)	MFF median (range) (%)	Muscle edema
Myotonic dystrophy 1	DMPK, 19q13.32	F	35	18.6	5.5 (3.7–6.2)	–
Myotonic dystrophy 1	DMPK, 19q13.32	M	36	28.7	5.5 (2.7–8.4)	–
Myotonic dystrophy 2	ZNF9, 3q21.3	M	73	237	8.6 (7.5–10.9)	ADM
Becker muscular dystrophy	DMD, Xp21.2- p21.1	M	34	30.5	5.3 (2.9–9.7)	–
Limb girdle muscular dystrophy 2A	CAPN 3, 15q15.1	F	66	27.0	26.5 (7.4–70.6)	–
Limb girdle muscular dystrophy not classified	n.d.	F	52	22.1	44.1 (8.0–71.9)	–
Facioscapulohumeral muscular dystrophy	D4Z4 array, 4q35	M	68	28.3	5.3 (4.6–7.8)	VL
Facioscapulohumeral muscular dystrophy	D4Z4 array, 4q35	F	49	28.3	45.1 (40.5–82.8)	–
Central core disease	RYR1, 19q13.2	M	56	22.1	11.3 (10.5–20.2)	VL
Inclusion-body myositis	Sporadic	M	72	32.1	21.0 (11.0–30.2)	VL, VI

VL, vastus lateralis muscle; VI, vastus intermedius muscle; ADM, adductor magnus muscle.

Magnetic resonance imaging (MRI)

Healthy volunteers and patients were examined in supine position, feet first, using a 32-channel sensitivity encoding torso array coil on a 3T MRI system (Ingenia, Software Release 5.1.7, Philips Medical Systems, Best, The Netherlands). After a gradient echo localizer sequence, an axial T2-weighted (T2W) turbo spin echo (TSE) sequence with Dixon fat-water separation, and a coronal T1-weighted (T1W) Dixon sequence with post-processing of axial in-/opposed phase and fat-/water only images, were performed for anatomical imaging and assessment of muscular or perineural edema. For DTI, an axial fat-suppressed spin-echo single-shot echo planar imaging sequence (EPI) with the number of signal averages (NSA) equal to 8, the number of diffusion-encoding gradient directions (NDGD) equal to 15, and two b-values (0 and 500 s/mm²) were used. The DTI, T1W imaging, and T2W imaging were co-localized and covered the same anatomical field of view (FOV). For details of the examination protocol, see Table 2.

Table 2.

Sequence parameters for T2 and T1 weighted imaging and DTI.

Sequence	Plane	Slice thickness (mm)	FOV	Voxel size (mm³)	Reconstruction voxel size (mm)	Acquisition matrix	Flip angle (°)	TR (ms)	TE (ΔTE) (ms)	b-value (s/mm²)	Gradient directions	Signal averaging (n)
T2 Dixon	Axial	3	270 × 221 × 300	0.65 × 0.65 × 3	0.469	416 × 342	90	8111	80	NA	NA	NA
T1 Dixon	Axial	1	250 × 250 × 250	2.0 × 1.0 × 0.5	0.488	124 × 249	7	4.2	1.19 (2.5)	NA	NA	NA
DTI	Axial	4	220 × 220 × 240	1.5 × 1.5 × 4.0	0.98	148 × 145	90	2479	66	0, 500	15	8

Image analysis

Analysis of DTI metrics was performed by two operators (SK with four years of MRI training and JY with 12 years of MRI training) blinded to clinical, e.g. electrophysiological, information. Pathological macroscopic alterations of the sciatic nerve, like edema and swelling, were visualized and assessed by the fat-suppressed high-resolution T2W and T1W Dixon sequences. Geometrically co-registered axial T1W imaging and DTI data were used as anatomical reference to identify the sciatic nerve structure (Fig. 1). MR neurography was performed using the manufacturer’s workstation and software (FiberTrak; release V2.1.3 Philips Healthcare, Cleveland, OH, USA). In both techniques (freehand ROI and single-point ROI), diffusion images were registered to the b = 0 image prior to performing fiber tracking. For fiber tracking, the software employed a fiber assignment by continuous tracking (FACT) method with thresholds previously defined by Mori et al. (24 –26).

In each patient and control, three ROIs were chosen for the extrapelvic sciatic nerve caudal to the greater sciatic foramen and proximal to the popliteal fossa, where the sciatic nerve branches into the common fibular and the tibial nerve.

For the freehand ROI technique, a ROI was manually drawn onto the outer border of the sciatic nerve, avoiding surrounding areas of connective and adipose tissue (Fig. 1a).

In a second approach, here called single-point ROI technique, a single voxel within the nerve was selected and processed for fiber tracking (Fig. 1b).

The maximum FTL, FA, and ADC of the sciatic nerve were calculated automatically by the software, using a single pixel-based ROI-line propagation technique (FA threshold =0.2, direction threshold =5.0°).

Muscle fat fraction (MFF%): The MFF% was obtained in four thigh muscles (rectus femoris [RF], semitendinosus [ST], biceps femoris [BF], and gracilis [G] muscle) using two-point modified Dixon-based MRI (2pt-MRIDIXON) with chemical-shift-encoded reconstruction of the fat and water signal (27,28). Using custom imaging software (OsiriX; version 6.5; Pixmeo), the signal intensities (SI) of three ROIs covering the whole muscle were drawn in the fat (SI_FAT) and water-only (SI_WATER) image of the proximal, middle and distal part of the RF, ST, BF, and G muscle. An additional ROI was placed in the subcutaneous fat for normalization.

The MFF% was calculated using the following algorithm:

\begin{matrix} MFF % = [{meanSI}_{FAT} \div ({meanSI}_{FAT} + {meanSI}_{WATER})] \\ \times 100 \end{matrix}

Statistical analysis

Statistical analysis was performed using GraphPad Prism 6.0f (GraphPad Software, La Jolla, CA, USA) and SPSS Statistics 22 software (IBM Inc. SPSS Statistics, Armonk, NY, USA). Normal distribution of data was assessed using the Shapiro–Wilk normality test. DTI metrics (ADC, FA, FTL) were expressed as the mean ± standard deviation of each single value. The MFF% of all four muscles analyzed was presented as median and interquartile range. Paired Student’s t test was conducted for comparison of DTI metrics in patients and controls. Data differences were considered statistically significant if a significance level α = 0.05 was reached.

The concordance of the freehand ROI and single-point ROI techniques in patients and controls was analyzed using paired Student’s t-test and Bland–Altman plotting (29). Inter-operator agreement was assessed for the single-point and freehand ROI technique in patients and controls using intra-class correlation (ICC). An ICC value of >0.81 was considered excellent agreement, ICC 0.61–0.80 substantial agreement, ICC 0.41–0.60 moderate agreement, ICC 0.21–0.40 fair agreement, and ICC < 0.20 slight agreement.

Results

In controls and in patients, no sign of nerve disease or trauma was detected in the initial anatomical T1W and T2W imaging. All patients showed a variable degree of muscular atrophy with concomitant muscular fatty infiltration (Table 1). Mild to moderate muscular edema on T2W Dixon images with fat suppression was observed in 4/10 patients, restricted to the vastus and adductor magnus muscle groups (Table 1). Three-dimensional MR neurography of the co-registered axial DTI-T1W imaging was produced without visible artifacts in patients with mild to moderate degrees of muscular wasting and fatty infiltration and controls (Figs. 2 –4).

Inter-operator agreement with the freehand ROI technique was excellent in both patients and controls and controls (ICC = 0.96–1.0). The ICC of inter-operator comparison using the single-point ROI technique was slightly lower, but still excellent (ICC = 0.83–0.97). A maximum FTL was reached using the single-point ROI technique, however at cost of a lower ICC (0.99 [95% confidence interval] (CI) = 0.96–1.0) versus 0.83 [95% CI = 0.3–0.95]) (Table 3).

Table 3.

Inter-operator agreement of DTI metrics.

	Freehand ROI			Single-point ROI
	Operator 1	Operator 2	ICC	Operator 1	Operator 2	ICC
Patients
ADC	1.42 ± 0.13	1.41 ± 0.12	0.98 (0.92–0.99)	1.48 ± 0.12	1.47 ± 0.11	0.97 (0.80–0.99)
FA	0.47 ± 0.07	0.46 ± 0.08	0.96 (0.87–0.99)	0.45 ± 0.06	0.46 ± 0.07	0.87 (0.55–0.96)
FTL	78.7 ± 26.4	78.5 ± 24.4	0.99 (0.96–1.00)	100.1 ± 24.9	90.2 ± 22.5	0.83 (0.30–0.95)
Controls
ADC	1.63 ± 0.13	1.62 ± 0.13	1.00 (0.99–1.00)	1.57 ± 0.11	1.57 ± 0.11	0.97 (0.87–0.99)
FA	0.43 ± 0.05	0.42 ± 0.05	0.99 (0.97–1.00)	0.44 ± 0.05	0.46 ± 0.07	0.84 (0.41–0.96)
FTL	110.9 ± 26.4	109.5 ± 24.26	0.99 (0.95–1.00)	136.7 ± 28.45	133.9 ± 25.23	0.97 (0.90–0.99)

All data are presented as mean ± standard deviation.

The inter-operator-agreement is presented as ICC (95% CI).

ADC, apparent diffusion coefficient (mm²/s × 10⁻³); FA, fractional anisotropy; FTL, fiber track length (mm); ICC, intra-class correlation.

The Bland–Altman analysis for comparison of the freehand ROI versus the single-point ROI showed very good agreement in patients and controls with a bias of −0.0007 ± 0.09 mm²/s × 10⁻³ (95% limits of agreement =−0.18 to 0.18 mm²/s × 10⁻³) and a bias of −0.01 ± 0.04 (95% limits of agreement =–0.09 to 0.07) for assessment of the ADC and FA, respectively. No significant inter-technical difference was found in the freehand ROI method compared to the single-point ROI method in patients or controls (ADC P = 0.3; FA P = 0.7). The Bland–Altman analysis for DTI metrics in patients and controls is illustrated in Fig. 5.

Mean ADC was significantly diminished in patients when averaged between the observers (freehand ROI: patients =1.42 ± 0.13 mm²/s × 10⁻³, controls =1.62 ± 0.13 mm²/s × 10⁻³; P = 0.002). No significant group difference was observed for FA (freehand ROI: patients = 0.46 ± 0.08, controls = 0.43 ± 0.05; P = 0.21).

Using both the freehand and single-point ROI techniques, the mean FTL was significantly shortened in patients compared to controls (freehand ROI: patients= 78.6 ± 25.2 mm, controls =110.2 ± 25.2 mm; P = 0.01). In general, the FTL obtained by the single-point ROI technique (patients = 95.1 ± 22.7 mm; controls =135.3 ± 26.5 mm) resulted in a longer continuous track length compared to the freehand ROI method, especially in the control group (bias =–25.1 ± 16.8; 95% limits of agreement =–58.1 to 7.8).

Discussion

In this study, we analyzed the feasibility of sciatic nerve MR neurography in patients with different types of myopathies and muscular dystrophies, which are associated with skeletal muscular wasting and concomitant increase of fat and connective tissue. To confirm the visual aspect of muscle wasting and fatty infiltration, muscle fat fractions were analyzed on representative muscle groups. Because of the rarity of muscular dystrophies and myopathies, we chose a patient cohort with diseases exhibiting probably similar levels of fatty replacement and muscle wasting, without known or previously described alterations of DTI or electrodiagnostic analysis of peripheral nerves.

Although recent studies emphasize the utilization of b-values > 1000 s/mm² for optimal nerve fiber visualization, the rationale of using a b-value of 500 s/mm² in this study was to test whether a single examination protocol, adjusted for properties of muscle tissue, could be of use to visualize peripheral nerves included in the examination field and to assess any co-existing nerve pathology. Referring to previous studies describing a lowering of the SNR with higher diffusion sensitivity, we supposed that an imaging protocol with lower b-values might be less prone to artifact (as SNR is degraded by fatty infiltration) and more time-efficient especially for immobilized patients (7,10,20). The results obtained in this study confirm good visualization and excellent inter-operator agreement for peripheral nerve MR neurography at b-values 500 s/mm² in patients and controls. The herein applied single-point as well as freehand ROI technique revealed comparable DTI metrics (FA and ADC). The single-point method detected slightly higher ADC values and higher maximum FTL in controls and patients. Whereas the single-point ROI method reflects the signal from a single image voxel chosen for tractography, the freehand ROI localization utilizes image voxels of the entire nerve structure including peri- and epineural tissue.

Comparing patients to healthy controls, both ROI localization techniques applied here showed a reduction of the sciatic nerve ADC and maximum FTL in patients, but without significant alteration of mean FA. The shortening of the fiber under the applied technical algorithms suggests that either the FA dropped along the tract below the determined FA threshold of 0.2 for fiber termination, or the direction of the track changed more than 5° between neighboring image slices. Because of the increased standard deviation of FA in patients versus controls, the nerve fibers in patients are more heterogeneous with segments of variably lower and higher FA. A reduction of FA was frequently observed in conditions of axonal damage, rather than demyelination (16).

FA and ADC values obtained in this study and previous published studies are highly variable, depending on several factors including b-values, number of gradient directions, voxel size and number of acquisitions in the DTI acquisition protocol (30). In comparison to previous studies analyzing the sciatic nerve in different polyneuropathies, the observed elevation of ADC and associated decrease of FA in polyneuropathy described by Vaeggemose et al. (8) and Mathys et al. (11) was not detected in our patient cohort. On the contrary, the relation of ADC and FA values in this study was shifted, with a decrease in ADC and unchanged to slight increase in FA values. The disagreement is most likely caused by different disease conditions. A decrease in ADC indicates greater restriction of water diffusion in nerve cells, which can be consistent with cytotoxic edema.

We speculate that the findings obtained here reflect a shifted appearance of the epineurium and perineurium in patients rather than a measurable defect of axonal fiber integrity, which could influence the extent but not direction of free water diffusion.

This study has some limitations. As this study focuses on the technical comparison of peripheral nerve DTI in neuromuscular disorders, the selected study cohort bears the limitation of heterogeneous disease, though all share a progressive loss of muscle associated with a variable degree of fatty infiltration. Future studies could focus on disease-specific DTI metrics under the technical conditions applied here. Another limitation of this study is that diffusion images with higher b-values than 500 s/mm² were not acquired for comparison. These limitations can be addressed in future studies.

In conclusion, both the freehand ROI and single-point ROI localization in DTI adapted for skeletal muscles are feasible and reproducible complementary MR tools to analyze peripheral nerves in neuromuscular disease. A maximum FTL can be achieved using the single-point ROI technique, however at cost of a reduced inter-observer agreement. Future studies focusing on disease-specific cohorts will be of interest.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

Mori

Zhang

. Principles of diffusion tensor imaging and its applications to basic neuroscience research. Neuron 2006; 51: 527–539.

Le Bihan

Breton

Lallemand

et al.

MR imaging of intravoxel incoherent motions: application to diffusion and perfusion in neurologic disorders. Radiology 1986; 161: 401–407.

Ponrartana

Ramos-Platt

Wren

et al.

Effectiveness of diffusion tensor imaging in assessing disease severity in Duchenne muscular dystrophy: preliminary study. Pediatr Radiol 2015; 45: 582–589.

Hooijmans

Damon

Froeling

et al.

Evaluation of skeletal muscle DTI in patients with duchenne muscular dystrophy. NMR Biomed 2015; 28: 1589–1597.

Chhabra

Lee

Bizzell

et al.

3 Tesla MR neurography–technique, interpretation, and pitfalls. Skeletal Radiol 2011; 40: 1249–1260.

Khalil

Budzik

Kermarrec

et al.

Tractography of peripheral nerves and skeletal muscles. Eur J Radiol 2010; 76: 391–397.

Kabakci

Gurses

Firat

et al.

Diffusion tensor imaging and tractography of median nerve: normative diffusion values. Am J Roentgenol 2007; 189: 923–927.

Vaeggemose

Pham

Ringgaard

et al.

Diffusion tensor imaging MR neurography for the detection of polyneuropathy in type 1 diabetes. J Magn Reson Imaging 2017; 45: 1125–1134.

Cage

Yuh

Hou

et al.

Visualization of nerve fibers and their relationship to peripheral nerve tumors by diffusion tensor imaging. Neurosurg Focus 2015; 39: E16–E16.

10.

Khalil

Hancart

Le Thuc

et al.

Diffusion tensor imaging and tractography of the median nerve in carpal tunnel syndrome: preliminary results. Eur Radiol 2008; 18: 2283–2291.

11.

Mathys

Aissa

Meyer Zu Horste

et al.

Peripheral neuropathy: assessment of proximal nerve integrity by diffusion tensor imaging. Muscle Nerve 2013; 48: 889–896.

12.

Bernabeu

Lopez-Celada

Alfaro

et al.

Is diffusion tensor imaging useful in the assessment of the sciatic nerve and its pathologies? Our clinical experience. Br J Radiol 2016; 89: 20150728–20150728.

13.

Breckwoldt

Stock

Xia

et al.

Diffusion tensor imaging adds diagnostic accuracy in magnetic resonance neurography. Invest Radiol 2015; 50: 498–504.

14.

Chen

Zhang

Lin

et al.

DTI metrics can be used as biomarkers to determine the therapeutic effect of stem cells in acute peripheral nerve injury. J Magn Reson Imaging 2017; 45: 855–862.

15.

Lehmann

Zhang

Mori

et al.

Diffusion tensor imaging to assess axonal regeneration in peripheral nerves. Exp Neurol 2010; 223: 238–244.

16.

Kakuda

Fukuda

Tanitame

et al.

Diffusion tensor imaging of peripheral nerve in patients with chronic inflammatory demyelinating polyradiculoneuropathy: a feasibility study. Neuroradiology 2011; 53: 955–960.

17.

Froeling

Nederveen

Nicolay

et al.

DTI of human skeletal muscle: the effects of diffusion encoding parameters, signal-to-noise ratio and T2 on tensor indices and fiber tracts. NMR Biomed 2013; 26: 1339–1352.

18.

Guggenberger

Eppenberger

Markovic

et al.

MR neurography of the median nerve at 3.0T: optimization of diffusion tensor imaging and fiber tractography. Eur J Radiol 2012; 81: e775–782.

19.

Chhabra

Thakkar

Andreisek

et al.

Anatomic MR imaging and functional diffusion tensor imaging of peripheral nerve tumors and tumorlike conditions. Am J Neuroradiol 2013; 34: 802–807.

20.

Stein

Neufeld

Pasternak

et al.

Diffusion tensor imaging of the median nerve in healthy and carpal tunnel syndrome subjects. J Magn Reson Imaging 2009; 29: 657–662.

21.

Baumer

Pham

Ruetters

et al.

Peripheral neuropathy: detection with diffusion-tensor imaging. Radiology 2014; 273: 185–193.

22.

Andreisek

White

Kassner

et al.

Diffusion tensor imaging and fiber tractography of the median nerve at 1.5T: optimization of b value. Skeletal Radiol 2009; 38: 51–59.

23.

Farrell

Landman

Jones

et al.

Effects of signal-to-noise ratio on the accuracy and reproducibility of diffusion tensor imaging-derived fractional anisotropy, mean diffusivity, and principal eigenvector measurements at 1.5 T. J Magn Reson Imaging 2007; 26: 756–767.

24.

Mori

Crain

Chacko

et al.

Three-dimensional tracking of axonal projections in the brain by magnetic resonance imaging. Ann Neurol 1999; 45: 265–269.

25.

Mori

Kaufmann

Davatzikos

et al.

Imaging cortical association tracts in the human brain using diffusion-tensor-based axonal tracking. Magn Reson Med 2002; 47: 215–223.

26.

Netsch

van Muiswinkel

. Quantitative evaluation of image-based distortion correction in diffusion tensor imaging. IEEE Trans Med Imaging 2004; 23: 789–798.

27.

Dixon

. Simple proton spectroscopic imaging. Radiology 1984; 153: 189–194.

28.

. Dixon techniques for water and fat imaging. J Magn Reson Imaging 2008; 28: 543–558.

29.

Bland

Altman

. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; 1: 307–310.

30.

Balbi

Budzik

Duhamel

et al.

Tractography of lumbar nerve roots: initial results. Eur Radiol 2011; 21: 1153–1159.