Muscle fat quantification using magnetic resonance imaging: case-control study of Charcot-Marie-Tooth disease patients and volunteers.

BACKGROUND: This study aimed to evaluate the potential value of 3D multiple gradient echo Dixon-based magnetic resonance imaging (MRI) sequence as a tool for thigh intramuscular fat quantification in Charcot-Marie-Tooth disease (CMT) patients.
METHODS: A prospective comparison study comprising 18 CMT patients and 18 age/sex-matched volunteers was performed. MRI including 3D multiple gradient echo Dixon-based imaging was performed for each subject. Region of interest analyses were performed at the upper and lower third of both thighs. The two-sample t-test or Wilcoxon rank sum test was used for intergroup comparison of the mean muscle fat fraction. Intraclass correlation coefficients were used to evaluate the interobserver agreement and test-retest reproducibility. Semiquantitive analysis using the Goutallier classification (Grades 0-4) was performed on T1-weighted images in upper thigh muscles. For Goutallier Grade 0 muscles, comparison of the mean intramuscular fat fraction between volunteers and CMT patients was performed.
RESULTS: The interobserver agreements were excellent for all measurements (intraclass correlation coefficients > 0.8). Mean muscle fat fractions were significantly higher in all the measured muscles of CMT patients (P < 0.05) except in the adductor magnus in the upper thigh (P = 0.109). Goutallier Grade 0 muscles of the CMT patients showed a significantly higher mean fat fraction compared with that of the volunteers (P < 0.05).
CONCLUSIONS: 3D multiple gradient echo Dixon-based MRI is a reproducible and sensitive technique which can reveal a significant difference in the fat fraction of thigh muscle, including comparison between Goutallier Grade 0 muscles, between CMT patients and volunteers.

Introduction

Charcot–Marie–Tooth disease (CMT) represents a group of hereditary neuromuscular disorders linked to various gene mutations responsible for either primary axonal degeneration or primary myelin changes with eventual axonal degeneration.1 The consequence is peripheral neuropathy with a characteristic distal predominant limb‐muscle wasting and sensory loss.1 It has been reported that there is a substantial variability in the clinical course of the disease among various subtypes and even within the same subtype of the disease.1, 2 Thus, careful monitoring of patients is necessary. Currently, there exists no approved pharmacological treatment and supportive measures remain the main treatment for CMT, but several preclinical and clinical studies using pharmacological agents have shown promising results.3, 4, 5, 6 In addition, recent advances in knowledge regarding the genetic basis of CMT with numerous genes identified to be associated with CMT suggest potentials for developing possible treatment options for the disease in the future.7 In this regard, an objective and reliable tool for assessing the status of patients including the evaluation of treatment response would be necessary.

In the management of CMT patients, careful assessment and monitoring of the affected muscle is important. One of the common problems encountered in these patients in clinical practice is the lack of a standardized method to assess muscle degeneration. Numerous scoring systems, which composed of multiple clinical parameters including limb strength, have been developed for diagnosis and monitoring of CMT patients.8, 9, 10 Despite several revisions, there remain some limitations including issues regarding inter‐rater and intra‐rater agreement.8 Muscle biopsy may provide some information on muscle fat infiltration, but it is impractical to use it on a regular basis due to its invasiveness and limited number of biopsy sites, which does not represent the true state of the disease with multiple muscles involved.

Magnetic resonance imaging (MRI) allows general visual assessment for the degree and distribution of intramuscular fat infiltration in multiple muscle compartments, which is a commonly associated finding in CMT patients.11, 12 A semiquantitative grading system using a conventional T1‐weighted imaging sequence described by Goutallier et al. is widely used scoring system for various muscular abnormalities.13 This system has a significant shortcoming of being highly observer dependent14, 15 and lacking quantitative data. Furthermore, because this classification relies on macroscopic fat signal and only has five grades, it is not optimized for assessment of early fat infiltration or interval progression that is not severe enough to result in change of grades.16 Dixon MRI is an emerging imaging technique for fat fraction measurement that exploits the capability to differentiate the individual contributions of water and fat in each voxel of tissue using the chemical shift difference between the two.17 Recent Dixon‐based MRI techniques generate fat fraction maps that allows direct quantitative measurement of the fat proportion within the designated region of interest (ROI).18 Studies have reported encouraging results with using Dixon‐based techniques for fat quantification in skeletal muscle.15, 16 Given the significant clinical implication of the muscle assessment in CMT patient management, a reproducible quantitative imaging parameter acquired through Dixon‐based MRI would be desirable.

In our study, the potential value of intramuscular fat quantification using a 3D multiple gradient echo Dixon‐based MRI sequence was sought by performing a prospective comparison study encompassing CMT patients and volunteers. We performed analyses in thigh, where fat infiltration is less prominent compared to that in calf due to the predominant nature of distal muscle involvement in CMT. Furthermore, we aimed to evaluate the potential of 3D multiple gradient echo Dixon sequence in terms of sensitivity by comparing the fat fraction of grossly normal muscle, based on Goutallier grading, between CMT patients and volunteers.

Materials and methods

Study population

The study was approved by our institutional review board. Between February and June 2017, 18 patients diagnosed with CMT Type I by genetic analysis and electrophysiologic study prospectively underwent MRI. Seventeen patients who received genetic analysis were diagnosed as CMT Type IA. One patient who did not receive genetic testing was diagnosed with CMT Type I based on electrophysiologic study and clinical history. The cohort was confined to patients aged between 20 and 40 years old (mean ± standard deviation: 30.1 ± 4.3 years; range: 23–37 years; 8 male and 10 female participants). Patients had no contraindication for MRI, such as claustrophobia or metallic implant. A public notice was posted to recruit 18 age‐matched and sex‐matched volunteers with no history of peripheral neuropathy or other pathologic condition in the lower extremities.

A neurologist (B.C, with 21 years of experience) performed a neurologic examination to assess for signs of abnormality prior to MRI. Finally, 18 age/sex‐matched healthy volunteers (mean age: 28.2 ± 1.2 years; age range: 20–36 years; 8 male and 10 female participants) were recruited. All participants gave written informed consent prior to MRI. No participant dropped out, so data acquired from 18 subjects in each group were used for analyses.

Magnetic resonance imaging acquisition

Magnetic resonance imaging (MRI) was obtained using a 3.0‐T MRI system (Ingenia; Philips Healthcare, Best, the Netherlands) with a 16‐channel anterior coil and posterior built‐in coil. The following MRI sequences were obtained for morphologic imaging: coronal and axial T1‐weighted turbo spin echo sequences, and axial T2‐weighted Dixon sequence. From the Dixon sequence, water‐only, fat‐only, in‐phase, and out‐of‐phase images were produced. The MRI protocols are detailed in Table 1.

Table 1

Summary of magnetic resonance imaging parameters

Imaging parameters	T1‐weighted TSE imaging in axial plane	T1‐weighted TSE imaging in coronal plane	T2‐weighted Dixon imaging in axial plane	mDixon‐Quant
Repetition time (ms)	613.7	450–650	4635.5	6.6
Echo time (ms)	16.7	15	80	1.01, 1.91, 2.81, 3.71, 4.61, 5.51
Flip angle (°)	90	90	90	3
No. of signals averaged	1	1	1	2
Reconstructed voxel size (mm)	0.68	0.68	0.68	1.82
Matrix size	320 × 320	320 × 317	320 × 320	192 × 192
Field of view (mm)	350 × 350	350 × 350	350 × 350	350 × 350
Section thickness (mm)	2	5	2	6
Gap (mm)	1	0.5	1	0
Number of slices	67	25	67	140
Imaging time (s)	162	200	194	72.5

mDixon‐Quant, 3D multiple gradient echo Dixon‐based sequence for fat quantification; N/A, not applicable; TSE, turbo‐spin echo.

A 3D multiple gradient echo Dixon‐based MRI sequence (mDixon‐Quant; Philips Healthcare, Best, the Netherlands) was obtained for fat fraction quantification which sampled six‐echo data. Images were obtained for the pelvic girdle and the thighs at levels from the anterior inferior iliac spine through the distal end of the femur, in the axial plane. Fat and water‐only images were consequentially reconstructed, thereby automatically generating axial fat fraction maps with correction for confounding effects of T2* decay.

Data analysis

Water‐only images and fat fraction maps were loaded into image‐processing software (IntelliSpace Portal, version 5.0; Philips Healthcare). The segmentation of all muscle compartments was manually performed by two independent radiologists (Y.C.Y and H.S.K, with 14 and 5 years of experience in musculoskeletal radiology, respectively) who were blinded to the clinical information. ROIs were initially drawn on water‐only images because it offers better visualization of muscle boundaries. Using copy‐and‐paste function at the workstation, ROIs with identical shape, size, and position were generated on fat fraction maps.

Thigh muscles were segmented to define seven muscles: rectus femoris, vastus lateralis, vastus medialis, biceps femoris (long head), semitendinosus, adductor magnus, and gracilis (Figures 1, 2, 3). These muscles were chosen for analysis as they were constantly visualized with relatively discrete margins. Two levels were considered to measure muscle fat fraction—the upper and lower third of both thighs. Among axial image slices containing the uppermost part of the femoral heads to the lower most part of the femoral condyles, these two levels were determined using the slice number of the axial images. Measurement was not performed for the biceps femoris in the upper third and adductor magnus in the lower third due to the small cross‐sectional area and inconsistent demonstration of muscle. The manual drawing of ROI was performed so that the boundaries were within 1–2 mm from the muscle fascia. A training session was conducted prior to the measurements, to familiarize both radiologists with the areas of measurement and defining ROI for muscle. One radiologist again drew ROIs that best fit the muscle boundaries to measure the cross‐sectional area of the muscles at the upper and lower third of both thighs.

Figure 1

An example of semiquantitative analysis and region of interest (ROI) measurement performed for intramuscular fat quantification in upper thighs of a 33 year‐old male volunteer. All the muscles of upper thigh were Graded 0 on axial T1‐weighted image (A) based on semiquantitative scale described by Goutallier et al. ROIs were initially drawn in muscles of corresponding level on water‐only image (B) acquired from 3D multiple gradient echo Dixon‐based magnetic resonance imaging. Using copy‐and‐paste function, ROIs with identical shape, size, and position were generated on fat fraction map (C). Measured intramuscular fat fraction ranged from 1.33 to 7.36%. [Rectus femoris, 1.48% (left) and 2.48% (right); vastus lateralis, 2.35% and 3.04%; vastus medialis, 1.33% and 1.68%; semitendinosus, 2.85% and 4.52%; adductor magnus, 2.92% and 2.50%; gracilis, 6.94% and 7.36%]. Notes: AM, adductor magnus; G, gracilis; RF, rectus femoris; S, semitendinosus; VL, vastus lateralis; VM, vastus medialis.

Figure 2

An example of region of interest (ROI) measurement performed for intramuscular fat quantification in lower thighs of 35 year‐old male Charcot–Marie–Tooth disease patient. ROIs were initially drawn on water‐only image (A) acquired from 3D multiple gradient echo Dixon‐based magnetic resonance imaging. ROIs with identical shape, size, and position were generated on fat fraction map (B). Measured intramuscular fat fraction ranged from 2.87% to 20.79%. [Rectus femoris, 7.03% (left) and 2.87% (right); vastus lateralis, 7.37% and 7.35%; vastus medialis, 5.78% and 4.72%; biceps femoris, 20.79% and 17.30%; semitendinosus,13.28% and 8.74%; gracilis, 16.18% and 10.36%]. Notes: BF, biceps femoris; G, gracilis; RF, rectus femoris; S, semitendinosus; VL, vastus lateralis; VM, vastus medialis.

Figure 3

An example of semiquantitative analysis and region of interest (ROI) measurement performed for intramuscular fat quantification in upper thighs of a 36 year‐old male Charcot–Marie–Tooth disease patient. All the muscles of upper thigh were Graded 0 on axial T1‐weighted image (A) based on semiquantitative scale described by Goutallier et al. ROIs were initially drawn in muscles of corresponding level on water‐only image (B) acquired from 3D multiple gradient echo Dixon‐based MRI. Using copy‐and‐paste function, ROIs with identical shape, size, and position were generated on fat fraction map (C). Measured intramuscular fat fraction ranged from 3.54% to 7.29%. [Rectus femoris, 4.71% (left) and 4.64% (right); vastus lateralis, 6.86% and 4.49%; vastus medialis, 4.33% and 5.04%; semitendinosus, 4.47% and 6.33%; adductor magnus, 3.54% and 3.92%; gracilis, 7.29% and 7.00%]. Notes: AM, adductor magnus; G, gracilis; RF, rectus femoris; S, semitendinosus; VL, vastus lateralis; VM, vastus medialis.

All subjects were examined two times to evaluate the test–retest reproducibility of the 3D multiple gradient echo Dixon‐based MRI. Using copy‐and‐paste function, each radiologist independently generated ROIs on fat fraction maps of the second image set.

Semiquantitative assessment

Two radiologists analysed T1‐weighted images of thigh muscles and graded the presence of fatty infiltration independently based on a five‐point semiquantitative scale described by Goutallier et al.19: Grade 0, normal; Grade 1, some fatty streaks; Grade 2, less fat than muscle; Grade 3, fatty degeneration of 50%; Grade 4, fatty infiltration of more than 50%. Radiologists were blinded to the clinical information, and analyses were performed at the same level where the fat fraction measurement for the upper thigh muscle was performed on the fat fraction map.

Clinical assessment

Demographic and clinical data of the CMT patients are shown in Table 2. Disease duration was determined by asking the patients when their symptoms, such as distal muscle weakness, foot deformity, and/or sensory change, first appeared. The severity of CMT was evaluated using the functional disability scale20 and the CMT neuropathy score (CMTNS).21 The British Medical Research Council scale was used to assess the thigh muscle strength.22 Knee extension and flexion were tested on both sides. One CMT patient showed Medical Research Council Grade 4+ for knee extension for both sides while all the other patients showed no abnormality. None of the patients experienced subjective weakness regarding thigh movements. Electrophysiologic study result in the peroneal and tibial nerves of the CMT patients are shown in Table 3.

Table 2

Demographic and clinical data of the Charcot–Marie–Tooth disease patients

Patient number	Gender	Age	Disease duration (years)	Charcot–Marie–Tooth neuropathy score	Functional disability scale	Knee extensiona (left/right)	Knee flexiona (left/right)
1	Male	22	11	5	1	5/5	5/5
2		24	2	7	1	5/5	5/5
3		26	0	14	2	5/5	5/5
4		27	12	16	2	5/5	5/5
5		29	16	16	2	5/5	5/5
6		34	19	11	1	5/5	5/5
7		36	29	21	3	5/5	5/5
8	Female	20	5	15	3	5/5	5/5
9		22	6	9	1	5/5	5/5
10		23	2	7	1	5/5	5/5
11		28	11	18	2	5/5	5/5
12		32	30	9	1	5/5	5/5
13		35	16	16	2	5/5	5/5
14		33	33	6	1	5/5	5/5
15		25	12	17	2	4+/4+	5/5
16		34	5	6	1	5/5	5/5
17		31	18	9	1	5/5	5/5

Assessment based on British Medical Research Council scale.

Table 3

Electrophysiologic study result of the Charcot–Marie–Tooth disease patients

Patient number	Peroneal nerve			Tibial nerve
	TL (ms) (left/right)	CMAP (mV) (left/right)	NCV (m/s) (left/right)	TL (ms) (left/right)	CMAP (mV) (left/right)	NCV (m/s) (left/right)
1	16.3/16.8	0.9/2.6	16.5/16.9	14.5/12.3	3.2/4.7	17.4/17.4
2	13.5/13.2	0.4/1.2	18.7/18.2	10.0/10.2	7.8/8.0	20.8/19.1
3	12.3/12.5	1.5/0.6	16.3/19.0	8.5/8.2	0.5/0.3	14.4/15.2
4	A/A	A/A	A/A	10.6/11.4	1.0/0.5	14.4/12.7
5	11/A	1.6/A	13.0/A	12.0/12.2	0.2/0.6	16.2/14.3
6	9/7.4	2.7/2.7	20.8/20.2	11.9/8.6	0.7/3.3	20.8/23.4
7	9.4/10.2	0.8/0.6	17.6/16.4	7.8/7.1	2.3/2.0	18.6/16.4
8	A/16.5	A/0.3	A/13.7	20.1/21.2	0.3/0.9	17.3/15.1
9	9.1/9.0	1.1/1.8	24.9/21.8	9.9/7.5	4.9/6.3	28.4/26.5
10	11.6/11.1	0.8/0.8	15.3/15.1	10.2/10.7	3.6/4.1	17.1/18.5
11	A/A	A/A	A/A	11.3/16.9	0.2/0.4	A/19.2
12	7.6/8.2	4.2/3.1	34.2/31.8	5.0/9.1	10.0/3.3	26.7/36.6
13	A/A	A/A	A/A	A/A	A/A	A/A
14	12.9/9.6	2.9/4.6	17.2/19.5	10.5/10.5	7.7/7.7	19.3/19.3
15	A/A	A/A	A/A	10.4/12.6	0.6/0.5	13.7/10.0
16	A/A	A/A	A/A	A/13.6	A/1.0	A/18.2
17	17.3/A	0.2/A	14.8/A	11.3/8.1	0.7/4.7	14.3/13.1

A, absent; CMAP, compound muscle action potential; NCV, nerve conduction velocity; TL, terminal latency.

Statistical analysis

Statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC, USA). Interobserver agreements and test–retest reproducibility of muscle fat quantification were calculated using intraclass correlation coefficients, interpreted as follows: 0.81–1.00, excellent; 0.61–0.80, good, and ≤0.60, poor agreement. Interobserver agreement of semiquantitative assessment was performed using Cohen's kappa analysis, interpreted as follows: 0.81–1.00, almost perfect; 0.61–0.80, substantial; 0.41–0.60, moderate; 0.21–0.40, fair; and ≤0.20, slight agreement. Using Bland–Altman analysis, mean difference and standard deviation of interobserver agreements and test–retest reproducibility were also acquired.

Comparison of mean muscle fat fraction and cross‐sectional area in both thighs between CMT patients and volunteers was conducted using the Wilcoxon rank sum test. For thigh muscles classified as Goutallier Grade 0, comparison of mean intramuscular fat fraction between the two groups were performed using the Wilcoxon rank sum test. Fat fraction of the upper and lower thigh muscles (rectus femoris, vastus lateralis, vastus medialis, and semitendinosus) was compared using Wilcoxon signed rank test in CMT patients.

The relationship between the thigh muscle fat fraction and clinical data (i.e. disease duration, CMTNS, and functional disability scale) were evaluated using Spearman's correlation analysis in the CMT patients. Comparison of body mass index (BMI) between both groups was conducted using the Wilcoxon rank sum test to confirm the presence of a significant difference that may affect the intramuscular fat fraction. In addition, Spearman's correlation analysis was performed to find possible correlation between BMI and thigh muscle fat fraction in both groups.

Result

Both sets of 3D multiple gradient echo Dixon‐based MRI obtained from one volunteer and one CMT patient, respectively, showed fat‐water swap. Also, a single set of 3D multiple gradient echo Dixon‐based MRI in one volunteer showed fat‐water swap. These image sets were not included in the analysis. Thus, interobserver agreement evaluation and comparison of muscle fat fraction between the two groups were conducted for 17 volunteers and 17 CMT patients. Semiquantitative analyses of the upper thigh muscles were performed in the same study population. Test–retest reproducibility evaluation was performed for 33 subjects, which composed of 16 volunteers and 17 CMT patients. There was no significant difference in BMI between the two groups (P = 0.270; volunteers: median ± standard deviation, 21.60 ± 3.09; interquartile range, 20.60–25.10; CMT patients: median ± standard deviation, 23.06 ± 4.99; interquartile range, 21.48–27.63).

The interobserver agreements in the muscle fat fraction measurement were excellent for all fat quantification analyses (Table 4). The result of the test–retest reproducibility of quantitative fat fraction measurement was excellent in both reviewers' evaluation except for the adductor magnus muscle in the left upper thigh measured by Reviewer 1, which showed good agreement (intraclass correlation coefficients = 0.67). Data obtained by Reviewer 1 using the first image set were used for comparison of the intramuscular fat fraction. Bland–Altman analysis result for the interobserver agreements and test–retest reproducibility are displayed in Table S1.

Table 4

Interobserver agreement and test–retest reproducibility of muscle fat quantification

		Interobserver agreement				Test–retest reproducibility (R1/R2)
		Upper thigh		Lower thigh		Upper thigh		Lower thigh
		ICC	95% CI	ICC	95% CI	ICC	95% CI	ICC	95% CI
Rectus femoris	Right	0.91	0.84–0.96	0.89	0.79–0.94	0.96/0.97	0.91–0.98/0.94–0.99	0.95/0.97	0.90–0.97/0.95–0.99
	Left	0.95	0.90–0.97	0.83	0.70–0.91	0.95/0.97	0.90–0.97/0.94–0.99	0.99/0.98	0.98–1.00/0.96–0.99
Vastus lateralis	Right	0.90	0.82–0.95	0.99	0.99–1.00	0.91/0.94	0.83–0.95/0.89–0.97	1.00/1.00	1.00–1.00/1.00–1.00
	Left	0.86	0.75–0.93	1.00	0.99–1.00	0.93/0.94	0.86–0.96/0.89–0.97	1.00/1.00	1.00–1.00/1.00–1.00
Vastus medialis	Right	0.95	0.91–0.98	0.97	0.95–0.99	0.87/0.95	0.76–0.93/0.90–0.97	0.99/0.99	0.99–1.00/0.99–1.00
	Left	0.98	0.95–0.99	0.96	0.93–0.98	0.95/0.94	0.90–0.97/0.89–0.97	0.99/0.99	0.99–1.00/0.98–1.00
Semitendinosus	Right	0.99	0.98–1.00	1.00	1.00–1.00	1.00/1.00	1.00–1.00/1.00–1.00	1.00/1.00	1.00–1.00/0.99–1.00
	Left	0.94	0.89–0.97	1.00	0.99–1.00	0.89/1.00	0.79–0.94/0.99–1.00	1.00/1.00	1.00–1.00/1.00–1.00
Biceps femoris	Right	NA	NA	1.00	1.00–1.00	NA	NA	1.00/1.00	1.00–1.00/1.00–1.00
	Left	NA	NA	0.98	0.97–0.99	NA	NA	1.00/1.00	0.99–1.00/1.00–1.00
Adductor magnus	Right	0.83	0.70–0.91	NA	NA	0.94/0.96	0.89–0.97/0.92–0.98	NA	NA
	Left	0.91	0.84–0.95	NA	NA	0.67/0.99	0.47–0.83/0.99–0.99	NA	NA
Gracilis	Right	0.91	0.83–0.95	0.98	0.95–0.99	0.98/0.99	0.97–0.99/0.98–0.99	1.00/1.00	0.99–1.00/0.99–1.00
	Left	0.91	0.84–0.96	0.97	0.93–0.98	0.99/0.99	0.99–1.00/0.98–1.00	0.99/0.99	0.99–1.00/0.99–1.00

CI, confidence interval; ICC, intraclass coefficient; NA, not applicable; R1, Reviewer 1; R2, Reviewer 2.

In the upper thigh muscles, significantly higher mean fat fractions were demonstrated in CMT patients compared with volunteers in all the measured muscle besides the adductor magnus (Table 5). In the lower thigh muscle, significantly higher mean fat fractions were seen in CMT patients in all the measured muscles. There was no significant difference in the mean muscle cross‐sectional area between the two groups (Table 6). Wilcoxon signed rank test result for comparison between upper and lower thigh muscle fat fractions in the CMT patients showed significantly higher fat percentage in the vastus lateralis muscle of the lower thigh with no significant difference in other muscles (rectus femoris, vastus medialis, and semitendinosus) (Table S2).

Table 5

Comparison of thigh muscle fat quantification between healthy volunteers and CMT patients

	Rectus femoris	Vastus lateralis	Vastus medialis	Semitendinosus	Adductor magnus	Biceps femoris	Gracilis
Upper thigh
Volunteers	3.066 ± 1.313	3.587 ± 1.425	3.730 ± 1.513	5.238 ± 2.181	3.195 ± 1.202	N/A	5.349 ± 1.914
CMT patients	4.916 ± 2.939	4.649 ± 1.854	5.669 ± 2.818	11.473 ± 11.965	4.581 ± 3.581	N/A	10.849 ± 5.754
P value	<0.001*	0.011*	0.001*	<0.001*	0.109	N/A	<0.001*
Lower thigh
Volunteers	2.684 ± 1.068	4.298 ± 1.585	3.273 ± 1.547	5.199 ± 1.691	N/A	4.231 ± 2.691	6.185 ± 2.760
CMT patients	6.694 ± 6.277	9.248 ± 8.393	5.648 ± 3.565	11.410 ± 8.440	N/A	13.381 ± 16.369	14.062 ± 10.012
P value	<0.001*	0.003*	<0.001*	<0.001*	N/A	<0.001*	<0.001*

Data are displayed as mean values (%) ± standard deviation. CMT, Charcot–Marie–Tooth disease; N/A, not applicable.

Indicates statistical significance.

Table 6

Comparison of thigh muscle cross‐sectional area between healthy volunteers and CMT patients

	Rectus femoris	Vastus lateralis	Vastus medialis	Semitendinosus	Adductor magnus	Biceps femoris	Gracilis
Upper thigh
Volunteers	916.014 ± 222.276	1704.049 ± 356.582	693.467 ± 115.891	379.169 ± 97.406	2041.183 ± 493.057	N/A	278.711 ± 64.237
CMT patients	918.407 ± 252.904	1732.060 ± 334.948	732.520 ± 124.404	418.864 ± 120.585	2102.977 ± 365.292	N/A	273.254 ± 65.575
P value	0.973	0.658	0.433	0.357	0.193	N/A	0.946
Lower thigh
Volunteers	263.305 ± 178.248	1617.485 ± 506.403	1432.448 ± 342.284	462.466 ± 122.657	N/A	1007.295 ± 249.165	220.881 ± 90.093
CMT patients	199.123 ± 112.889	1575.501 ± 535.228	1512.382 ± 180.755	448.270 ± 111.431	N/A	920.659 ± 276.730	231.924 ± 75.237
P value	0.496	0.734	0.170	0.734	N/A	0.586	0.563

Data are displayed as mean values (mm2) ± standard deviation. CMT, Charcot–Marie–Tooth disease; N/A, not applicable.

The agreement between the two radiologists for semiquantitative analysis was almost perfect (kappa value of 0.843). There was disagreement of grading in nine muscles (9/408, 2.20%; Grade 0 vs. 1, two cases; Grade 1 vs. 2, six cases; and Grade 0 vs. 2, one case), all of which belonged to the patient group. One of the readers' data was used for analysis. All the upper thigh muscles were Grade 0 in volunteers (100%, 204/204). The results in CMT patients showed majority of muscles rated as Grade 0 (86.3%, 176/204), followed by Grade 1 (12.7%, 26/204), and Grades 2 and 4 (0.05%, 1/204). Fat fractions of each grade were as follows: Grade 0 (mean: 4.72; standard deviation, 2.61; range, 0.87–18.60), Grade 1 (mean: 14.40; standard deviation, 5.19; range, 7.07–26.34), Grade 2 (27.37%), and Grade 4 (72.08%) (Figure 4). Goutallier Grade 0 muscles in the CMT patient group (median ± standard deviation, 4.54 ± 3.03; interquartile range, 3.58–6.82) showed significantly higher mean fat fraction compared with that of the volunteer group (median ± standard deviation, 3.58 ± 1.86; interquartile range, 2.70–5.00) (P < 0.001).

Figure 4

Boxplots showing distribution of fat fraction in 408 thigh muscles of all subjects in each Goutallier grade (A) and Goutallier Grade 0 muscles of volunteer and Charcot–Marie–Tooth disease patient group (B). Numbers in the parentheses represent number of the measured muscles. ○, outliers.

There was a correlation between intramuscular fat fraction of the upper thigh vastus medialis and CMTNS with correlation coefficient of 0.5191 (P < 0.05) (Table S3). There was no correlation between fat fraction of other muscles and other clinical parameters. No significant correlation was found between BMI and thigh muscle fat fraction in either CMT patient or volunteer group (Table S4).

Discussion

In this study, the intramuscular fat fraction of the thigh obtained using 3D multiple gradient echo Dixon‐based MRI showed significantly higher mean values in the CMT patients compared to those in the volunteers for all the muscles besides the adductor magnus. The excellent results of interobserver agreement and test–retest reproducibility suggest that fat measurement using this technique is highly reliable. To our knowledge, little has been reported on MRI evaluation of fat infiltration of thigh muscles in CMT patients. The significant difference in the degree of intramuscular fat fraction in the thigh between the two groups demonstrated in our study may be attributed to the sensitivity of the 3D multiple gradient echo Dixon‐based MRI technique, which may not have been revealed by less advanced MRI techniques. However, further study with larger number of enrolled subjects and longitudinal data is warranted to elucidate the true significance.

With the characteristic predominant distal muscle wasting in CMT patients, only a small proportion of patients are reported to develop severe thigh muscle weakness late in the clinical course.1, 23 Most studies in the past also focused on evaluation of fat infiltration in lower leg muscles.11, 12 A recent study by Morrow et al.24 comparing the intramuscular fat fraction using a Dixon‐based MRI technique between a CMT Type IA group and a volunteer group showed a significantly higher muscle fat fraction in the lower leg but not at the thigh level. Besides the clinical factors of the enrolled subjects, the difference between our study result and this recent study may be related to technical factors. A three‐point Dixon technique, with no information regarding T2* correction, was used for fat quantification in the study of Morrow et al. while T2*‐corrected six‐point Dixon technique was used in our study. A previous study using MR spectroscopy as a reference standard reported that intramuscular fat quantification using T2*‐corrected six‐echo Dixon sequences showed a significantly better concordance with the spectroscopic data compared with those of T2*‐corrected three‐echo Dixon or non‐T2*‐corrected two‐echo Dixon technique.25 The importance of T2* correction has been proposed considering the presence of iron that causes local magnetic inhomogeneity and has been emphasized in liver fat quantification.26 Skeletal muscles are also reported to contain non‐negligible amounts of iron.27 Also, it has been suggested that it is necessary to acquire at least six echoes for the optimal separation of water and fat signals with T2*‐correction28 as in our study. Our study result may suggest that 3D multiple gradient echo Dixon‐based MRI can reveal intramuscular fat infiltration in the thigh which may be present at a relatively early stage of CMT. As with previous studies using Dixon‐based MRI for evaluation of CMT or other neuromuscular disorder patients,24, 29, 30 our result supports further application of this imaging technique for quantitative data acquisition regarding intramuscular fat fraction in future research.

Interobserver agreement and test–retest reproducibility were excellent for 3D multiple gradient echo Dixon‐based MRI. Intramuscular fat quantification using Dixon‐based quantitative MRI sequence is reported to show high interobserver and test–retest reproducibility.16 Our study result is also in agreement with the previous results and suggests that Dixon‐based MRI can provide a reliable fat measurement that is advantageous over subjective analysis based on Goutallier classification, which is reported to be highly observer dependent.14, 15 Goutallier Grade 0 muscles in the CMT patient group showed a significantly higher mean intramuscular fat fraction compared to that of the volunteer group (Figure 4). This result may imply that intramuscular fat measurement using a Dixon‐based quantitative MRI sequence can be a more sensitive and objective tool for screening of muscular degeneration. In addition, 3D multiple gradient echo Dixon‐based MRI could be obtained in a reasonable scan time (72.5 s) which was even shorter than the time taken for other conventional sequences (Table 1). Thus, it may easily be incorporated as part of a routine imaging sequence in neuromuscular disease patient evaluation.

None of the patients presented with subjective weakness of the thigh muscles, and only one patient showed slightly decreased strength for knee extension on physical examination. In this regard, our patients are subclinical in terms of proximal lower limb weakness, and intramuscular fat infiltration observed through 3D multiple gradient echo Dixon‐based MRI may reflect early manifestation of the degeneration prior to clinically evident muscle weakness. Morrow et al. reported that calf muscle fat fraction measured using Dixon‐based MRI significantly increased over 12 months in their cohort of CMT patients.29 It would be beneficial to investigate the longitudinal change of thigh muscle fat fraction in a large cohort of CMT patients in the future. Furthermore, because substantial portion of the patients eventually develop hip muscle weakness,31 it would be interesting to expand the evaluation to encompass hip muscles in future research.

In our study, a correlation was found between fat fraction of the upper thigh vastus medialis and CMTNS with correlation coefficient of 0.5191. A recent study by Morrow et al.29 reported a strong correlation between calf muscle fat fraction with CMTNS. For thigh muscle, little is known about relation between the degree of fat infiltration and clinical status in CMT patients. Because our study mostly included patients with relatively early thigh muscle fatty degeneration, relation between thigh muscle fat fraction and clinical parameters should also be sought in patients with more advanced fat infiltration. Further investigation with larger cohort is needed.

Among a total 72 sets of 3D multiple gradient echo Dixon‐based MRI, five image sets (6.9%) showed fat‐water swap. Fat‐water swaps are commonly encountered problems in Dixon technique attributed to phase shift errors.32 To overcome this artefact by dealing with B0 field inhomogeneity, multipoint Dixon techniques have been developed.32 However, fat‐water swaps still remain a problem in 3D multiple gradient echo Dixon‐based MRI.33 For further clinical application and exact measurement of intramuscular fat quantification by Dixon‐based MRI, technical advances are warranted to overcome this issue.

There were several limitations in our study. First, the number of enrolled subjects was relatively small. However, we used an age‐ and sex‐matched volunteers for comparison. Second, we lack longitudinal data of the thigh muscle fat fraction to clarify how these fat infiltration will change and affect clinical status in the future. Third, there may have been measurement errors in analyses. Fourth, the study population was limited to subjects in their 20s and 30s to evaluate MRI in detection of muscular changes in early stage of CMT because the disease onset usually occurs in the first two decades of life. Lack of patients with severe thigh muscular fatty atrophy and overt clinical symptom limited correlation of muscle fat fraction with clinical parameters such as muscle strength or electrophysiologic parameters of the sciatic nerve. Furthermore, the ability of fat quantification MRI for the distinction between intramuscular fat infiltration due to a normal age‐related process and that in denervated muscle could not be sought. This may be beyond the range of our study.

In conclusion, muscle fat quantification using 3D multiple gradient echo Dixon‐based MRI revealed a significant difference in the fat fraction in thigh muscle between CMT patients and normal volunteers, where intramuscular fat infiltration is less prominent compared with that in the calf muscle in CMT. A significant difference in fat fraction in Goutallier Grade 0 muscle between the two groups demonstrated in our result may suggest the sensitivity of the technique to indicate early fat infiltration which may be difficult to identify through visual assessment of T1‐weighted image. This technique was highly reproducible and could be obtained in a relatively short scan time. Incorporation of 3D multiple gradient echo Dixon‐based MRI as a part of the routine MRI for assessment of CMT and other neuromuscular disease patients may have clinical value.

Conflict of Interest

All of the authors (H.S.K, Y.C.Y, B.C, W.J, and J.G.C) declare that they have no conflict of interest.

Table S1. Bland‐Altman analysis for interobserver agreement and test‐retest reliability

Click here for additional data file.

Table S2. Wilcoxon signed rank test for comparison of upper and lower thigh muscle fat fraction in the CMT patients

Click here for additional data file.

Table S3. Spearman's correlation analysis for the relation between thigh muscle fat fraction and clinical data

Click here for additional data file.

Table S4. Spearman's correlation analysis for the relation between body mass index and thigh muscle fat fraction

Click here for additional data file.