Magnetic resonance imaging based anatomical assessment of tongue impairment due to amyotrophic lateral sclerosis: A preliminary study

Abstract

Amyotrophic Lateral Sclerosis (ALS) is a neurological disorder, which impairs tongue function for speech and swallowing. A widely used Diffusion Tensor Imaging (DTI) analysis pipeline is employed for quantifying differences in tongue fiber myoarchitecture between controls and ALS patients. This pipeline uses both high-resolution magnetic resonance imaging (hMRI) and DTI. hMRI is used to delineate tongue muscles, while DTI provides indices to reveal fiber connectivity within and between muscles. The preliminary results using five controls and two patients show quantitative differences between the groups. This work has the potential to provide insights into the detrimental effects of ALS on speech and swallowing.

1. Introduction

Amyotrophic Lateral Sclerosis (ALS) is a progressive and neurological disorder, which affects motor neurons in the motor cortex that control voluntary movements including speech and swallowing (Paganoni et al., 2014; Benatar et al., 2016). The continuous weakening and wasting of the tongue musculature is a major contributing factor to both speech and swallowing decline in ALS. In people with ALS, hypoglossal neurons gradually degenerate, causing muscle atrophy and weakness, which thus affect speech and swallowing as well as the necessary muscles to complete those tasks (Cha and Patten, 1989). Complex muscular behaviors during speech and swallowing are produced through the flexible and efficient coordination of tongue muscles to create various deformations (Stone et al., 2004; Kuruvilla et al., 2012; Ramanarayanan et al., 2013; Woo et al., 2014). In this work, therefore, we are interested in examining anatomical differences of tongue muscles between controls and ALS patients, since the effectiveness of tongue movements for these activities is dependent on sufficient muscle bulk and a highly organized myoarchitecture. Although analyses on speech movements in both controls and ALS patients using three-dimensional (3D) motion capture technology such as electromagnetic articulography have been studied previously (Green, 2015), the effects of the disease on tongue muscle volume and fiber muscle orientation using medical imaging are currently not well documented. Such information is, however, essential for understanding how ALS impairs tongue control and mobility for speech and swallowing.

Magnetic resonance imaging (MRI) has been used in both anatomical assessment and speech production research to date (Narayanan et al., 2004; Gaige et al., 2007; Parthasarathy et al., 2007; Woo et al., 2012; Stone et al., 2014; Fu et al., 2015; Xing et al., 2016; Töger et al., 2017; Woo et al., 2017). In particular, both high-resolution magnetic resonance imaging (hMRI) and diffusion weighted imaging (DWI) have been widely used to image tongue structure for decades due to their abilities to provide information about 3D internal muscle anatomy and fiber myoarchitecture (Gaige et al., 2007; Woo et al., 2012; Stone et al., 2014; Töger et al., 2017). hMRI provides detailed 3D anatomy of each individual and internal muscle of the tongue. In addition, DWI provides a means for non-invasively characterizing the structure and geometry of tongue muscles as it allows for quantification of water diffusion and its directional anisotropy in tissues (Lansdown et al., 2007). Furthermore, DWI provides measurements of fiber orientation and connectivity such as fractional anisotropy (FA), mean diffusivity (MD), the number of fibers within each muscle, and information on the connecting fibers between muscles. These measurements from DWI offer information that facilitates comparisons before and after a disease, injury, or treatment. For an extensive review on the anatomical and functional assessment of the tongue, readers can refer to the paper by Töger et al. (2017) and the references therein.

The objective of this study is to assess anatomical differences using quantities related to fiber connectivity in both controls and ALS patients using hMRI and DWI. To achieve this, we make use of a widely used DTI analysis pipeline, demonstrating the differences in quantities related to fiber connectivity within and between superior longitudinal (SL) and genioglossus (GG) muscles that are interdigitated with each other. The GG muscle is an extrinsic muscle located in the tongue's base; it is used to examine the function of the hypoglossal nerve. The SL muscle, on the other hand, is an intrinsic muscle that spans the top of the tongue surface from the tip to far back root; it allows the tongue to retract and elevate, thereby making it become shorter or curved, respectively (Zemlin, 1968).

In the rest of this paper, we first describe our MRI protocol and analysis pipeline in Sec. 2. Then in Sec. 3, we compare quantitative measurements between normal controls and ALS patients. Finally, the discussion and summary are presented in Sec. 4.

2. Materials and methods

2.1. Subjects and MRI acquisition

Subjects. We examined seven subjects including five normal controls and two patients diagnosed with ALS. Informed consent was obtained prior to the scan. All images were acquired with the subjects in a resting supine position, and the subjects were instructed to remain still during the scan. For normal controls, two males and three females between the ages of 35 and 67 years (mean: 46) participated in this study. For ALS patients, both patients had bulbar symptoms and were evaluated with the Amyotrophic Lateral Sclerosis Functional Rating Scale-Revised (ALSFRS-R) (Cedarbaum et al., 1999). The ALSFRS-R, a 12-question survey, is a validated rating instrument to reliably assess overall ALS severity. In brief, each of the 12 questions examines the level of functional impairment on a scale from 0 (severe symptoms) to 4 (no symptoms); the total scores range from 0 to 48. In addition, the Bulbar subscore specifically examines bulbar symptoms, which uses the first three questions on the ALSFRS-R. Bulbar subscores range from 0 (severe bulbar symptoms) to 12 (no symptoms). Table 1 summarizes the characteristics of the patients used in this study.

Table 1.

Characteristics of patients.

Subjects	Age	Gender	Diagnosis	Disease duration (month)	ALSFRS-R	Bulbar Subscore
Patient 1	60	M	Bulbar onset	34	39/48	6/12
Patient 2	58	F	Bulbar onset	32	28/48	5/12

High-resolution MRI. In this study, MRI datasets of five normal controls and two ALS patients were acquired. All MRI scanning was performed on a Siemens 3.0 T Prisma scanner (Erlangen, Germany) using a 64-channel head and neck coil. We used T2-weighted Turbo Spin Echo sequence with echo train length of 17 and echo time (TE)/repetition time (TR) of 107 ms/3380 ms. Each data set covered the tongue and surrounding structures. The image size for hMRI was 256 × 256 × 200 with 1 mm × 1 mm × 1.2 mm (one patient was acquired at 0.6 mm× 0.6 mm × 2.5 mm).

DWI. We used a single-shot spin-echo, echo-planar imaging (EPI) sequence using diffusion sensitizing gradients (Tuch, 2004). Specific imaging parameters for the EPI sequence were as follows. Diffusion weighting was applied in 64 directions with a b value of 500 s/mm². Three averaging was performed. The image size for DWI was 128 × 128 × 72 with 2.3 mm × 2.3 mm × 4.6 mm. TE/TR was 55 ms/3200 ms, and the flip angle was 90°.

2.2. Extraction of measurements using our DTI analysis pipeline

Our DTI analysis pipeline to extract measurements comprises multiple steps. First, our pipeline is dependent on accurate segmentation and quantification of tongue muscles from hMRI. Thus we carry out manual segmentation of the GG and SL muscles by two experienced observers. The manual segmentations of GG and SL muscles were guided primarily by the protocol used in the vocal tract atlas (Woo et al., 2015). Any anatomical deviations across subjects were carefully differentiated based on studies of the general anatomy of tongue muscles and the vocal tract atlas. Inter-rater reliability between the two observers was estimated using intraclass correlation coefficient (ICC) (Töger et al., 2017) of muscle volumes and Dice similarity coefficient (DSC). Second, DWIs are preprocessed using ANTs (Avants et al., 2008) and Camino (Cook et al., 2006) software packages. To correct distortion and motion artifacts, affine registration of each DWI with the unweighted (i.e., first b = 0) image of each subject is carried out. This step is of particular importance to deal with potential motion artifacts caused by swallowing or fasciculation in ALS. Finally, we reconstruct diffusion tensor, visualize the muscle fibers using Trackvis (Wang et al., 2007), and compute FA, MD, the number of fibers, and connecting fibers within and between the GG and SL muscles, deformed via tongue surface-based diffeomorphic registration (Avants et al., 2008) between hMRI and the first b = 0 image. For fiber tracking, we use a deterministic tracking method by setting the angle threshold to 45° for all subjects. FA is a measure describing the directionality of molecular displacement by diffusion, which quantifies the anisotropic fraction of diffusion. High FA values mean a diffusion of water molecules along the fiber orientation. MD is an inverse measure of membrane density describing how isotropic the region of interest is as values increase. More specifically, a spin-echo signal from water diffusion can be described by the six components of the diffusion tensor, D, each component of the calculated b-matrix, b_ij, the signal intensity with no diffusion sensitization, A(b = 0), and the measured echo signal, A(b) as given by

$$\text{Ln}\left(\frac{A\left(b\right)}{A(b=0)}\right)=-\sum _{i=1}^3\sum _{j=1}^3{b}_{\mathit{i}\mathit{j}}{\mathbf{D}}_{\mathit{i}\mathit{j}},$$ (1)

where a symmetric positive definite matrix is used to characterize each voxel's water diffusion

$$\mathbf{D}=(\begin{array}{ccc}{D}_{\mathit{x}\mathit{x}}\hfill & D_{\mathit{x}\mathit{y}}\hfill & D_{\mathit{x}\mathit{z}}\hfill \\ D_{\mathit{x}\mathit{y}}\hfill & D_{\mathit{y}\mathit{y}}\hfill & D_{\mathit{y}\mathit{z}}\hfill \\ D_{\mathit{x}\mathit{z}}\hfill & D_{\mathit{y}\mathit{z}}\hfill & D_{\mathit{z}\mathit{z}}\hfill \end{array}).$$ (2)

Here, all the measurements are computed in the reference frame, x, y, and z of the gradients, and DWI indices including FA and MD are obtained by diagonalization of the diffusion tensor D,

$$\text{FA}=\sqrt{\frac{1}{2}}\frac{\sqrt{{({\lambda }_1-{\lambda }_2)}^2+{({\lambda }_2-{\lambda }_3)}^2+{({\lambda }_3-{\lambda }_1)}^2}}{\sqrt{{\lambda }_1^2+{\lambda }_2^2+{\lambda }_3^2}},$$ (3)

$$\text{MD}=\frac{{\lambda }_1+{\lambda }_2+{\lambda }_3}{3},$$ (4)

where λ₁, λ₂, and λ₃ are the eigenvalues computed from the diffusion tensor D.

3. Results

Figure 1 shows the GG and SL muscles defined on hMRI. As stated above, the GG muscle shown in blue is located in the base of the tongue and the SL muscle shown in pink is a thin muscle located along the superior surface of the tongue. The inter-rater agreement of the manual segmentations by the two observers was quantified using the ICC of muscle volumes and DSC, where the agreement was strong (ICC: 0.98 and DSC: 0.86). hMRI, fiber anatomy including the GG and the whole tongue, and FA and MD maps derived from DWI for both a normal control (first row) and an ALS patient (second row) are shown in Fig. 2. Figure 2(a) depicts hMRI of one ALS patient (second row), where the tongue did not fill the anterior oral cavity and the shape in that region was not rounded anymore. In addition, normal curvilinear and radial bands were altered, and the tongue tip and body did not touch the incisors or hard palate compared with the tongue of the normal control (first row). Furthermore, the fibers of the GG and the whole tongue in the patient were more disorganized than those in the control as illustrated in Figs. 2(b) and 2(c). Figure 3 shows FA, MD, and the number of tracks within and between muscles. In FA, there was a decrease in the values of the whole tongue, GG, and SL muscles by 21%, 16%, and 20%, respectively. In MD, there was an increase in the values of the whole tongue, GG, and SL muscles by 59%, 68%, and 60%, respectively. In addition, it was observed that there was a decrease in the number of tracks of the whole tongue, GG, SL, and the number of connecting tracks between the two muscles by 32%, 0.2%, 9%, and 79%, respectively. Although there was variability in tongue size in both controls and patients as in Table 2, we found that the number of fibers that connected the GG and SL muscles was smaller in ALS patients. Additionally, in ALS patients, MD clearly had much higher values than controls in each muscle, which indicates muscle degeneration in ALS patients.

Fig. 1.

(Color online) The GG (blue) and SL (pink) muscles are segmented on hMRI. A mid-sagittal slice of hMRI overlaid with (a) the two muscles and (b) 3D rendering of the muscles are shown. It is noted that the two muscles are interdigitated with each other.

Fig. 2.

(Color online) Qualitative comparison between a normal control and an ALS patient. The first and second row show mid-sagittal views of hMRI, the tractography of the GG, the whole tongue in the b = 0 space, FA, and MD of a normal control and a patient, respectively.

Fig. 3.

(Color online) Quantitative comparison between controls and patients (Mean ± Standard Deviation): (a) FA, (b) MD, and (c) the number of fibers within and between GG and SL muscles (GG, SL, and GG-SL indicate the GG muscle, the SL muscle, and connecting fibers between the two muscles, respectively).

Table 2.

Volumes of the whole tongue, GG, and SL muscles.

Volume (ml)	Whole Tongue	GG	SL
Control 1	98.4	24.7	17.9
Control 2	96.3	32.2	17.7
Control 3	116.2	21.4	17.8
Control 4	136.2	31.5	18.8
Control 5	106.3	36.5	24.5
Patient 1	91.5	47.5	9.9
Patient 2	87.3	26.4	13.8

4. Discussion and summary

In this work, we measured various quantities derived from DWI using a widely used analysis pipeline to shed light on the differences of tongue anatomy between normal controls and ALS patients. Our MRI protocol was assembled from clinical sequences with minimum modifications. These modifications, however, did not affect the contrast that might reveal structural changes due to ALS, and that could be adequately seen using clinical sequences. ALS affects tongue structure and the orientation within tongue muscles, which is critically responsible for performing voluntary activities such as speaking and swallowing. We focused mainly on fiber connectivity between the GG and SL muscles, in which we measured quantities including FA, MD, and the number of fiber tracks connecting the two muscles. Our preliminary DWI findings have shown that there was a decrease in FA, an increase in MD, and fewer fibers in patients than in controls, which supports the decrease in connectivity caused by ALS; the disease causes deterioration in tracks, especially those that connect different muscles together. This work will help with early detection of the disease and show its progression as well as other clinical studies conducted on the tongue. In addition, our pipeline has the potential to provide insight about how ALS-related structural changes impair speech and swallowing when combined with tongue motion data.

There are several ways to expand on this work for the future. First, in the present work, we used seven subjects in total. In our future work, we will increase the number of subjects and stratify them by age, sex, and disease stage. In addition, variability in tongue measures observed from controls will be studied to establish normative ranges of the measures for the accurate assessment and diagnosis of ALS. Furthermore, we compared the number of fibers directly within the same muscle to avoid any bias caused by the muscle atrophy or degeneration in ALS patients. In future studies with a larger sample size, we will additionally normalize by each muscle volume to assess statistical significance of tract density. Second, we will further investigate the fiber connectivity of all the muscles using the vocal tract atlas (Woo et al., 2015) to build a comprehensive connectivity map. As the size of the MRI datasets has increased, the cost and time needed for the laborious manual segmentation has become prohibitive. In addition, some muscles may not be visible well in individual volumes. Therefore, the use of the vocal tract atlas is of great importance as in other fields such as brain image analysis. Third, we will collect clinical assessment data from ALS patients such as visible atrophy and weakness in addition to ALSFRS-R evaluation for further studies of the structural integrity of the tongue. Finally, there is a lack of understanding of how muscle fiber and geometry contribute to tongue function for speech or swallowing. We will also collect tongue motion data during speech or swallowing using tagged-MRI (Xing et al., 2016) or use four-dimensional (3D space with time) atlases of tongue motion (Woo et al., 2017) to compute strain in the line of action along the muscle fiber directions from tractography to further investigate the detrimental effects of ALS on speech or swallowing.