Lower limb muscle morphology, composition and force generation capacity in typically developing children

IN

DEGREE PROJECT MEDICAL ENGINEERING, SECOND CYCLE, 30 CREDITS

STOCKHOLM SWEDEN 2019 ,

Lower Limb Muscle Morphology, Composition and Force Generation Capacity in Typically Developing Children

ANTEA DESTRO

A Gioele e Zina.

All’Africa, che mi ha

guarito il cuore.

Lower limb muscle morphology, composition and force generation capacity in typically developing children

Nedre benmuskelmorfologi, -sammans¨ attning och

kraftgenereringsf¨ orm˚ aga hos barn under typisk utveckling

Antea Destro

Degree Project in Medical Engineering Supervisor: Ruoli Wang

Reviewer: Rodrigo Moreno

KTH Royal Institute of Technology, Stockholm, Sweden

School of Engineering Sciences in Chemistry, Biotechnology and Health

A ^BSTRACT

Abstract

This study investigated the architecture and composition of eight lower limb mus-

cles in typically developing children using diffusion tensor imaging and mDixon

techniques, respectively. Moreover, the correlation between intramuscular fat frac-

tion and force generation capacity was studied. It was observed that intramuscular

fat fraction differed in the considered muscles. A positive correlation was observed

between the maximum voluntary contraction and the intramuscular fat fraction of

gastrocnemius, soleus and tibialis anterior in four subjects, implying that maxi-

mum voluntary contraction increases proportionally with intramuscular fat frac-

tion. This outcome disproves the primary hypothesis which states that lower in-

tramuscular fat fraction corresponds to a higher amount of produced force. It was

concluded that intramuscular fat fractions do not affect force generation capacity

in typically developing children.

S AMMANFATTNING

Sammanfattning

Denna studie unders¨ okte strukturen och sammans¨ attningen av ˚ atta nedre ben- muskler, hos typiskt utvecklande barn, med diffusionstensorsavbildning samt Dixon- metoder. Ut¨ over det studerades korrelationen mellan intramuskul¨ ar fettfraktion och kraftgenereringsf¨ orm˚ aga. Avvikelser i intramuskul¨ ar fettfraktion hos de un- ders¨ okta musklerna observerades. En positiv korrelation observerades mellan den maximala frivilliga kontraktionen och den intramuskul¨ ara fettfraktionen av gas- trocnemius, soleus och tibialis anterior hos fyra studiedeltagare, vilket inneb¨ ar att en den maximala frivilliga kontraktionen ¨ okar proportionellt med den intra- muskul¨ ara fettfraktionen. Det h¨ ar resultatet motbevisar den prim¨ ara hypotesen som s¨ ager att l¨ agre intramuskul¨ ar fettfraktion motsvarar h¨ ogre kraftproduktion.

Det fastst¨ alldes slutligen att intramuskul¨ ara fettfraktioner inte p˚ averkar

kraftgenereringsf¨ orm˚ agan hos typiskt utvecklande barn.

A CKNOWLEDGEMENTS

Acknowledgements

I would like to thank my supervisor Ruoli Wang for the opportunity to discover this interesting field of study and for constructively criticizing my work and guid- ing me during the whole thesis project. I would like to thank Clara K¨ orting for her kindness and availability in teaching me the fiber tracking procedure, answering my questions and commenting the segmentations. I would like to thank Arkiev D’Souza (Neuroscience Research Australia, NeuRA) for his precious help and en- thusiam in sharing his method for the intramuscular fat calculation.

I would like to thank Dr. Sven Petersson (Karolinska Universitetssjukhuset, Hud- dinge), involved in the data collection, and Prof. Taija Juutinen Finni (Neuromus- cular Research Center, University of Jyv¨ askyl¨ a) for correcting the segmentations. I would also like to thank my supervision group for all the valuable comments and feedback during the meetings.

I am very grateful to all my friends here for making this time in Sweden unforget- table and edifying. In particular, I thank Teresa for her true friendship, constant presence and wise support, even before arriving here. I want to thank those friends in Italy and those spread all over the world for their support, for visiting me, for all the long Skype calls and for not forgetting me, despite time and distance.

Voglio ringraziare di cuore la mia famiglia per avermi sempre incoraggiata e ap- poggiata nelle mie scelte, per quanto onerose e difficili. Grazie del vostro amore, della vostra comprensione, della vostra pazienza e della vostra presenza nella mia vita. Spero di continuare a rendervi orgogliosi.

Lastly, I would like to thank Carlos for encouraging me, helping me and making

my life colorful and joyful as nobody before. Thanks for never giving up on me

and being my best team mate ever.

C ^ONTENTS

Contents

List of Figures vi

List of Tables viii

List of Abbreviations ix

1 Introduction 1

2 Materials and methods 3

2.1 Study subjects . . . . 3

2.2 Experimental setup and data acquisition . . . . 3

2.3 Muscle segmentation . . . . 4

2.4 Muscle reconstruction and architecture measurements . . . . 5

2.5 Intramuscular fat fraction quantification . . . . 6

2.6 Maximum voluntary contraction measurements . . . . 9

2.7 Correlation between IFF and force generation capacity . . . . 10

3 Results 11 3.1 Architectural parameters . . . . 11

3.2 IFF and MVC correlation . . . . 11

4 Discussion 15

5 Conclusions 18

Appendix A State of the Art A1

A1 Introduction A1

A2 Lower leg anatomy A1

A3 Force generation in skeletal muscles A4

A3.1 Muscle contraction . . . . A4

A3.2 Muscle architecture . . . . A4

A4 In vivo imaging technologies to measure morphological parameters A7

A4.1 Diffusion tensor imaging . . . . A7

A4.1.1 Fiber tracking . . . . A7

C ^ONTENTS

A5 Adipose tissue in skeletal muscles A8

A5.1 Origin of fat infiltration . . . . A8 A5.2 Fat quantification methods . . . . A9 A5.2.1 Muscle biopsy . . . . A9 A5.2.2 T1-weighted images . . . A10 A5.2.3 Magnetic Resonance Spectroscopy . . . A10 A5.2.4 mDixon method . . . A10

A6 Summary A12

References vii

L ^{IST OF} F ^IGURES

List of Figures

1 Axial T1-weighted MRI slice approximately midway between the

ankle and the knee with all the outlined muscles. . . . 4

2 Fiber tracking procedure. . . . . 6

3 A slice of the mDixon scan. . . . . 7

4 Overlays in ITK-SNAP. . . . 8

5 Example of motion artifact in FF mDixon image. . . . 8

6 Example of incorrect overlay of the mDixon coordinates-transformed segmentation on the FF mDixon image. . . . 9

7 Frontal (a) and side (b) view of the used HHD. . . . 10

8 Distribution of IFF in respect to the considered muscles. . . . 12

9 Correlation between PF MVC and gastrocnemius IFF. . . . 13

10 Correlation between PF MVC and soleus IFF. . . . 13

11 Correlation between DF MVC and tibialis anterior IFF. . . . 14 A1 Axial view of the lower leg compartments. . . . A2 A2 The gastrocnemius, the soleus and its subcompartments. . . . A3 A3 The tibialis posterior and anterior. . . . A3 A4 Schematics of a) the trasmission of the AP from the motor neuron

to the muscle fiber and b) the sarcomere’s contraction. . . . A4

A5 Schematics of PA (a) and PCSA (b). . . . A5

A6 Schematics of the isometric length-tension relationship. . . . . A6

A7 Intramuscular and intermuscular fat. . . . A8

A8 Cells populations in skeletal muscles and their association to fat

L ^{IST OF} F ^IGURES

A9 Example of muscle biopsy. . . . A9 A10 Four images generated from a single sagittal mDixon sequence of

the knee. . . . A12

L ^{IST OF} T ^ABLES

List of Tables

1 Subjects characteristics expressed as mean ± SD. . . . 3 2 Mean measurements and standard deviation for architectural pa-

rameters. . . . 11

3 Mean measurements and standard deviation for IFF. . . . 12

L ^{IST OF} A BBREVIATIONS

List of Abbreviations

3D Three-dimensional

B0 Main static magnetic field

B1 Magnetic field from radio frequency coil

CP Cerebral palsy

DF Dorsi flexion

dMRI Diffusion magnetic resonance imaging DTI Diffusion tensor imaging

DWI Diffusion weighted imaging

EMG Electromyography

EPI Echo-planar imaging

F Fat image

FA Fractional anisotropy FF Fat fraction image

FL Fascicle length

FOV Field of view

GRE Gradient echo sequence

HHD Hand-held dynamometer

IFF Intramuscular fat fraction

LPCA Local Principle Component Analysis MRI Magnetic resonance imaging

MVC Maximum voluntary contraction

PA Pennation angle

PA

Deep pennation angle PA

Superficial pennation angle PCSA Physiological cross sectional area

PF Plantar flexion

SD Standard deviation

L ^{IST OF} A BBREVIATIONS

TD Typically developing

TE Echo time

TR Repetition time

TSE Turbo spin echo

W Water image

WF Water fraction image

1 I NTRODUCTION

1 Introduction

Muscles exert forces and mainly contribute to human movement. Since ancient times, they have been studied by generations of scientists who used to dissect ca- davers to observe the anatomy. Among these, Leonardo da Vinci (1452-1519) was the first one who tried to explain the mechanical relations in the human body [1], giving origin to the science of biomechanics.

Thanks to the technological and scientific progress over the centuries, biomechan- ical knowledge has been greatly broadened. Nowadays, it is known that the me- chanical potential of muscles depends on morphology, fiber type and composi- tion [2]. In particular, muscle morphology, or architecture, can be quantified by parameters such as fascicle length (FL), pennation angle (PA) and physiological cross sectional area (PCSA). These parameters can notably differ among mus- cles and individuals [3], and they change with disease status [2, 3], sex [4, 5], age [4, 6], and exercise level [3, 7–10] (see Appendix A3.2). To investigate these differences, quantitative methods to measure muscle- and subject-specific archi- tectural parameters are required.

Magnetic resonance imaging (MRI) can provide non-invasive muscle parameter measurements and can generate a wide range of imaging contrasts over large volumes of muscle [11]. However, it lacks the resolution to examine individual muscle fibers [3]. In recent years, diffusion MRI (dMRI) has been introduced to study skeletal muscle architecture which can be reconstructed through particular techniques such as diffusion tensor imaging (DTI). Based on the measurement of the apparent diffusion of water in a biological tissue [12] (see Appendix A4.1), this technique can be used for 3D fiber tractography [13] where it was shown that DTI fiber directions well resemble fascicle directions visible in high-resolution im- ages [12].

Composition is another feature for assessing the mechanical potential of muscles.

It can be altered by an increase of intramuscular fat content due to aging [14], dis-

ease status and reduced physical activity [2, 15]. Recently, fat content was proved

to represent a biomarker of disease progression [16, 17] and a potential outcome

measure for the assessment of treatments in clinical trials [16]. Several techniques

mostly based on MRI have been introduced for the assessment and quantification

of fat (see Appendix A5.2.1, A5.2.2, A5.2.3, A5.2.4). In particular, mDixon method

(see Appendix A5.2.5) is gaining importance because of its repeatability [15] and

a direct correlation to fat levels based on muscle biopsy [16]. Increased fat content

is often associated with poor muscle function, presupposing that muscle compo-

sition may contribute to muscle weakness. Schlaeger et al. [18] stated that this

occurs when non contractile tissue replaces muscles fibers. Detecting and quantify-

ing changes of intramuscular fat may help initiate individualized therapy protocols

1 I NTRODUCTION

to maintain or improve muscle function.

Literature about muscle morphology and composition in children is limited. Also, the correlation between intramuscular fat content and muscle strength has never been accurately investigated. This may be due to unsuitable equipment or little compliance with measurement protocols due to their young age.

The aim of the thesis is to provide reference values for architectural parameters

and intramuscular fat fraction (IFF) of lower leg muscles in typically developing

(TD) children, and to study the correlation between IFF and force generation ca-

pacity. The primary hypothesis is that a lower IFF corresponds to a higher amount

of produced force.

2 M ATERIALS AND METHODS

2 Materials and methods

2.1 Study subjects

The study was approved by the Regional Ethics Committee in Stockholm, Sweden, and it involved a group of nine TD children (Table 1). Prior to participation, par- ents provided informed consent according to the Declaration of Helsinki (2008).

All children met the safety requirements of MRI and none of them had recent lower leg injuries or any developmental disorder affecting the lower limbs.

Table 1: Subjects characteristics expressed as mean ± SD.

Characteristics Value

Age (years) 9.0 ± 2.2

Gender (M:F) 4:5

Heigth (cm) 135.2 ± 12.0

Weight (kg) 33.0 ± 0

Shank length (cm) 30.4 ± 3.6

2.2 Experimental setup and data acquisition

Each participant was scanned using a 3T MRI scanner (Siemens Trio) in a supine position, feet first, with 20

knee flexion and 10

ankle plantar flexion. The pro- tocol consisted of three sequences: T1-weighted images for anatomical reference, DT images for muscle architecture reconstruction, and mDixon images for IFF quantification. The following imaging parameters were used:

• T1-weighted: TSE sequence, TR/TE 605/23 ms, FOV 201 x 340 mm

, acqui- sition matrix 512 x 258 pixel, slice thickness 5 mm, voxel size 0.7 x 0.7 x 5.0 mm

, flip angle 120

, number of signal averages 1, scan time 89 s.

• dMRI: EPI sequence, TR/TE 5800/63 ms, FOV 350 x 350 mm

, acquisition matrix 140 x 140 pixel, slice thickness 2.5 mm, voxel size 2.5 x 2.5 x 2.5 mm

, 20 diffusion directions, EPI factor 140, number of signal averages 1, b = 500 s/mm

(reference image with b = 0 s/mm

), scan time 495 s.

• mDixon: 3-point 3D VIBE mDixon GRE sequence,

TR/TE1/TE2/TE3 6.79/1.70/3.44/4.89 ms, FOV 308 x 350 mm

, acquisition

matrix 352 x 310 pixel, slice thickness 3 mm, voxel size 1.0 x 1.0 x 3.0 mm

,

flip angle 4

, number of signal averages 2, scan time 69 s.

2 M ATERIALS AND METHODS

Based on simulations, a b-value between 400 and 500 s/mm

with at least 10 gradient directions is recommeneded [19]. Moreover, such value, in contrast with b = 1000 s/mm

for brain imaging, suggests lower eddy current effect [20].

2.3 Muscle segmentation

Muscle segmentation was manually performed using the open source platform 3D Slicer (http://www.slicer.org, version 4.10.1). Eight muscles were segmented:

tibialis anterior (TA) and posterior (TP), medial anterior soleus (MAS), lateral an- terior soleus (LAS), medial posterior soleus (MPS), lateral posterior soleus (LPS), lateral gastrocnemius (LG) and medial gastrocnemius (MG) (Figure 1). Muscle boundaries were outlined on each slice of the T1-weighted images and a 3D tri- angulated surface model of the muscle was generated [21]. Two external experts verified all the segmentations.

Segmentation was performed on the right lower leg. Due to motion artifact dur- ing the scanning, two of the nine enrolled subjects were excluded from the study.

Thus, seven children were analised. Among them, due to the impossibility to clearly visualise muscle boundaries, one subject had the soleus segmented as one while another subject had the left leg segmented because the images were less noisy. In fact, in this study, it is assumed that left and right lower limbs are sym- metric in TD children.

Figure 1: Axial T1-weighted MRI slice approximately midway between the ankle and the knee

with all the outlined muscles.

2 M ATERIALS AND METHODS

2.4 Muscle reconstruction and architecture measurements

Muscle fascicles reconstruction from dMR images was modified from a previous study by Bolsterlee et al. [22]. The dMR image was first denoised using a Local Principle Component Analysis (LPCA) filter that considers the four dimensionality of the data. LPCA denoising exploits the multi-directional redundancy of diffusion weighted imaging (DWI) patterns [23]. Manj´ on et al. [24] showed that the pro- duced diffusion parameters of the filter better reflect the tissue characteristics and that tractography results could be improved. The filtered data were imported into DSI-Studio (http://www.dsi-studio.labsolver.org) which reconstruct tracts through deterministic DTI fibre tracking algorithms [25].

Fiber tracking was performed for each muscle on the whole volume which corre- sponds to the seed region. Then, tracking parameters had to be set. Only tracts in the range of 0.1 ≤ FA ≤ 0.7 were considered. According to [22], these values give reasonable results. Once fiber tracking was initiated, tract was propagated in both directions of the primary eigenvector with a step size of 1 mm until the tract entered a region with FA ≤ 0.1 or until the angle between subsequent tracts exceeded 10

(Figure 2). Instead, all tracts lying within the region with FA ≥ 0.7 were set to terminative, enforcing a termination if tracks enter the region. The tract search was stopped when 4000 tracts where found in the range between 10 mm and 200 mm.

All tracts were exported into MATLAB (version R2017b, The MathWorks Inc., Nat- ick, MA, USA) and overlayed over the 3D triangulated surface model generated from 3D-Slicer (Figure 2). All tracts were visually checked for plausibility. If im- plausible in their orientation or path, the tracking had to be repeated.

To reconstruct the endpoints of the fiber tracts, the median x-, y- and z-coordinates of the endpoints at either end of the tracts were determined and translated towards the surface of the muscle model along the line connecting both median endpoints until the surface was intersected [23]. The fascicle length (FL) was calculated as the distance between the two endpoints on the superficial and deep aponeurosis.

The pennation angle (PA) was calculated as the mean value of the deep and su- perficial pennation angle (PA

and PA

). These represent the angle of the median tract to the the normal vectors of all the surface triangles inside a radius of 5 mm around the respective endpoint (Figure 2). The physiological cross-sectional area (PCSA) was calculated as:

P CSA = V

F L cos(P A) (1)

In Equation 1, V is the muscle volume.

The measurements were performed for all subjects. The soleus was not investi-

gated in the subject where it was segmented as one.

2 M ATERIALS AND METHODS

Figure 2: Fiber tracking procedure. The T1-weighted images were manually segmented in 3D Slicer. The DTI data was reconstructed in DSI-Studio and fascicle tracts (in red) were tracked within the whole muscle volume. Then, tracts and volumes were imported in MATLAB. The median endpoints of every tract were calculated and translated towards the muscle surface in both directions. When the surface was intersected, the FL was calculated. PA

and PA

represent the angle of the median tract to the surface of the muscle volume in both endpoints. Modified from [23].

2.5 Intramuscular fat fraction quantification

Intramuscular fat is adipose tissue located between muscle fibers [26–30]. The first step of the quantification was converting from DICOM to NIfTI format the FF image which was directly obtained from the mDixon scan. The water image (W), the fat image (F) and the water fraction (WF) image (Figure 3) were obtained and converted as well. The segmentation was performed on the T1-weighted image.

Thus, a 3D rigid body transformation was used to align the segmentation (in T1

coordinates) to the FF mDixon image. To verify the alignment of the transformed

segmentation file to mDixon coordinates, ITK-SNAP (http://www.itksnap.org, ver-

sion 3.8.0) was used to view the mDixon image with the transformed segmen-

tation as an overlay (Figure 4.a). In fact, the subject may have moved between

the T1 and mDixon scan, or image distortion in either one of the images could

result in poor segmentation overlap on the FF image. To quantify the exact IFF,

it was necessary to isolate the muscle of interest in the mDixon image by taking

the dot product of the segmentation image and the FF image. In MATLAB, the

segmentation image is a matrix of zeros and ones, where 1 corresponds to the

2 M ATERIALS AND METHODS

muscle of interest and 0 corresponds to areas that are not the muscle of interest.

When doing a dot product, voxels in the segmentation image are multiplied by the corresponding voxel in the FF image. So, the two images need to be of the same size. The resulting matrix contains zeros in regions that are not the region of interest and the image intensity of the FF image in the regions that are the muscle of interest. The voxel intensity was evaluated in each voxel to get the mean and standard deviation for the whole muscle volume and then converted into fat frac- tion considering that in the FF mDixon image, voxel intensity of 1000 corresponds to 100% fat. Moreover, two pixel layers were removed so that the segmentation does not overlap the muscle boundary. The idea of removing layers is to minimize small segmentation inaccuracies which would result in the inclusion of intermus- cular fat between muscle groups [14, 26, 28, 29]. Finally, it had to be checked in ITK-SNAP that the FF mask correctly overlayed on the FF image (Figure 4.b). IFF values were not considered in three kids due to motion artifact (Figure 5) and in- correct overlay of the mDixon-transformed segmentation on the FF mDixon image (Figure 6).

Figure 3: A slice of the mDixon scan. This imaging technique produces a water image and

a fat image which are used to calculate the IFF in the muscle of interest. In this

study, the WF and the FF images were automatically obtained during the scan. In

particular, voxels in the FF image are colour coded according to their fat fraction.

2 M ATERIALS AND METHODS

Figure 4: Overlays in ITK-SNAP. (a) The segmentation in mDixon coordinates on the FF image and (b) the FF mask on the FF image.

Figure 5: Example of motion artifact in FF mDixon image. (A = anterior, P = posterior, L =

left, R = right).

2 M ATERIALS AND METHODS

Figure 6: Example of incorrect overlay of the mDixon coordinates-transformed segmentation on the FF mDixon image. (A = anterior, P = posterior, L = left, R = right).

2.6 Maximum voluntary contraction measurements

For studying the correlation between IFF and force generation capacity, the maxi- mum voluntary contraction (MVC) had to be measured. In clinical settings, MVC is commonly measured with a portable hand-held dynamometer (HHD) (Figure 7) [31–33] which was customised with a fixation rig for better stabilization during the tests. The measurements took place at Astrid Lindgren Children’s Hospital, Karolinska University Hospital, and the obtained data was processed by another student. During the tests, the subjects were sitting with the knee fixed at 30

, the hip fixed at 90

and the foot in neutral position, enhancing MVC production [34].

They were asked to perform three plantar-flexions and three dorsi-flexions for 4-5

s while surface EMGs were simultaneously measured from gastrocnemius, soleus

and tibialis anterior.

2 M ATERIALS AND METHODS

Figure 7: Frontal (a) and side (b) view of the used HHD.

2.7 Correlation between IFF and force generation capacity

The correlation between IFF and force generation capacity was studied in both

plantar- and dorsi flexion in the four subjects with valid IFF data. First, an aver-

aged IFF was calculated for the gastrocnemius and the soleus. An averaged MVC

value was calculated as well for the three plantar- and dorsi flexion trials of each

subjects. Then, linear regression analysis was performed in EXCEL. The linear re-

gression equation and the determination coefficient R

were calculated. Finally,

the Pearson correlation coefficient R was computed to interpret the correlation

according to [35].

3 R ^ESULTS

3 Results

3.1 Architectural parameters

The values of the architectural parameters obtained from DTI data are given in Table 2. Values for the soleus compartments were obtained only in six subjects.

Table 2: Mean measurements and standard deviation for fascicle length (FL), pennation an- gle (PA), muscle volume (V) and physiological cross-sectional area (PCSA) obtained from DTI data off seven subjects. (MG - medial gastrocnemius, LG - lateral gastroc- nemius, TA - tibialis anterior, TP - tibialis posterior, MPS - medial posterior soleus, LPS - lateral posterior soleus, MAS - medial anterior soleus, LAS - lateral anterior soleus).

Muscle FL [mm] PA [

] V [cm

] PCSA [cm

] MG 31.9 ± 5.1 24.9 ± 5.6 72.2 ± 2.7 20.7 ± 5.4 LG 18.7 ± 5.1 25.7 ± 7.0 29.1 ± 0.7 14.3 ± 2.3 TA 54.1 ± 13.0 21.1 ± 6.4 44.4 ± 1.3 7.7 ± 1.4 TP 40.7 ± 11.4 21.2 ± 5.1 45.0 ± 1.4 10.8 ± 3.3 MPS 38.8 ± 6.2 23.0 ± 2.3 70.3 ± 1.9 17.0 ± 5.6 LPS 21.8 ± 4.6 27.6 ± 4.2 53.9 ± 1.8 21.7 ± 4.2 MAS 21.1 ± 2.7 22.5 ± 2.0 12.2 ± 3.7 5.3 ± 0.7 LAS 22.0 ± 6.2 22.2 ± 3.3 14.1 ± 3.6 5.9 ± 0.9

3.2 IFF and MVC correlation

The values of the IFF obtained from mDixon FF images of four subjects are given in

Table 3. The distribution of IFF within each muscle is showed in Figure 8. Figure 9

and 10 show the correlation between MVC data and the IFF of gastrocnemius and

soleus, respectively, during plantar flexion (PF). Figure 11 shows the correlation

between MVC and the IFF of the tibialis anterior during dorsi flexion (DF).

3 R ^ESULTS

Table 3: Mean measurements and standard deviation for IFF. (MG - medial gastrocnemius, LG - lateral gastrocnemius, TA - tibialis anterior, TP - tibialis posterior, MPS - medial posterior soleus, LPS - lateral posterior soleus, MAS - medial anterior soleus, LAS - lateral anterior soleus).

Muscle FF [%]

MG 3.9 ± 1.1 LG 2.8 ± 0.8 TA 7.7 ± 5.9 TP 4.8 ± 2.0 MPS 4.1 ± 1.5 LPS 2.9 ± 1.0 MAS 2.2 ± 0.4 LAS 2.4 ± 0.6

Figure 8: Distribution of IFF in respect to the considered muscles. Data is displayed as box-

plots where the red mark indicates the median value and the bottom and top edges

indicate, respectively, the 25

and 75

percentile. The whiskers extend to the most

extreme data points that are not considered outliers. (MG - medial gastrocnemius,

LG - lateral gastrocnemius, TA - tibialis anterior, TP - tibialis posterior, MPS - medial

posterior soleus, LPS - lateral posterior soleus, MAS - medial anterior soleus, LAS -

lateral anterior soleus).

3 R ^ESULTS

Figure 9: Correlation between PF MVC averaged on the three trials and gastrocnemius IFF averaged on medial and lateral gastrocnemius IFF. The scatter plot displays indi- vidual results and linear regression with determination coefficient R

and Pearson correlation coefficient R.

Figure 10: Correlation between PF MVC averaged on the three trials and soleus IFF averaged

on the 4 compartments. The scatter plot displays individual results and linear

regression with determination coefficient R

and Pearson correlation coefficient R.

3 R ^ESULTS

Figure 11: Correlation between DF MVC averaged on the three trials and tibialis anterior IFF.

The scatter plot displays individual results and linear regression with determina-

tion coefficient R

and Pearson correlation coefficient R.

4 D ^ISCUSSION

4 Discussion

In the first part of this study, DTI-based techniques were used to examine the ar- chitecture of lower limb muscles in TD children.

Medial gastrocnemius FL, volume and PCSA were larger than in the lateral gas- trocnemius, while the PA does not greatly differ. Previous studies about the com- parison between children with cerebral palsy (CP) and TD children mostly focused only on the medial gastrocnemius because of its functional significance in locomo- tion. D’Souza et al. [36] reported FL (38.7 ± 6.8 mm)and PA (25.6 ± 3.6

) values for TD children similar to Table 2 while volume (153.9 ± 74.6 cm

) and PCSA (38.6 ± 12.9 cm

) differ. Among the tracking parameters, only the minimum tract length was different (20 mm). The differences in D’Souza’s results may be due to a larger group of older subjects (13 males, 11.2 ± 3.6 years). The soleus was studied only in six subjects. It presents larger volume and PCSA in the poste- rior compartments while PA does not greatly differ between anterior and posterior compartments. MAS and LAS present almost the same values for each parameter while there is a considerable difference in MPS and LPS parameters, except for PA.

It is to note that FL in LPS more similar to FL of the anterior compartment. The two tibialis present same PA and volume. Moreover, TA has smaller PCSA than TP but it has the longest FL (54.1 ± 13.0 mm). This value is similar to those obtained in previous studies on adults (50.0 ± 8.0) [37]. It is possible that errors in the segmentation occured. On the basis of such results, it can be affirmed that precise and standardised values for architectural parameters cannot be collected. In fact, not only different methodologies but also the variability of growing muscles can significantly influence the results. In this regard, it would be interesting to inves- tigate the inter-subject variability based on sex, age and exercise level.

In the second part of the study, the IFF was quantified from FF mDixon images and the correlation with maximum muscle strength was investigated. This kind of study was previously done only for paraspinal muscles [18] and quadriceps mus- cle [38] in adults.

The medial compartment of the gastrocnemius shows a higher IFF than the lat- eral compartment, and the same can be observed for MPS and LPS. LAS and MAS have similar IFF. TA and TP show the highest IFF. In particular, the whisker in TA (Figure 8) is due to an incorrect inclusion of subcutaneous fat in one subject. The obtained values are in sufficient agreement with the findings reported by Noble et al. [2]. However, a precise comparison is not possible because of the larger subject group (6 males and 4 females) and higher age range (22.8 ± 3.0 years). Also, [2]

did not consider the four compartments of the soleus. A positive correlation was

observed between PF MVC and mean IFF of the gastrocnemius (R = 0.9631) and

the soleus (R = 0.9721). There is a positive correlation (R = 0.6638) also between

4 D ^ISCUSSION

DF MVC and TA IFF. A positive correlation denotes that MVC increases proportion- ally with IFF which goes against the primary hypothesis. Given such results, it is plausible that IFF of lower limb muscles of the four considered TD children is negligible compared to the amount of contractile tissue. For instance, IFF in the medial gatrocnemius is 3.9% and it is possible to intepret the remaining 96.1% as contractile tissue. Thus, force production is not affected by IFF.

Some limitations of the present study have to be addressed. First, the studied sub- ject group is small and, especially for the correlations, it is difficult to draw general conclusions. Moreover, many comparisons are not possible due to the scarcity of previous studies on this topic. Motion artifacts during data acquisition can affect the quality of the collected images and, consequently, of the final results. Second, a (semi-)automatic muscle segmentation algorithm based on both anatomical and DTI scans would be beneficial to obtain more accurate measurements of architec- tural parameters than manual segmentation. Third, during fiber tracking, it was noticed that some fascicles were not reconstructed in a plausible direction based on knowledge about muscle structure. This may be due to two reasons. A first rea- son can be the presence of fat in the voxels that can distort DTI measurements [36]

so additional constraints are needed. Secondly, only one set of tracking parame- ters was chosen based on studies by Bolsterlee et al. [22] on young adults (see Section 2.4). Thus, such parameters should be chosen to be optimal for children muscles. However, it is not clear how to objectively determine them. For example, fascicle length range can be set visually by decreasing it until no tracts are out of the volume. However, this is not a standard criteria. Thus, further studies are needed.

In the future, it would be interesting to compare the same measurements and cor- relation with children suffering from cerebral palsy. For this group of subjects, changes in the scan protocol aiming to reduce scan time are necessary. For in- stance, the number of DTI directions could be decreased but the possible variations in the outputs should be evaluated. Another possibility consists in acquiring only DT and mDixon images. By doing so, as long as muscle boundaries are visible, segmentation could be performed directly on the water image, avoiding possible inaccuracies in the transformation from T1 to mDixon coordinates and in the IFF quantification. Furthermore, it would be interesting to investigate whether IFF is muscle specific and if it represents a better force predictor than PCSA in lower limb muscles. Schlaeger et al. [18] found that IFF measurements improved the prediction of paraspinal muscle strength beyond PCSA.

According to Bolsterlee et al. [37], fiber tracts can be fitted with a three-dimensional

third-order polynomial curve and FL is calculated as the length of such curve. By

doing this, the fascicle curvature is calculated and the fiber is not considered as a

4 D ^ISCUSSION

Last, the extent to which inaccuracies in manual segmentation have propagated

to errors in muscle architecture measurements could be determined. This kind of

studies could give more insight into the pathophysiology of muscles, helping phys-

iotherapists, physicians and engineers to predict muscle degeneration, designed

personalised therapy protocols and improve the accuracy of subject-specific mus-

culoskeletal models.

5 C ^ONCLUSIONS

5 Conclusions

When muscle architecture is studied in children, results must be interpreted by

considering the high variability due to growing muscles that quickly change mor-

phology over time. Fiber tracking using diffusion tensor imaging generates good

architectural measurements. However, the procedure needs to be adapted to chil-

dren to obtain more reliable results. IFF can be successfully quantified through

mDixon images. Even though the primary hypothesis was not proved, this is, to

the author’s knowledge, the first study on the correlation between IFF and force

generation capacity in lower limb muscles in TD children. According to the results,

intramuscular fat fractions do not influence force generation capacity of such sub-

ject group. Overall, mDixon images can provide clinically important information

beyond muscle composition and potentially track early biomechanical changes in

muscles that are not atrophied or fatty infiltrated. A comparison with children

suffering from cerebral palsy could provide more clinically important information

in assessing morphological differences and IFF-force correlation.

A ^PPENDIX A S TATE OF THE ART

Appendix A

State of the Art

A1 Introduction

The purpose of this master thesis is to provide reference values for architectural parameters and intramuscular fat fraction (IFF) of lower leg muscles in typically developing (TD) children using, respectively, diffusion tensor (DT) and mDixon images. Moreover, the focus is to investigate the correlation between IFF and force generation capacity. The obtained results can eventually increase the accuracy of personalised musculoskeletal models used in the study of movement disorders and in the design of medical treatments.

Muscle architecture and muscle composition are key aspects of skeletal muscle dy- namics. In particular, several studies have shown that fat content increases with age [14, 26, 27, 39], obesity [14, 39, 40], disease status [40], muscle injuries and inactivity [28]. As a result, muscular [29, 41], mobility and metabolic dysfunc- tions [14, 26, 29, 41] may occur. Recent developments in medical imaging have enabled quantitative measurements of fat infiltration to evaluate the progress of specific diseases. Among the available image sequences, mDixon images are gain- ing importance.

The state of the art reviews the anatomy of the lower leg, with particular attention to those muscles considered in the thesis. It presents the fundamental notions of muscle force generation, the correlation with muscle architectural parameters, and the imaging technique used in this study to obtain such parameters. It describes the origin of adipose tissue in skeletal muscles. In the end, it gives an overview of the different methods for fat quantification.

A2 Lower leg anatomy

The lower leg lies between the knee joint and the ankle joint. Here, 14 muscles are

divided into three compartments, namely anterior, lateral and posterior, separeted

by intermuscular septa and the interosseous membrane between tibia and fibula.

A ^PPENDIX A S TATE OF THE ART

The posterior compartment is subdivided into superficial and deep compartment (Figure A1).

The muscles in the posterior compartment mainly act to plantar-flex and invert the foot. Plantar-flexion consists in pointing the foot downwards while inversion consists in tilting the sole of the foot inwards. The opposite movements are called, respectively, dorsi-flexion and eversion.

The gastrocnemius (Figure A2) is included in the superficial compartment. It has a medial and a lateral head which upperly originate, respectively, on the medial and lateral condyles of the femur. Distally, it inserts onto the calcaneus via the Achilles tendon.

Figure A1: Axial view of the lower leg compartments. Figure adapted from [42].

The soleus (Figure A2) is an ankle plantar flexor which lies underneath the gas- trocnemius. It originates on the proximal ends of the tibia and fibula and it joins the Achilles tendon to insert onto the calcaneus. Many studies on cadavers and magnetic resonance images indicate that the soleus is subdivided in four compart- ments: medial-anterior, lateral-anterior, medial-posterior and lateral-posterior [3]

(Figure A2).

The tibialis posterior (Figure A3) is included in the deep posterior compartment.

It originates between the tibia and the fibula on the interosseous membrane. Dis- tally, it inserts on the navicular and cuneiform bones. It does not only plantar flex and invert the foot but it also supports the medial arch of the foot.

The muscles of the anterior compartment mainly act to dorsi-flex the foot. In par-

ticular, the tibialis anterior (Figure A3) attaches on the lateral side of the tibia and

A ^PPENDIX A S TATE OF THE ART

on the adjacent interosseous membrane. Its tendon runs down the front of the leg and inserts medially on the foot, providing support to the arch.

Figure A2: The gastrocnemius, the soleus and its subcompartments. Figure adapted from Fig.1 in [43].

Figure A3: The tibialis posterior and anterior.

A ^PPENDIX A S TATE OF THE ART

A3 Force generation in skeletal muscles

Skeletal muscles contribute to locomotion, joint stability and body positioning.

Therefore, it is crucial to comprehend the force generation process, especially the mechanism of muscle contraction. Multiple factors influence force generation in- cluding fiber type and neural activation [27, 29, 44, 45], age [4, 27, 40, 44, 45], sex [4, 46] and muscle architecture [29, 44]. For the purposes of the thesis, only the latter will be discussed.

A3.1 Muscle contraction

Muscle contraction is produced by an orderly sequence of electrical and chemical events, beginning with an action potential (AP), transmitted from the motor neu- ron, that reaches the neuromuscular junction (Figure A4.a). Here, a synapse takes place, the AP spreads into the muscle membrane and calcium ions are released into the myofibrils. As a result, the interaction between the actin and myosin filaments leads to sarcomeres’ contraction (Figure A4.b).

Figure A4: Schematics of a) the trasmission of the AP from the motor neuron to the muscle fiber and b) the sarcomere’s contraction.

A3.2 Muscle architecture

Muscle architecture determines muscles’ mechanical function. It is defined as the arrangement of fibers within the muscle either in a parallel or pennation pattern [29] where fibers run at an angle relative to the line of pull of the muscle [44].

Force generation depends on such pattern because it determines fascicle length

A ^PPENDIX A S TATE OF THE ART

The pennation angle (PA) (Figure A5.a) is the angle made by the fascicles and the line of action of the muscle [44] represented by the external tendon or the aponeurosis [47].

The PCSA represents the sum of the cross-sectional areas of all the muscle fibers within the muscle [48] (Figure A5.b) and it is calculated as:

P CSA = V

F L cos(P A) (2)

In Equation 1, V is the muscle volume.

Figure A5: Schematics of PA (a) and PCSA (b).

It is believed that CSA is proportional to the maximal muscle force [7, 48], thus it represents the most relevant index to estimate the maximal force production capacity [5]. However, several studies suggested that this relation may not al- ways be consistent because of age, different training condition and sex. Castro et al. [8] reported higher force to CSA ratio (F/CSA) in trained young males and females than untrained ones. Maughan et al. [9] assessed F/CSA in leg extensors in elite male sprinters, elite endurance runners and an untrained control group.

Results revealed that sprinters were stronger than endurance subjects but no signif-

icant differences in CSA were observed. Furthermore, the control group exhibited

greater individual F/CSA ratios than the athletes. Martel et al. [10] investigated

changes in individual muscle fibre types corresponding to strength training pro-

grammes. Whereas strength increased by 29% and 34% for men and women,

respectively, the percentage of fibre types that made up total muscle volume as

a result of strength training programmes varied between the sexes [7, 9]. The

A ^PPENDIX A S TATE OF THE ART

lengths and velocities of muscle fibers have a dramatic effect on muscle force gen- eration [49]. Mechanical studies of single fibers isolated from the tibialis anterior of frogs reported that maximum contraction velocity is directly proportional to fiber length, as is the width of the isometric length–tension relationship [48, 50]

(Figure A6). However, it is not yet fully understood how much length-tension and force–velocity properties affect force generation during locomotion [49]. There have been reported differences in fiber length due to age [4, 6] and sex [5].

Figure A6: Schematics of the isometric length-tension relationship. Shortened muscle fibers cannot generate tension (A) because the myofilaments have exceeded their over- lapping capability and fewer cross bridges can be formed. The greatest tension is generated at resting length (L

) (B) because of the maximum overlap of myofila- ments and maximal number of cross-bridges. During elongation (C), cross-bridges are pulled apart thus passive tension accounts for most of the force generation.

Lastly, PA influences muscle force since the cosine in Equation 1 normalizes fiber angulation to the line of action of the muscle [51]. PA can vary from 0

to 30

[47]

but it can increase during contraction up to 90

when the force becomes zero. By

means of real-time ultrasonography, it is possible to noninvasively measure the

changes in PA. Agaard et al [52] observed a positive relationship between PA and

muscle volume in response to resistance training. Ichinose et al. [53] reported

that differences in the PA of olympic athletes are very event-related and reflect

differences in muscle thickness. Maxwell et al. [54] observed that hypertrophy

must accompany increments in PA and thickness in muscles with constant muscle

length, fiber length and fiber number. Consequently, not only more contractile

tissue would be attached to the tendon but also muscle volume and PCSA would

increase [51]. Some studies reported differences in PA due to sex [4] and age [6].

A ^PPENDIX A S TATE OF THE ART

A4 In vivo imaging technologies to measure mor- phological parameters

A4.1 Diffusion tensor imaging

Diffusion tensor imaging (DTI) is a magnetic resonance technique that provides a measurement of the extent and direction of diffusion of water molecules [55]

in a biological tissue. Used for long time to reconstruct brain neuroanatomy, this technique has been recently applied to characterize muscle architecture in terms of parameters, such as fractional anisotropy (FA), primary, secondary, and tertiary eigenvectors. The diffusion properties allow the measurement of 3D fascicle length and fiber orientation in skeletal muscles. In particular, fiber orientation measure- ments are based on the principle that water diffusion occurs primarily in the axial direction of muscle fibers because radial direction diffusion might be obstructed by cell membranes and intracellular obstructions [55] Previous studies evaluated the repeatability of muscle DTI [22, 56] and attempted to validate DTI measurements of muscle architecture through direct comparison with cadaver measurements [3].

A4.1.1 Fiber tracking

Fibre tracking is the procedure of generating curves that, starting from a seed point, follow the primary direction of diffusion bi-directionally through a DTI scan volume [3]. In muscle fiber tracking, these curves follow the fibre orientation throughout a muscle and, when appropriate stopping criteria are defined, they resemble muscle fibres [3]. This approach is possible thanks to the anisotropic diffusion of water within muscle tissue, and it is accomplished by combining prin- cipal eigenvector information for consecutive voxels. The fibers are tracked from starting points along the aponeurosis to the muscle border [22, 23].

DTI-based fiber tracking can be done in DSI-Studio, an open-source diffusion MRI

analysis tool. Here, DTI data is imported and diffusion tensors, eigenvalues, eigen-

vectors and FA maps can be extracted within the software using a deterministic

fiber-tracking algorithm [25]. The first step consists in choosing a seed point and

a tract is bi-directionally propagated in the direction of the primary eigenvector

until the tract enters a region with a FA value below a certain threshold or until

the angle between subsequent tract segments exceeds a certain value. Another cri-

terion ensures that the tract length is within a reasonable range and that at least

one tract point ends in the superficial muscle boundary region and another one in

the deep muscle boundary region [23].

A ^PPENDIX A S TATE OF THE ART

A5 Adipose tissue in skeletal muscles

Adipose tissue is present within many organs and tissues in the body. In particular, the storage of adypocites underneath the deep fascia of muscles, called fat infiltra- tion, represents a potential contributor to functional decline of skeletal muscles.

For instance, it may alter muscle fiber orientation and hence the force producing capabilities of the whole muscle [26].

Fat infiltration includes adipose tissue located between muscle fibers (intramuscu- lar fat) [2,26–30] and also between muscle groups (intermuscular fat) [14,26,28, 29] (Figure A7). Only intramuscular fat is investigated in this thesis.

Figure A7: Intramuscular and intermuscular fat. Figure adapted from Fig.1 in [41].

A5.1 Origin of fat infiltration

Muscle fibers are surrounded by multipotential cells of mesenchymal origin called fibro-adipogenic progenitors (FAPs) [57], and a population of stem cells located adjacent to the plasma membrane of myofibers [28] called satellite cells (SCs) (Figure A8.a). Fat infiltration arises through two different pathways:

1. Direct accumulation of intramuscular fat (Figure A8.b). In particular, accu- mulation of the sphingolipid ceramide appears to have a particularly detri- mental effect on skeletal muscle function [57, 58].

2. Accumulation of intermuscular fat (Figure A8.c). In fact, under condition of

muscle injury, FAPs readily differentiate into adipocytes [57].

A ^PPENDIX A S TATE OF THE ART

Figure A8: Cells populations in skeletal muscles and their association to fat infiltration. (a) Muscle fibers (pink) are surrounded by SCs (blue) and FAPs (green). (b) Inter- muscular fat (IMF) accumulates within muscle fybers. (c) FAPs can differentiate to adypocytes (ACs) contributing to the accumulation of intermuscular fat. Figure reprinted from [57], p. 2, with permission from Elsevier.

A5.2 Fat quantification methods

Muscle composition represents the link between muscle anatomy and function.

Advances in medical imaging provided non invasive methods for quantifying mus- cle composition that overcome limitations from biopsy, the reference standard to assess fat infiltration in skeletal muscles.

A5.2.1 Muscle biopsy

Historically, fat infiltration in skeletal muscles has been measured through biopsy (Figure A9). Apart from invasiveness and pain, the possibility to explore only few muscular sites and obtain small samples is very limiting. Thus, a biopsy may not be truly representative in case of heterogeneous distribution of disease [59].

Figure A9: Example of muscle biopsy. After skin incision, (a) fascia and (b) muscle are directly

exposed to ensure clear visualization of the muscle and precisely draw the (c)

selected part of the muscle. Figure adapted from Fig.1 in [59].

A ^PPENDIX A S TATE OF THE ART

A5.2.2 T1-weighted images

According to [2] and [60], T1-weighted imaging accurately quantifies intermus- cular fat but not intramuscular fat. Burakiewicz et al. [11] argues that the signal intensity does not directly quantify changes in muscle fat content and needs to be referenced to the subjects’ bone marrow intensity of the same image section and expressed as a percentage of the bone marrow signal intensity. Moreover, this approach may be sensitive to B1 and B0 inhomogeneities [11].

A5.2.3 Magnetic Resonance Spectroscopy

In magnetic resonance spectroscopy (MRS), the water and fat signal can be deter- mined by their exact position on the x-axis of the spectra. In particular, fat signals from human muscle change frequencies with the angle between the muscle and the magnetic field in the magnet. Thus, when the muscle fibres are arranged paral- lel to the magnetic field, the geometric arrangement of inter- and intramuscular fat leads to different magnetic resonance characteristics that can be used to separate them [27]. Thus, MRS can only be used to measure intermuscular fat in muscles, such as the tibialis anterior, that can be aligned in the MRS unit so that the fibers run parallel to the direction of the magnetic field [27]. Although it is non-invasive and it provides repeatable measurements, MRS is expensive, time consuming and can only be performed in a limited number of muscles [27].

A5.2.4 mDixon method

mDixon method is a water-fat separation technique which is gaining clinical inter- est in intramusuclar fat quantification thanks to the technological advances such as robust algorithms for reliable water–fat separation and powerful reconstruction hardware for their rapid execution [61].

Developed in 1984 [62], mDixon method exploits the chemical shift difference between fat and water to produce in-phase (IP) and out-of-phase (OP) images.

Summation of IP and OP images gives a pure water image while the subtraction of OP from IP images gives a pure fat image. In total, four images are generated in a single acquistion (Figure A10), resulting in a shorter examination time [62].

In the case of muscles, after segmentation, the fat and water signals (S

and S

respectively) can be used to calculate the fat fraction (FF) expressed as the fraction of fat signal in the total signal in each voxel [11]:

F F = S

S

+ S

(3)

A ^PPENDIX A S TATE OF THE ART

This method, called 2-point Dixon, is commonly used in morphological imaging for many reasons:

• Higher spatial resolution than MRS [15]

• Reliable combination with nearly every type of pulse sequence and signal- contrast weighting [63]

• Shorter acquisition time due to high signal-to-noise ratio (SNR) and parallel imaging [62, 64]

• Insensitivity to B1 heterogeneity [60]

However, this method is limited by sensitivity to inhomogeneities in the main mag-

netic field B0, leading to phase errors [60,64]. This complication can be accommo-

dated by acquiring a third image to calculate a phase correction algorithm. In this

way, the separation effectiveness increases, thereby avoiding fat–water swapping

artifacts that represent potential sources for processing failure [65]. This approach

is called 3-point Dixon method. However, the longer acquisition time can lead to

increased risk for motion and breathing artifacts. These can be partially mitigated

with a multi-transmit coil [60].

A ^PPENDIX A S TATE OF THE ART

Figure A10: Four images generated from a single sagittal mDixon sequence of the knee. a) IP image, b) water image, c) OP image, d) fat image. IP images are obtained with a time to echo at which fat and water protons have the same phase. OP images are obtained with a time to echo at which fat and water protons are 180

out of phase. Figure adapted from Fig.1 in [62].

A6 Summary

The present study is based on the research by Bolsterlee [37] and D’Souza [66].

The aim of the thesis is to provide reference values for architectural parameters

and intramuscular fat fraction of lower leg muscles in typically developing chil-

dren. The correlation between IFF and force generation capacity is also investi-

gated.

R ^EFERENCES

References

[1] Mingazzini, P. ”Leonardo e l’anatomia. Magnificamente Leonardo: spettacolo di narrazione sulla vita e le opere di Leonardo Da Vinci”. November 22, 2017. [Online]. Available at https://leonardodavinciteatro.wordpress.com/category/anatomia/. Ac- cessed: July 22, 2019.

[2] Noble et al., “Intramuscular fat in ambulant young adults with bilateral spas- tic cerebral palsy,” BMC Musculoskeletal Disorders, vol. 15, no. 236, pp. 1–8, 2014.

[3] Bolsterlee et al., “Three-dimensional architecture of the whole human soleus muscle in vivo,” PeerJ, vol. 6, no. 6, pp. 1–22, 2018.

[4] Wu et al., “Effects of age and sex on neuromuscular-mechanical determinants of muscle strength,” American Ageing Association, vol. 38, no. 57, pp. 1–12, 2016.

[5] Four´ e et al., “Diffusion properties and 3D architecture of human lower leg muscles assessed with ultra-high-field-strength diffusion-tensor mr imaging and tractography: reproducibility and sensitivity to sex difference and intra- muscular variability,” Radiology, vol. 287, no. 2, pp. 593–607, 2018.

[6] Stenroth et al., “Age-related differences in Achilles tendon properties and triceps surae muscle architecture in vivo,” Journal of Applied Physiology, vol.

113, pp. 1537–1544, 2012.

[7] Jones et al., “Cross-sectional area and muscular strength: A brief review,”

Sports Medicine, vol. 38, no. 12, pp. 987–994, 2008.

[8] Castro et al., “Peak torque per unit cross-sectional area differs between strength-trained and untrained young adults,” Medicine & Science in Sports

& Exercise, vol. 27, pp. 397–403, 1995.

[9] Maughan et al., “Relationships between muscle strength and muscle cross- sectional area in male sprinters and endurance runners,” European Journal of Applied Physiology, vol. 50, pp. 309–318, 1983.

[10] Martel et al., “Age and sex affect human muscle fibre adaptations to heavy- resistance strength training,” Experimental Physiology, vol. 91, no. 2, pp.

457–463, 2006.

R ^EFERENCES

[11] Burakiewicz et al., “Quantifying fat replacement of muscle by quantitative MRI in muscular dystrophy,” The Journal of Neuroscience, vol. 264, pp. 2053–

2067, 2017.

[12] Reiser, Hricak and Knauth, Magnetic Resonance Imaging of the Skeletal Mus- culature. Springer: Marc-Andr´ e Weber Editor, 2014.

[13] Keller et al., “Diffusion tensor imaging of dystrophic skeletal muscle: Com- parison of two segmentation methods adapted to chemical-shift-encoded water-fat MRI,” Clinical Neuroradiology, vol. 29, no. 2, pp. 231–242, 2018.

[14] Zoico et al., “Adipose tissue infiltration in skeletal muscle of healthy elderly men: relationships with body composition, insulin resistance, and inflam- mation at the systematic and tissue level,” Journal of Gerontology: MEDICAL SCIENCE, vol. 65A, no. 3, pp. 295–299, 2010.

[15] Grimm et al., “Repeatability of Dixon magnetic resonance imaging and mag- netic resonance spectroscopy for quantitative muscle fat assessments in the thigh,” Journal of Cachexia, Sarcopenia and Muscle, vol. 9, pp. 1093–1100, 2018.

[16] Triplett et al., “Chemical shift-based MRI to measure fat fractions in dys- trophic skeletal muscle,” Magnetic Resonance in Medicine, vol. 72, no. 8, pp.

8–19, 2014.

[17] Fischmann et al., “Exercise might bias skeletal-muscle fat fraction calculation from Dixon images,” Neuromuscular Disorders, vol. 22, pp. S107–S110, 2012.

[18] Schlaeger et al., “Association of paraspinal muscle water–fat MRI-based measurements with isometric strength measurements,” European Radiology, vol. 29, pp. 599–608, 2019.

[19] Oudeman et al., “Techniques and applications of skeletal muscle diffusion tensor imaging: A review,” Journal of Magnetic Resonance Imaging, vol. 43, pp. 773–778, 2016.

[20] Sinha et al., “Reproducibility analysis of diffusion tensor indices and fiber ar- chitecture of human calf muscles in vivo at 1.5 t in neutral and plantarflexed ankle positions at rest,” Journal of Magnetic Resonance Imaging, vol. 34, pp.

107–119, 2011.

[21] Fedorov et al., “3D Slicer as an image computing platform for the Quantita-

tive Imaging Network,” Magnetic Resonance Imaging, vol. 30, pp. 1323–1341,

R ^EFERENCES

[22] Bolsterlee et al., “Comparison of measurements of medial gastrocnemius ar- chitectural parameters from ultrasound and diffusion tensor images,” Jour- nal of Biomechanics, vol. 48, pp. 1133–1140, 2015.

[23] K¨ orting, C., “Determination of in vivo muscle architecture: Comparison of ul- trasound and diffusion tensor imaging and analysis of muscle morphology in post-stroke patients,” Degree project in Medical Engineering, KTH Royal Insti- tute of Technology, School of Engineering Sciences in Chemistry, Biotechnology and Health, pp. 1–21, 2018.

[24] Manj´ on Herrera et al., “Diffusion Weighted Image Denoising using overcom- plete Local PCA,” PLoS ONE, vol. 8, no. 9, pp. 1–12, 2013.

[25] Yeh et al., “Deterministic diffusion fiber tracking improved by quantitative anisotropy,” PLoS ONE, vol. 8, no. 11, pp. 1–16, 2013.

[26] Marcus et al., “Intramuscular adipose tissue, sarcopenia, and mobility func- tion in older individuals,” Journal of Aging Research, vol. 2012, 2012.

[27] Janssen et al., “Linking age-related changes in skeletal muscle mass and com- position with metabolism and disease,” The Journal of Nutrition, Health and Aging, vol. 9, no. 6, pp. 408–416, 2005.

[28] Vettor et al., “The origin of intermuscular adipose tissue and its pathophys- iological implications,” American Journal of Physiology-Endocrinology and Metabolism, vol. 297, pp. 987–998, 2009.

[29] McGregor et al., “It is not just muscle mass: a review of muscle quality, com- position and metabolism during ageing as determinants of muscle function and mobility in later life,” Longevity and Healthspan, vol. 3, no. 9, pp. 1–8, 2014.

[30] Tuttle et al., “Intermuscular adipose tissue is muscle specific and associated with poor functional performance,” Journal of Aging Research, vol. 2012, pp.

1–7, 2012.

[31] Ross et al., “Comparison of three different methods to analyze ankle plan- tarflexor stiffness in children with spastic diplegia cerebral palsy,” Archives of Physical Medicine and Rehabilitation, vol. 92, no. 12, pp. 2034–2040, 2011.

[32] Taylor et al., “Test–retest reliability of hand-held dynamometric strength test-

ing in young people with cerebral palsy,” Archives of Physical Medicine and

Rehabilitation, vol. 85, pp. 77–80, 2004.

R ^EFERENCES

[33] Berry et al., “Intrasession and intersession reliability of handheld dynamom- etry in children with cerebral palsy,” Pediatric Physical Therapy, vol. 16, pp.

191–198, 2004.

[34] Lidbeck et al., “Muscle strength does not explain standing ability in children with bilateral spastic cerebral palsy: a cross sectional descriptive study,” BMC Neurology, vol. 15, no. 188, pp. 1–7, 2015.

[35] Mukaka M., “Statistics corner: A guide to appropriate use of correlation co- efficient in medical research,” Malawi Medical Journal, vol. 24, no. 3, pp.

69–71, 2012.

[36] D’Souza et al., “Muscle architecture in children with cerebral palsy and an- kle contractures: an investigation using diffusion tensor imaging,” Clinical Biomechanics, vol. 68, pp. 205–211, 2018.

[37] Bolsterlee et al., “Reliability and robustness of muscle architecture mea- surements obtained using diffusion tensor imaging with anatomically con- strained tractography,” Journal of Biomechanics, vol. 86, pp. 71–78, 2019.

[38] Baum et al., “Association of quadriceps muscle fat with isometric strength measurements in healthy males using chemical shift encoding-based water- fat magnetic resonance imaging,” Journal of Computer Assisted Tomography, vol. 40, no. 3, pp. 447–451, 2016.

[39] Marcon et al., “Normative values for volume and fat content of the hip ab- ductor muscles and their dependence on side, age and gender in a healthy population,” Skeletal Radiology, vol. 45, pp. 465–474, 2016.

[40] Delmonico et al., “Longitudinal study of muscle strength, quality, and adi- pose tissue infiltration,” The American Journal of Clinical Nutrition, vol. 90, pp. 1579–1585, 2009.

[41] Addison et al., “Intermuscular fat: A review of the consequences and causes,”

International Journal of Endocrinology, vol. 2014, pp. 1–11, 2014.

[42] ”Lower Leg Compartment”. Wikem, The Global Emergency Medicine Wiki.

July 31, 2016. [Online]. Available at https://wikem.org/wiki/Compartment- syndrome. Accessed: March 10, 2019.

[43] ”Muscles in Posterior Compartment of the Leg”. Teach me Anatomy. Septem-

ber 28, 2018. [Online]. Available at https://teachmeanatomy.info/lower-

limb/muscles/leg/posterior-compartment/. Accessed March 10, 2019.