The digital human forearm and hand

Abstract

How changes in anatomy affect joint biomechanics can be studied using musculoskeletal modelling, making it a valuable tool to explore joint function in healthy and pathological joints. However, gathering the anatomical, geometrical and physiological data necessary to create a model can be challenging. Very few integrated datasets exist and even less raw data is openly available to create new models. Therefore, the goal of the present study is to create an integrated digital forearm and make the raw data available via an open‐access database. An un‐embalmed cadaveric arm was digitized using 7T MRI and CT scans. 3D geometrical models of bones, cartilage, muscle and muscle pathways were created. After MRI and CT scanning, physiological muscle parameters (e.g. muscle volume, mass, length, pennation angle, physiological cross‐sectional area, tendon length) were obtained via detailed dissection. After dissection, muscle biopsies were fixated and confocal microscopy was used to visualize and measure sarcomere lengths. This study provides an integrated anatomical dataset on which complete and accurate musculoskeletal models of the hand can be based. By creating a 3D digital human forearm, including all relevant anatomical parameters, a more realistic musculoskeletal model can be created. Furthermore, open access to the anatomical dataset makes it possible for other researchers to use these data in the development of a musculoskeletal model of the hand.

Introduction

Variations in tendon attachments, muscle volume and 3D articular surface geometry will influence how bones move and how joints are loaded. How changes in anatomy affect joint biomechanics can be studied using musculoskeletal modelling, making it a valuable tool to explore joint function in healthy and pathological joints. In fundamental research, models can be used to gather insights into hand function, motion control and cartilage deformation (Lieber & Fridén, 2001). In a clinical setting, musculoskeletal models can be used for pre‐operative planning (Dvinskikh et al. 2011; Galarraga et al. 2017) and prosthesis development (Gustus et al. 2012; Wolf et al. 2004).

Although sophisticated musculoskeletal models of the knee (Beidokhti et al. 2016), hip (Blanco & Gambini, 2006; Blemker & Delp, 2005; Wesseling et al. 2015) and shoulder (Nikooyan et al. 2011; Seth et al. 2016; Wu et al. 2016) have been developed in the past, a complete model of the hand is still lacking. One of the main reasons for this is the high complexity of the human hand. Within the small space of the hand, many different intrinsic hand muscles influence the motion of a multitude of small joints. In addition, the extrinsic hand muscles drive wrist, (meta)carpal and phalangeal movement. The small size of the bones and muscles, combined with the complex joint motions and the immense quantity of degrees of freedom, make biomechanical analysis, and in particular modelling, of the hand difficult.

To keep the complexity of the model manageable, models of a single finger have been developed since the 1970s. Cooney and colleagues calculated the forces on the base of the thumb when a load was applied at the tip of the thumb using a rudimentary thumb model (Cooney & Chao, 1977). Over the years, models of the wrist (Fischli et al. 2009; Iwasaki et al. 1998; Wayne & Mir, 2015), fingers (Goislard de Monsabert et al. 2014; Stillfried & der van Smagt, 2010; Valero‐Cuevas et al. 1998; Vigouroux et al. 2007) and the thumb (Valero‐Cuevas et al. 2003; Vigouroux et al. 2011, 2009; Wu et al. 2009) have gradually gained in complexity. Additionally, wrist models have focused on the carpal bones. More complete hand models have also been introduced but often did not include accurate anatomical data. Physiological muscle parameters were lacking (An et al. 1979), the intrinsic hand muscles were omitted (Holzbaur et al. 2007), embalmed specimens (which influences muscle characteristics) were used or joint geometry and muscle pathways or physiological parameters were obtained from different specimens (Buchholz & Armstrong, 1992; Sancho‐Bru et al. 2014). More recently, an effort was made to create an integrated model with morphological parameters from a single specimen (Gustus et al. 2012; Mirakhorlo et al. 2016). However, the specimen used from an 87‐year‐old female and 3D geometrical data of the bones and cartilage were not included.

We want to expand the current biomechanical knowledge of the hand and thumb by further increasing the integration of different anatomical datasets, as such providing the building blocks for a more lifelike musculoskeletal model of the hand. With the increased availability of high Tesla magnetic resonance imaging (MRI) scanners that have a bore large enough to fit an entire forearm and hand, medical imaging may provide a way to quantify muscle pathways in a way that better represents the in vivo situation. By combining the MRI data with computed tomography (CT) images, 3D bone and cartilage geometry can be accurately obtained. This makes it possible to calculate the inertial properties of bone segments instead of estimating these by fitting cylinders on the segments. In addition, the unavailability of raw data used to build existing musculoskeletal models forms a major restriction for the development of new models. A freely accessible database with precise anatomical data of muscle pathways, physiological muscle parameters and the 3D geometry of bone and cartilage could be useful for all researchers interested in the modelling of hand function. Therefore, the goal of the present study is to create an integrated digital forearm and make the raw anatomical data available via an open‐access database.

Materials and methods

Specimen details

Through the voluntary human body donation programme of the university, an un‐embalmed left forearm with hand from a 60‐year‐old man was acquired. The arm was segmented at the humerus, well above the origin of the wrist extensor muscles and stored at −20°C.

Medical imaging protocols

After thawing the specimen at room temperature, the arm was placed in a thermoplastic cast and three MRI‐compatible nifedipine capsules (Adalat 10, Bayer Vital, Cologne, Germany) were placed onto the skin of the proximal forearm, distal forearm and wrist (Fig. 1). A surgical marker was used to delineate the position of the capsules onto the skin of the specimen. The cast ensured a standardization of the position of the hand, wrist and elbow between different MRI and computed tomography (CT) scans. The capsules were used as extra reference points when subsequent MRIs were fused in a single large image.

Figure 1.

Anterior view of the thawed, fresh‐frozen arm placed in a thermoplastic cast. The yellow capsules can be seen at the proximal forearm, distal forearm and wrist.

The in situ soft tissue geometry was visualized by a 7 Tesla MRI scanner with a bore diameter of 60 cm and a magnet length of 270 cm (MAGNETOM 7T, Siemens Healthcare, Erlangen, Germany) with a custom‐made coil (Erwin L. Hahn Institute for MRI, Essen, Germany). A 3D FL3D1 sequence was used, with the following specifications: Repetition Time = 10 s, Echo Time = 3.58 s, flip angle = 25°, slice thickness = 0.38 mm and pixel size = 0.38 × 0.38 mm. Due to the limited coil size, three subsequent acquisitions were needed to scan the entire specimen.

To obtain the 3D geometry of the forearm and hand bones, a CT scan of the entire specimen was made using a 64‐slice Discovery HD 750 CT scanner (GE Healthcare, Little Chalfont, UK) with the following settings: display field of view (DFOV): 250 mm, slice thickness: 0.625 mm, voxel size: 0.150 mm³, 100 kV, 180 mA.

Image processing

Image processing software (mimics 19, Materialise, Leuven, Belgium) was used to combine the three MRI scans in a single MRI of the entire forearm and hand. This software was also used to combine the MRI scan with the CT scan. Next, 3D models of the muscle volumes, muscle pathways, bones and cartilage were created through manual segmentation of the MRI and CT images. The ‘optimal’ reconstruction setting in mimics was used for all 3D reconstructions. These models were exported in a standard computer‐aided‐design file format: American Standard Code for Information Interchange STereoLithography (ASCII STL). Volumes of the muscle and bone models was determined using an anatomical computer‐aided design programme (3‐matic 19, Materialise, Leuven, Belgium).

Physiological muscle parameters

After CT and MRI scanning, a stepwise dissection of the entire cadaveric forearm and hand was performed. Each muscle tendon unit (MTU) was isolated and the following measurements were taken: muscle mass, MTU length, muscle length, fibre length, pennation angle, muscle volume, external tendon length and internal tendon length (Channon et al. 2009). Distances were measured with digital callipers (ABS Coolant Proof Caliper, Mitutoyo, Kawasaki, Japan). Muscle mass was determined up to 1 g using a digital scale (Precision Balance PC 4500/C2, Radwag, Radom, Poland). To prevent shrinkage during the dissection due to dehydration, the specimen was kept moist using a 0.9% saline solution (Baxter, Deerfield, IL, USA).

To determine pennation angle of each muscle, the muscle belly was cut lengthwise and photographs were taken perpendicular to the muscle using a tripod. Photo processing software (imagej, NIH, Bethesda, MD, USA) was used to determine the pennation angle at three places within the muscle belly where the angulation of the fascicles relative to the internal tendon was clearly visible (Fig. 2). These three values were averaged to give a single pennation angle per muscle.

Figure 2.

Measurement of pennation angle (α) relative to the internal tendon.

Muscle volume was determined as the volume of saline solution displaced by submersion of the muscle belly. However, for the smallest intrinsic muscles, the displacement of the saline solution was too small (> 1 mL) to be determined accurately. For these muscles, muscle volume was calculated by multiplying the muscle mass by the average muscle density (as obtained for the larger hand and forearm muscles).

After volume measurement, biopsies were cut from each muscle belly, measured with digital callipers and placed in 4% formaldehyde solution. After 14 days, the formaldehyde solution was removed and the biopsies size were re‐measured to determine the amount of shrinkage. Each muscle biopsy was placed in an incubator at room temperature for 24 h with a blocking buffer [phosphate‐buffered saline (PBS) 1×, 0.2% Triton‐X, 1% bovine serum albumin (BSA)]. After 24 h, the blocking buffer was removed and the tissue was stained using a 1/10 000 4’,6‐diamidino‐2‐phenylindole, dihydrochloride (DAPI)/PBS solution for 72 h at 4°C. After washing the tissue samples with PBS, confocal microscopy (Zeiss Observer.Z1, Zeiss International, Oberkochen, Germany) was used to visualize sarcomeres. The software imagej (Rueden et al. 2017) was used to measure the average sarcomere length over 10 sarcomeres (Fig. 3).

Figure 3.

Visualization of sarcomeres after DAPI staining under confocal microscopy.

The average sarcomere length was adjusted for shrinkage due to the formaldehyde solution and used to compute optimal muscle fascicle lengths using the following formula (Burkholder et al. 2001; Walker & Schrodt, 1974) with a normalized optimal fascicle length of 2.6 μm:

$$\begin{array}{cc}& \text{Optimal muscle fascicle length}=\frac{\text{average fascicle length}}{\text{sacromere}}\hfill \\ & \times \text{normalized optimal fascicle length}\hfill \end{array}$$

This optimal muscle fascicle length was used to calculate the physiological cross‐sectional area (PCSA):

$$\text{PCSA}=\frac{\text{muscle mass}\times \text{cosine}(\text{pennation angle})}{\text{muscle density}\times \text{optimal muscle fascicle length}}$$

The PCSA is directly related to the force‐generating capacity of the muscle (i.e. the maximal isometric force) (Brand et al. 1986).

Results

Figure 4 shows a visual impression of the digitized hand, based on the CT and MRI data. The DICOM files, STL surface meshes of the bones, muscles and cartilage can be downloaded at http://www.morphosource.org under project number P419, ‘The integrated digital human forearm and hand’.

Figure 4.

(a) 3D model of intrinsic hand muscle and metacarpal bones. Dorsal view of MC V (left) until MC II (right). In red are shown (from left to right) m. abductor digiti minimi, m. interosseous dorsalis IV till II. (b) 3D impression of the extensor muscles of the wrist (dorsal view).

All measured and calculated muscle parameters can be found in Table 1.

Table 1.

The measured and calculated muscle parameters for optimal fibre length, muscle mass, muscle volume, pennation angle and PCSA

Muscle name		Average fiber length (mm)	Muscle belly mass (g)	Muscle belly length (mm)	Muscle volume (mm3)	Average pennation angle (°)	PCSA (mm²)
Wrist
Extensor carpi radialis longus	ECRLb	160	74	12	82	12	483
Extensor carpi radialis brevis	ECRBb	160	74	199	82	12	483
Extensor carpi ulnaris	ECU	45	27	226	25	14	620
Flexor carpi radialis	FCR	63	25	99	24	16	358
Flexor carpi ulnaris	FCU	99	33	259	34	20	321
Palmaris longus	PL	a	a	a	a	a	a
Palmaris brevis	PB	28	1	28		0	32
Pronator teres	PT	70	42	137	37	24	529
Pronator quadratus	PQ	39	11	49	12	0	308
Supinator	SUP	76	18	19	18	0	184
Brachioradialis	BR	263	46	277	45	0	149
Anconeus	ANC	42	13	113	125	31	256
Thumb
Abductor pollicis longus	APL	171	18	177	7	0	33
Abductor pollicis brevis	APB	58	7	6	6	0	92
Extensor pollicis longus	EPL	62	11	161	15	8	149
Extensor pollicis brevis	EPB	58	2	84	15	0	24
Opponens pollicis	OPP	51	5	6	45	0	92
Adductor pollicis	ADP		17		17
	Transversum	52		52		0
	Obliquum	51		6		0
Flexor pollicis brevis	FPB
	Superficiale	56	3	56	3	0	56
	Profundum	40	0	4	5	0	9
Flexor pollicis longus	FPL	76	25	64	33	14	395
Index finger
Extensor indicis	EI	90	6	127	5	0	53
Extensor digitorum II	ED	66	57		46	22	665
	II			279
Flexor digitorum superficialis II	FDS		89		94	0	1142
	II	96		287
Flexor digitorum profundus II	FDP		110		116	14	935
	II			254
Lumbricalis I	LUMB						46
	I	82		82		0
Palmar interosseus II	IOP				7		178
	II/III	60	2	38		0
	IOD
Dorsal interosseus I MC1	FDI (IOD I/II)	46	9	56	9	0	196
Dorsal interosseus I MC2	II/III	80	4	66
Middle finger
Extensor digitorum III+IV	ED	66	57		46	22	665
	III			235
	IV			253
Flexor digitorum superficialis III	FDS		89		94		1142
	III	58		266
Flexor digitorum profundus III	FDP		110				935
	III			271
Lumbricalis II	LUMB						46
	II	89		89		0
	IOD						196
Dorsal interosseus II	II/III	80	4	66
Dorsal interosseus III	III/IV	63	4	53
Ring finger
Extensor digitorum IV	ED	66	57		46	22	665
	IV			253
Flexor digitorum superficialis IV	FDS		89		94		1142
	IV	76		229
Flexor digitorum profundus IV	FDP IV		110 223
Lumbricalis III	LUMB						46
	III	86		86		0
Palmar interosseus III	IOP				7		178
	III/IV	58	2	33		0
Dorsal interosseus IV	IOD						196
	III/IV	63	4	53
Little finger
Extensor digitorum V	ED	66	57		46	22	665
	V			251
Extensor digiti minimi	EDM	251		251		0	0
Flexor digitorum superficialis IV+V	FDS		89		94	0	1142
	V	73		289
Flexor digitorum profundus IV+V	FDP		110				935
	V			253
Flexor digiti minimia	FDM
Opponens digiti minimi	ODM	44	3	44	3	0	67
Abductor digiti minimi	ADM	96	11	84	1	0	111
Lumbricalis IV	LUMB						46
	IV	40		4		0
Palmar interosseus V	IOP				7		178
	IV/V	69	3	56		0

^aMuscle was not present.

^bECRL and ERCB muscle bellies were fused.

The measured lengths of tendons and muscle‐tendon units can be found in Table 2.

Table 2.

The properties of the muscle tendon units: MTU length, muscle belly length, internal tendon lengths, external tendon lengths

Muscle name		MTU length (mm)	Origin Internal tendon length (mm)	Insertion Internal tendon length (mm)	Origin External tendon length (mm)	Insertion External tendon length (mm)
Wrist
Extensor carpi Radialis longus	ECRL	299		43		189
extensor carpi Radialis brevis	ECRB	322		75		123
Extensor carpi ulnaris	ECU	300	13	164	189	56
Flexor carpi radialis	FCR	210	46	147	13	99
Flexor carpi ulnaris	FCU	272		201		13
Palmaris longus	PLa
Palmaris brevis	PB	28
Pronator teres	PT	173	39	11	36
		173	117
Pronator Quadratus	PQ	27	16
		49	8
Supinator	SUP	109
Brachioradialis	BR	363		49		86
Anconeus	ANC	113	85
Thumb
Abductor pollicis longusb	APL	246		131		82
	APL	246		112		55
Abductor pollicis brevisc	APB	69		18		5
		62		18		5
Extensor pollicis longus	EPL	301	88	107		140
Extensor pollicis brevis	EPB	172	13	26		88
Opponens pollicis	OPP	51
Adductor pollicis	ADP
	Transversum	52
	Obliquum	51	24
Flexor pollicis Brevis	FPB
	Superficiale	56
	Profundum	48		22		8
Flexor pollicis longus	FPL	307			118	126
Index finger
Extensor indicis	EI	296		74		169
Extensor digitorum II	ED
	II	433		144		154
Flexor digitorum superficialis II	FDS
	II	418		253	5	126
Flexor digitorum profundus II	FDP
	II	425		135		171
Lumbricalis I	LUMB
	I	82
Palmar Interosseus II	IOP
	II/III	6				22
	IOD
Dorsal interosseus I	FDI (IOD I/II)	76				20
Middle finger
Extensor digitorum III+IV	ED
	III	438		137		203
	IV	445		134		192
Flexor digitorum superficialis III	FDS
	III	416		153	5	145
Flexor digitorum profundus III	FDP
	III	42		100		149
Lumbricalis II	LUMB
	II	89
	IOD
Dorsal interosseus II	II/III	8		36		14
Dorsal interosseus III	III/IV	63		29		1
Ring finger
Extensor digitorum IV	ED
	IV	445		134		192
Flexor digitorum superficialis IV	FDS
	IV	405		145	5	171
Flexor digitorum profundus IV	FDP
	IV	41		179		187
Lumbricalis III	LUMB
	III	86
Palmar interosseus III	IOP
	III/IV	58				25
Dorsal interosseus IV	IOD
	III/IV	63		29		1
Little finger
Extensor digitorum V	ED
	V	43		137		179
Extensor digiti minimi	EDM	43		137		179
Flexor digitorum superficialis IV+V	FDS
	V	40		215	5	106
Flexor digitorum profundus IV+V	FDP
	V	39		133		137
Flexor digiti minimi	FDM
Opponens digiti minimi	ODM	44
Abductor digiti minimi	ADM	96				11
Lumbricalis IV	LUMB
	IV	4
Palmar interosseus V	IOP
	IV/V	69				13

^aMuscle was not present.

^bAPL with two tendons.

^cABP with two tendons.

Cross‐checking

To determine the internal consistency of the dataset, the volume of the 3D models obtained from the MRI images are compared with the measurement obtained during dissection in Table 3. Additionally, bone volumes that were calculated from 3D CT images can be used, in addition to mass and mass distribution parameters, to calculate the inertial properties of the body segments (Table 4).

Table 3.

Muscle volumes of several intrinsic hand muscles determined by MRI (3D model) and by submersion during dissection

Muscle	Volume 3D model (mm3)	Volume by submersion (mm3)	Difference (%)
Abductor digiti minimi	8979	10 000	+10
Adductor pollicis	15 745	17 000	+13
Flexor & Abductor pollicis brevis	13 484	12 500	+10
Interosseus I	7854	9000	+11
Lumbricalis I–IV	4100	4000	+1
Opponens pollicis	4292	4500	+2

Table 4.

The volumes of the bones in the hand and forearm (without cartilage). These values can be used, in addition to mass and mass distribution parameters, to calculate the inertial properties of the body segments

Bone	Volume (mm3)
Radius	54 305
Ulna	56 558
Scaphoideum	2857
Lunatum	2562
Triquetrum	2007
Pisiforme	1185
Trapezium	2782
Trapezoideum	1415
Capitatum	3938
Hamatum	3347
MC1	3606
MC2	9108
MC3	9248
MC4	5882
MC5	5844
Prox. Phal1	3354
Prox. Phal2	8471
Prox Phal3	5435
Prox Phal4	5470
Prox Phal5	5435
Int. Phal2	887
Int. Phal3	2276
Int. Phal4	1448
Int. Phal5	1438
Dist. Phal1	331
Dist Phal2	835
Dist Phal3	848
Dist Phal4	539
Dist Phal5	536

These values can be used, in addition to mass and mass distribution parameters, to calculate the inertial properties of the body segments.

Discussion

The goal of this research was to create an integrated digital forearm and hand and to make the raw data available via an open‐access database. By combining high resolution MR and CT images with quantitative anatomical data obtained by dissection of the same un‐embalmed specimen, an integrated dataset was created. This study provides quantitative anatomical information on the structures within the human forearm and hand that can be used to create complete and realistic musculoskeletal models of the forearm and/or hand.

In the current study, anatomical data were collected from one specimen. This is both a strength and a weakness. Gathering all the geometrical and physiological information from a single specimen provides a realistic representation of this single specimen. However, this single specimen is not necessarily representative of all human forearms and hands. There is a well‐known wide variety of anatomy between humans. This large diversity makes it difficult to create a dataset that captures most anatomical variations. More importantly, larger datasets are not by definition better. The wide anatomical variety would cause the creation of an ‘average’ forearm that does not represent a realistic forearm or hand. A recent study of Goislard de Monsabert and colleagues highlights the problems with mixing datasets in a single model. They found differences in the load‐sharing of wrist muscles of up to 180% when combining several incomplete datasets vs. the same model using one complete dataset (Goislard De Monsabert et al. 2018). Ideally, the current parameters can be used as basis for a subject‐specific model. However, care has to be taken when simply scaling a model, in particular because scaling a generic model simply by bone size is probably not very accurate. Unpublished data from seven dissections in which we compared radius length and muscle length for multiple forearm muscles showed no correlation between the two (see Supporting information). For the comparison of anatomical datasets as well as scaling models, a muscle volume (for in vivo use) or muscle mass (ex vivo) percentage appears to be more suitable due to the well‐established correlation between muscle volume and isometric joint‐moment generating capacity (Bolsterlee et al. 2015; Fukunaga et al. 2001; Vidt et al. 2012). By expressing the volume of a muscle as a percentage of the total muscle volume of the forearm and hand, different datasets can be compared. For example, the flexor carpi radialis made up 4.5% of total forearm muscle mass in our study and 4.8% in the research of Mirakhorlo and colleagues, this despite the absolute muscle mass being 2.5 times higher in the latter (Mirakhorlo et al. 2016).

When comparing the obtained data with existing anatomical data, the most striking difference is that for several muscles we provide a pennation angle of 0°. This is probably due to how pennation angle is defined. When a muscle belly is cut lengthwise following the insertion tendon and is folded open, most muscles show muscle fibres attached at an angle to the internal tendon due to the flattening of the cylindrical shape of the muscles. However, this does not mean that in situ the muscle is a multipennate muscle. In addition to the difficulties of measuring pennation angles, the relevance of the pennation angle for musculoskeletal modelling is debatable as well. During muscle contraction, the pennation angle of the contracting muscle will change as can be observed (and quantified) using ultrasound (Zhou et al. 2015, 2012). However, ex vivo, pennation angles can only be measured on isolated muscle tissue during dissection, which does not always correlate well to the in vivo configuration. Even though pennation angle is used in the calculation of PCSA, its influence is relatively minor (i.e. multiplication by cosine of the pennation angle). In our study, the pennation angle varied between 0° and 31°, resulting in a factor of 1–0.86 to be included in the calculation of PCSA. Considering the above‐mentioned concerns, the pennation angle seems best suited to describe whether a certain muscle has pennate or parallel fibres.

A caveat concerning the use of models based on literature values is the need for validation. When using the data presented here, the testing of moment arms or forces by loading tendons and measuring pressure is, indeed, not possible. Although the scaling of generic models from, for example, the AnyBody datasets is common practice, validation by biomechanical testing of the adapted model remains essential (e.g. by high density grid‐based electromyography data from living subjects, van Beek et al. 2018).

To further increase the detail of this dataset, future work could use geometrical muscle parameters to estimate PCSA within different parts of the same muscle (Lee et al. 2012). Additionally, reconstructing muscle pathways while the hand is in different functional positions would provide information about the changes in moment arms during movement. Such changes were not taken into account in the current dataset. However, considering the amount of work required to collect muscle moment arm, a sensitivity analysis on a small scale would be a useful next step.

Conclusion

By creating a digital human forearm and hand, with relevant physiological parameters, a more realistic musculoskeletal model of the forearm and hand can be created. Furthermore, due to the open nature of this study, anyone who wants to make a model of the hand, can use this dataset to create a complete hand model. Re‐inventing the wheel or combining outdated datasets is no longer necessary. This will lead to an increase in the development of musculoskeletal models that will provide insights into joint functioning in healthy and pathological joints or can be used for preoperative planning and prosthesis development.

Author contributions

Faes Kerkhof contributed to the experimental design, data collection, data analyses, data interpretation and writing, revising and editing the manuscript. Timo van Leeuwen contributed to the experimental design, data collection, data analyses, data interpretation and writing of the manuscript. Evie Vereecke contributed to the experimental design, data interpretation and writing, revising and editing of the manuscript.

Supporting information

Figure S1. Radius length vs. forearm muscle belly length in seven un‐embalmed specimens.

Figure S2. Radius length vs. forearm muscle tendon unit length in seven un‐embalmed specimens.