technologies

Article
Design, Construction and Tests of a Low-Cost Myoelectric Thumb
Murat Ayvali, Inge Wickenkamp and Andrea Ehrmann *

Faculty of Engineering and Mathematics, Bielefeld University of Applied Sciences, 33619 Bielefeld, Germany;
[email protected] (M.A.); [email protected] (I.W.)
* Correspondence: [email protected]

Abstract: Myoelectric signals can be used to control prostheses or exoskeletons as well as robots, i.e.,
devices assisting the user or replacing a missing part of the body. A typical application of myoelectric
prostheses is the human hand. Here, the development of a low-cost myoelectric thumb is described,
which can either be used as an additional finger or as prosthesis. Combining 3D printing with
inexpensive sensors, electrodes, and electronics, the recent project offers the possibility to produce an
individualized myoelectric thumb at significantly lower costs than commercial myoelectric prostheses.
Alternatively, a second thumb may be supportive for people with special manual tasks. These
possibilities are discussed together with disadvantages of a second thumb and drawbacks of the
low-cost solution in terms of mechanical properties and wearing comfort. The study shows that
a low-cost customized myoelectric thumb can be produced in this way, but further research on
controlling the thumb as well as improving motorization are necessarily to make it fully usable for
daily tasks.

Keywords: myoelectric prosthesis; 3D printing; biomechatronics; 6th finger; 3rd thumb







Citation: Ayvali, M.; Wickenkamp, I.;

Ehrmann, A. Design, Construction 1. Introduction
and Tests of a Low-Cost Myoelectric Prostheses can be subdivided into body-powered, pneumatically powered, and elec-
Thumb. Technologies 2021, 9, 63. trically powered ones [1]. While body-powered devices, using the energy of the muscles of
https://doi.org/10.3390/ a human, are relatively inexpensive and lightweight, they often have other shapes than
technologies9030063 the original part of the body which they replace, and they can be complicated to use [2].
Electrically powered prostheses, on the other hand, are usually much more expensive
Academic Editor: Manoj Gupta
and heavier, but for many users easier to handle. Such prostheses can get their signals
by different body signals, amongst them by electromyographic (EMG) signals from the
Received: 24 August 2021
residual muscles [3,4]. In this way, there is a better possibility to mimic the natural human
Accepted: 2 September 2021
movement than with body-powered prostheses. While the best control is reached by an
Published: 3 September 2021
implanted sensor, collecting intramuscular myoelectric signals, usually surface electrodes
are applied on the residual limb as a non-invasive technique is often preferred [5,6]. The
Publisher’s Note: MDPI stays neutral
efficiency of surface electrodes can be improved by transferring nerves to the positions
with regard to jurisdictional claims in
published maps and institutional affil-
where the surface electrodes are to be placed [7,8]. Previous studies found, e.g., that using
iations.
myoelectric prostheses can help avoiding cortical reorganization related with phantom
limp pain [9].
Myoelectric control is especially often used for upper-limb prostheses [10–13]. An-
other important part of the body is the hand, which is also often replaced by myoelectric
prostheses [14–16]. While research on moving the fingers of a hand prosthesis indepen-
Copyright: © 2021 by the authors.
dently is ongoing [17,18], single finger prostheses are usually not controlled by myoelectric
Licensee MDPI, Basel, Switzerland.
signals. One of the reasons for this finding may be the high costs for myoelectric prostheses
This article is an open access article
distributed under the terms and
in the range of 4000–75,000 $ for upper-limb prostheses [19].
conditions of the Creative Commons
Especially in developing countries, low-cost myoelectric prostheses would be of
Attribution (CC BY) license (https://
large help for people who lost a limb. This is why some research groups concentrate
creativecommons.org/licenses/by/ on developing cost-effective myoelectric prosthetic devices, especially hands. Atique
4.0/). and Rabbani, e.g., developed a myoelectric prosthetic hand with only the thumb being

Technologies 2021, 9, 63. https://doi.org/10.3390/technologies9030063 https://www.mdpi.com/journal/technologies

Technologies 2021, 9, 63 2 of 12

controllable by myoelectric signals, taken by surface electrodes, in this way enabling

holding small objects [20]. Three-dimensional printed prostheses are partly commercially
available at lower cost, e.g., approx. 1500$ for the Mark V (Seoul, Korea) [21]. Nevertheless,
these prices are still high for people in many countries around the world. A highly
interesting initiative, supporting people or also animals who need new limbs, is the e-
NABLE online global community, delivering wrist-powered hands and elbow-driven arms
in different designs, but not custom-made, by open-source 3D printing designs [22].
Besides these and other low-cost prostheses, there are also attempts to train people
with neurological injuries or to extend the physical properties of non-handicapped people
by exoskeletons [23–25]. Regarding hands or fingers, there are a few studies and design
approaches investigating the effect of a third thumb, e.g., controlled by the feet [26],
allowing to give people extended motor skills and at the same time enabling neuro-scientific
research on neuroplasticity [27,28].
Going one step further, myoelectric prostheses have recently been investigated to be
controlled commonly by user intention and robotic automation to improve the prosthesis
motions [29]. The interaction between a human and a prosthetic hand is still the most
complicated task to be solved, especially due to the problematic or non-existing sensory
feedback of the prosthesis [30]. This is why some groups concentrate on evaluating the
effect of different kinds of control strategies, such as raw control with raw feedback or
filtered control with filtered feedback [31]. Other groups investigated tolerant pattern
recognition for the myoelectric prosthesis control to avoid reaction to signal deviations
due to untrained conditions [32] or the effect of different learning durations on grasp
patterns [33]. On the other hand, different ways of learning, e.g., with augmented reality of
game-based training, are recently investigated [34,35].
Here, we present a project combining some of the aforementioned aims. We developed
a myoelectric third thumb—in case of able-bodied people—and tested at which positions
of the hand it could be ideally added. This thumb could also be used as prosthesis for
people who lost their thumb. Low-cost solutions for scanning are discussed, leading to the
possibility to prepare custom-made prostheses, which can be 3D printed and controlled
with inexpensive electronics by myoelectric signals taken by surface electrodes. The study
shows that developing a low-cost customized myoelectric thumb is possible, but more
research is needed regarding the exact positioning of the thumb as well as partly improving
the motor and the material of the fishing line, actuating the thumb.

2. Materials and Methods

In the conception phase, a requirement specification for an extra thumb was developed,
containing the following points:
- Enabling holding round objects with diameter 60 mm;
- Custom-made hand part with good fit to desired position;
- Rotation of minimum 40◦ of the thumb;
- Maximum dimensions of length 80 mm and width 20 mm to make the extra thumb
similar to a normal one;
- No sharp edges, ergonomic construction;
- Using a spring to let the thumb relax into the neutral position;
- Maximum mass, including servomotor, of 50 g;
- Production by 3D printing;
- Low-cost materials and parts, ideally leading to an overall price below 100 €.
The approaches to reach these goals are described below.

2.1. Customization
Different ways of producing a 3D model of the own hand were tested: Scanning by a
smartphone using the app Scann3D (from SmartMobileVision, Budapest, Hungary); or using
the free computer software 3DF Zephyr Free (3Dflow SRL, Verona, Italy), which enables
producing a 3D object from 2D photographs that were taken with a Nikon Coolpix P510.
 2.1. Customization
Different ways of producing a 3D model of the own hand were tested: Scanning by a
smartphone using the app Scann3D (from SmartMobileVision, Budapest, Hungary); or
using the free computer software 3DF Zephyr Free (3Dflow SRL, Verona, Italy), which
Technologies 2021, 9, 63 enables producing a 3D object from 2D photographs that were taken with a Nikon Cool- 3 of 12
pix P510.
The gained 3D object was imported in Autodesk Meshmixer (Autodesk, Inc., San
Rafael, CA, USA) where errors in the 3D shape were corrected. Afterwards, the 3D model
The gained
was imported in 3D object (Visual
MeshLab was imported in Autodesk
Computing Meshmixer
Lab, Institute (Autodesk,
of Information Inc., and
Science San
Rafael, CA, USA) where errors in the 3D shape were corrected. Afterwards, the 3D model
Technologies “Alessandro Faedo”—ISTI, National Research Council of Italy, Pisa, Italy),
was imported in MeshLab (Visual Computing Lab, Institute of Information Science and
where meshing errors were corrected.
Technologies “Alessandro Faedo”—ISTI, National Research Council of Italy, Pisa, Italy),
Next, the mesh was imported into Fusion 360 (Autodesk, Inc.), to scale the model to
where meshing errors were corrected.
the correct dimensions. Finally, holders fitting to two different positions of the meshed
Next, the mesh was imported into Fusion 360 (Autodesk, Inc.), to scale the model to
hand were constructed, one between thumb and wrist joint and another one between
the correct dimensions. Finally, holders fitting to two different positions of the meshed
small finger and wrist joint. Besides, a holder for the electronics and electromechanical
hand were constructed, one between thumb and wrist joint and another one between small
parts at the arm near the wrist was developed.
finger and wrist joint. Besides, a holder for the electronics and electromechanical parts at
The whole development process is shown in the chart in Figure 1.
the arm near the wrist was developed.
The whole development process is shown in the chart in Figure 1.

CAD Final
Aim 3D scanning 3D printing Assembly Evaluation
construction product

Figure 1.
Figure 1. Development
Development process of the
process of the myoelectric
myoelectric thumb.
thumb.

2.2.
2.2. 3D
3D Printing
Printing
The
The printed
printed parts
parts were
were exported
exported from
from Fusion
Fusion 360
360 as stl files
as stl files and
and sliced
sliced with
with the
the
software Cura (Ultimaker
software Cura (Ultimaker B. B.V.,
V.,Utrecht,
Utrecht,The
TheNetherlands).
Netherlands).Printing
Printing was
was performed
performed us-
using
ing an Ender
an Ender V2 (Shenzhen
V2 (Shenzhen Creality
Creality 3D Technology
3D Technology Co.,Shenzhen,
Co., Ltd., Ltd., Shenzhen,
China)China) with a
with a nozzle
nozzle
diameter diameter
of 0.4 mm.of 0.4
As mm. As filament,
filament, polylacticpolylactic
acid (PLA)acid
was(PLA)
chosen, was chosen,
which which has
has advantages
advantages like good printability, being eco-friendly and degradable, and
like good printability, being eco-friendly and degradable, and its shape memory properties its shape
memory properties allow for bending parts that cannot be printed easily
allow for bending parts that cannot be printed easily in the desired shape [36,37]. in the desired
The
shape [36,37].
following The following
parameters parameters
were used were used
for printing: layerfor printing:
thickness layer
0.16 mm,thickness 0.16 mm,
3 perimeters, fill
3density
perimeters, fill density
20%, printing temperature 210 ◦ temperature
20%, printing 210temperature
C, printing bed °C, printing50bed ◦ C,temperature
printing speed 50
°C, printing
50 mm/s, andspeed 50 mm/s,
the angle from and
whichtheanangle from needs
overhang whichsupport
an overhang
structure needs
wassupport
chosen
as 51◦ . These
structure was support
chosen as 51°. These
structures support
were structures
carefully werefrom
separated carefully separated
the printed fromafter
objects the
printed objects after printing, partly
printing, partly supported by a knife. supported by a knife.

2.3. Springs
2.3. Springs to
to Let
Let the
the Thumb
Thumb Relax
Relax
To enable
To enablerelaxation
relaxationofofthe
thethumb
thumbwithout muscle
without contraction,
muscle a torsion
contraction, spring
a torsion as well
spring as
as a tension spring were investigated. In both cases, different possibilities to mount
well as a tension spring were investigated. In both cases, different possibilities to mount the
springs
the at the
springs 3D printed
at the parts
3D printed were
parts investigated.
were investigated.
2.4. Electronics, Sensors and Programming
2.4. Electronics, Sensors and Programming
The third thumb was controlled by an Arduino Uno, testing two different programs
The third thumb was controlled by an Arduino Uno, testing two different programs
(cf. Appendix A). Figure 2 shows the wiring.
(cf. Appendix A). Figure 2 shows the wiring.
The servomotor used in this work is an SG90 (Tower Pro, Taipei, Taiwan) with the
following parameters [38]: dimensions 23 mm × 12.2 mm × 29 mm, mass 9 g, rotation
speed 0.1 s for 60◦ , torque 1.8 kg/cm, voltage 4.8 V. This servomotor pulls a string (here a
fishing line) to actuate the phalanges.
The myoelectric signals were taken by three surface electrodes including a reference
electrode (standard ECG gel electrodes) on the forearm and evaluated by the muscle sensor
board V3 (Advancer Technologies, Raleigh, NC, USA). The electrodes were placed on the
Flexor carpi ulnaris and the Flexor carpi radialis, i.e., the muscles responsible for deflection
and extension of the wrist joint. It should be mentioned that rotation of the third thumb
was included into the design for 3D printing, but not yet controlled by myoelectric signals.
 Technologies 2021, 9, 63 4 of 12

Technologies 2021, 9, 63 4 of 12

+5V

-5V

+5 V

GN

-5V

SIG

GN

Figure
Figure 2. 2. Electronicsetup
Electronic setupwith
withArduino
ArduinoUno
Uno(upper
(upper left),
left), servomotor
servomotor (upper
(upperright),
right),breadboard
breadboardconnections
connections(lower left)
(lower left)
and muscle sensor board (lower right).
and muscle sensor board (lower right).

The servomotor used in this work is an SG90 (Tower Pro, Taipei, Taiwan) with the
3. Results
following parameters [38]: dimensions 23 mm × 12.2 mm × 29 mm, mass 9 g, rotation
While scanning a hand did not lead to satisfactory results, using 3DF Zephyr princi-
speed 0.1 s for 60°, torque 1.8 kg/cm, voltage 4.8 V. This servomotor pulls a string (here a
Technologies 2021, 9, 63 pally enabled preparing a 3D model of a hand with several optimization steps, as visible5 of 12 in
fishing line) to actuate the phalanges.
FigureThe
3. myoelectric signals were taken by three surface electrodes including a reference
electrode (standard ECG gel electrodes) on the forearm and evaluated by the muscle
sensor board V3 (Advancer Technologies, Raleigh, NC, USA). The electrodes were placed
on the Flexor carpi ulnaris and the Flexor carpi radialis, i.e., the muscles responsible for
deflection and extension of the wrist joint. It should be mentioned that rotation of the
third thumb was included into the design for 3D printing, but not yet controlled by my-
oelectric signals.

3. Results
While scanning a hand did not lead to satisfactory results, using 3DF Zephyr prin-
cipally enabled preparing a 3D model of a hand with several optimization steps, as visi-
ble in Figure 3.

Figure
Figure 3. Different
3. Different results
results ofofphotographs
photographsinserted
insertedinto
into3DF
3DF Zephyr.
Zephyr. (a–d)
(a–d) show
show different
differentapproaches,
approaches,asasdescribed
describedinin
thethe
main text.
main text.

Here,different
Here, differenterrors
errorsoccur.
occur. Figure
Figure 3a
3a shows
shows aaversion
versionininwhich
whichthe
thehand
handwas not
was not
held at the identical position during taking 50 photographs, resulting in a “shadow
held at the identical position during taking 50 photographs, resulting in a “shadow hand”.
hand”. Insufficient contrast between hand and background may lead to insufficient edge
Insufficient contrast between hand and background may lead to insufficient edge detection
detection and, thus, incomplete parts of the hand, here especially to “holes” in the fingers
and, thus, incomplete parts of the hand, here especially to “holes” in the fingers (Figure 3b).
(Figure 3b). In Figure 3c, an effect is visible that may occur when photographs are taken
In Figure 3c, an effect is visible that may occur when photographs are taken in front of a
in front of a monochromatic background. Finally, Figure 3d shows the best result gained
monochromatic background. Finally, Figure 3d shows the best result gained by this technique,
by this technique, combining 50 photographs for a chosen reconstruction “human body”
combining 50 photographs for a chosen reconstruction “human body” and the selection “surface
and the selection “surface scan”, applying the highest possible resolution. Generally,
scanning body parts is much more complicated than scanning rigid objects since the lat-
ter keep their position, while the hand here could not be stabilized by any rigid object
since all sides had to be photographed without other objects blocking the line of sight.
Next, the model shown in Figure 3d was introduced in Meshmixer (Figure 4a)
 Here, different errors occur. Figure 3a shows a version in which the hand was not
held at the identical position during taking 50 photographs, resulting in a “shadow
hand”. Insufficient contrast between hand and background may lead to insufficient edge
Technologies 2021, 9, 63 detection and, thus, incomplete parts of the hand, here especially to “holes” in the fingers 5 of 12
(Figure 3b). In Figure 3c, an effect is visible that may occur when photographs are taken
in front of a monochromatic background. Finally, Figure 3d shows the best result gained
by this technique, combining 50 photographs for a chosen reconstruction “human body”
scan”,
and theapplying the highest
selection “surfacepossible
scan”, resolution.
applying the Generally,
highestscanning
possible body parts isGenerally,
resolution. much more
complicated
scanning body parts is much more complicated than scanning rigid objects sincehand
than scanning rigid objects since the latter keep their position, while the here
the lat-
could not be stabilized by any rigid object since all sides had to be photographed
ter keep their position, while the hand here could not be stabilized by any rigid object without other
objects blocking
since all the to
sides had linebeofphotographed
sight. without other objects blocking the line of sight.
Next, the model shown
Next, the model shown in in
Figure 3d was
Figure introduced
3d was in Meshmixer
introduced in Meshmixer(Figure 4a) where,
(Figure 4a)
e.g., ringe.g.,
where, finger and
ring middle
finger andfinger
middlewere separated,
finger as shownas
were separated, inshown
Figure in
4b.Figure
Next, 4b.
open areas
Next,
and
open meshing errors
areas and were corrected
meshing (Figure
errors were 4c), before
corrected the final
(Figure 4c), 3D handthe
before model
finalwas
3D derived
hand
(Figure
model 4d).
was derived (Figure 4d).

a b c d

Figure
Figure 4. Reconstruction
4. Reconstruction of of
thethe hand
hand model
model in Meshmixer.
in Meshmixer. (a)(a) Hand
Hand taken
taken from
from 3DF
3DF Zephyr;
Zephyr; (b)(b)
3D3D hand
hand model
model with
with ring
ring finger and middle finger separated; (c) correction of open areas and meshing errors; (d) final 3D hand model.
finger and middle finger separated; (c) correction of open areas and meshing errors; (d) final 3D hand model.

Next,the
Next, thedifferent
different holders
holders (one
(one forfor
thethe motor,
motor, twotwo for different
for the the different possible
possible posi- of
positions
tions of the third thumb) as well as the third thumb itself were constructed.
the third thumb) as well as the third thumb itself were constructed. The basic constructions The basic
constructions
are depicted in are depicted
Figure in Figure
5. Here, 5. Here,
the holders werethefitted
holders were
to the fitted to reconstructed
previously the previously3D
hand by constructing a plane, changing it roughly to the desired shape ofdesired
reconstructed 3D hand by constructing a plane, changing it roughly to the shape
the holder and
of the holder and drawing it to the surface of the 3D hand. In some areas,
drawing it to the surface of the 3D hand. In some areas, the holders were constructed6smaller the holders
Technologies 2021, 9, 63 of 12
were
than constructed
the hand to allowsmaller thanfit
a tighter the
onhand to allow
the soft a tighterNext,
hand regions. fit onall
the soft were
edges hand rounded.
regions.
Next, all edges were rounded.

Figure 5. Third thumb at two different positions and motor holder.

Figure 5. Third thumb at two different positions and motor holder.

On the holders, supporting systems (grey parts in Figure 5) were constructed on

which the actual thumbs were placed. The whole thumb construction is depicted in Fig-
ure 6. Connections between adjacent phalanges are made by screws. All holders contain
slits to fix textile bands, ensuring a reliable fixation of the holders on the hand or forearm,
Technologies 2021, 9, 63 6 of 12
Figure 5. Third thumb at two different positions and motor holder.

On the holders, supporting systems (grey parts in Figure 5) were constructed on

On the holders, supporting systems (grey parts in Figure 5) were constructed on which
which the actual thumbs were placed. The whole thumb construction is depicted in Fig-
the actual thumbs were placed. The whole thumb construction is depicted in Figure 6.
ure 6. Connections between adjacent phalanges are made by screws. All holders contain
Connections between
slits to fix textile adjacent
bands, phalanges
ensuring are
a reliable made by
fixation screws.
of the All on
holders holders contain
the hand slits to fix
or forearm,
textile bands,
respectively. ensuring a reliable fixation of the holders on the hand or forearm, respectively.

Technologies 2021, 9, 63 7 of 12

To enable rotation of the thumb, a deep groove ball bearing (623-2Z, 3 mm ID × 10

mm OD × 4 mm wide, light blue part in Figure 6) was pressed into the borehole of the
base phalanx. The latter was connected with the supporting system (beige part in Figure
5) by an M3 screw. Annular grooves for a pressure spring (Figure 7a) were added to the
supporting system as well as to the base phalanx, letting the thumb rotate back into the
neutral position when rotated. The length of the base phalanx is 27.3 mm. The interme-
diate phalanx can be bent by 50° with respect to the base phalanx and has a height of 28.8
mm. The end phalanx can be bent by 45° with respect to the intermediate phalanx. On its
tip, it has an M2 borehole on which a cover is plugged.
Moving the finger is enabled by a fishing line fixed at the end phalanx by this M2
screw and with its other end at the servomotor. The path of the fishing line is depicted in
Figure
Figure6.5 by thethumb
6.Third
Third greenincurved
(a) grooves (b) in bent
the thumb part and the grey groove in the sup-
thumb in (a) stretched
stretched and (b) bentposition.
Figure
port system part, respectively. At and
either ends position.
of the phalanges, the channel guiding the
fishing line was extended to avoid clamping it.
To enable rotation of the thumb, a deep groove ball bearing (623-2Z, 3 mm ID × 10 mm OD
× 4 mmTo let the light
wide, thumb move
blue partback into a6)neutral
in Figure position
was pressed intowithout detected
the borehole muscle
of the signals,
base phalanx.
springs
The latterwere
wasadded between
connected the supporting
with the phalanges. system
Firstly,(beige
a torsion
partspring was5)tested;
in Figure however,
by an M3 screw.
placement
Annular of the
grooves forspring wasspring
a pressure complicated,
(Figure 7a)and noadded
were fittingtotorsion springssystem
the supporting were asfound.
well
Thus,
as to thea tension springletting
base phalanx, was used instead,
the thumb as depicted
rotate back intointhe Figure 7. position
neutral While first
whentests were
rotated.
carried
The lengthoutofbythe
attaching the springs
base phalanx to 3DThe
is 27.3 mm. printed extensions
intermediate (Figure
phalanx can 7b), thisby
be bent 50◦ with
procedure
was found not reliable enough, with some extensions breaking during
respect to the base phalanx and has a height of 28.8 mm. The end phalanx can be bent by separation of 45 ◦
the
support
with material
respect to theorintermediate
being abraded during
phalanx. On attaching
its tip, itthe
hassprings. Thus, inon
an M2 borehole thewhich
final version
a cover
small screws were placed in thread inserts at the ends of the phalanges (Figure 7c).
is plugged.

Figure
Figure 7.
7. (a)
(a) Annular
Annular grooves
grooves for
for aa pressure
pressure spring; tension springs
spring; tension springs attached
attached to
to (b)
(b) 3D
3D printed
printed parts
parts
of the phalanges; (c) additional M2 screws at the ends of the phalanges.
of the phalanges; (c) additional M2 screws at the ends of the phalanges.

After these iteration steps, the final 3D parts (available at

https://www.thingiverse.com/thing:4938144, accessed on 24 August 2021) were printed
in the orientations shown in Figure 8. Light blue parts show support structures. Positions
Technologies 2021, 9, 63 7 of 12

Moving the finger is enabled by a fishing line fixed at the end phalanx by this M2
screw and with its other end at the servomotor. The path of the fishing line is depicted in
Figure 5 by the green curved grooves in the thumb part and the grey groove in the support
system part, respectively. At either ends of the phalanges, the channel guiding the fishing
line was extended to avoid clamping it.
To let the thumb move back into a neutral position without detected muscle signals,
springs were added between the phalanges. Firstly, a torsion spring was tested; however,
placement of the spring was complicated, and no fitting torsion springs were found. Thus,
a tension spring was used instead, as depicted in Figure 7. While first tests were carried out
by attaching the springs to 3D printed extensions (Figure 7b), this procedure was found not
reliable enough, with some extensions breaking during separation of the support material
or being abraded during attaching the springs. Thus, in the final version small screws were
placed in thread inserts at the ends of the phalanges (Figure 7c).
After these iteration steps, the final 3D parts (available at https://www.thingiverse.
com/thing:4938144, accessed on 24 August 2021) were printed in the orientations shown
in Figure 8. Light blue parts show support structures. Positions were chosen in such a way
that boreholes and especially the channel guiding the fishing line were printable without
support. Printing of all parts for two additional thumbs took 21 h. After chamfering the
parts, the thread inserts were heated by a soldering gun and pressed into the respective
boreholes. The fishing line was guided through the channel and fixed with the M2 screw
at the end phalanx. From the M5 thread insert at the base phalanx, the fishing line was
Technologies 2021, 9, 63
led into the connector (cf. lower left of Figure 6b) and through the tube (marked black in
8 of 12
Figure 5) to the connector at the servo holder. The different parts of the thumb were finally
connected with screws, and the springs (Figure 7) were added. Detailed descriptions of
the7)assembly process can be found in the animations for the third thumb in both possible
were added. Detailed descriptions of the assembly process can be found in the anima-
positions in the Supplementary
tions for the third thumb in bothMaterial.
possible positions in the Supplementary Material.

FigureFigure 8. Printing
8. Printing orientations
orientations ofparts
of the the parts of third
of the the third thumb:
thumb: (a)(a)
endend phalanx,(b)
phalanx, (b)bearing
bearingcap,
cap,(c)
(c) end
end phalanx
phalanx cap,
cap,(d)
(d) base
base phalanx, I hand holder for the outer side (between small finger and wrist), (f) hand holder for the inner side (be-
phalanx, I hand holder for the outer side (between small finger and wrist), (f) hand holder for the inner
tween thumb and wrist), (g) intermediate phalanx, (h) wrist holder for servo motor, (i) housing for servo motor.
side (between
thumb and wrist), (g) intermediate phalanx, (h) wrist holder for servo motor, (i) housing for servo motor.
4. Discussion
During the iterative development of this third thumb or myoelectric finger prosthe-
sis, several potential problems were recognized. A first test was performed with hand
holders of thickness 2 mm, which was found to be mechanically insufficient; the slits
 Technologies 2021, 9, 63 8 of 12

4. Discussion
During the iterative development of this third thumb or myoelectric finger prosthesis,
several potential problems were recognized. A first test was performed with hand holders
of thickness 2 mm, which was found to be mechanically insufficient; the slits holding the
textile strips broke several times. Thus, the design had to be changed to a holder thickness
of 3 mm.
Unexpectedly, the fishing line could be elongated too much, which made it less suitable
for rigidly holding objects. Besides, the channel guiding the fishing line was found to be
too broad; this additional degree of freedom increased the problem of differing thumb
positions with and without load.
On the other hand, the hook-and-loop fastener used here as textile band to fix the
21, 9, 63 third thumb and servo holders was too rigid and should be exchanged by a9 slightly of 12 elastic
textile material in a next step.
Practical tests of the developed third thumb (cf. third-thumb-tests.avi in the Supple-
mentary Material and Figure 9) showed that a third thumb at the inner side (between
of water. Regarding the thumb
original holders,andthe third
wrist) thumb
could was better
be helpful to grabwearable between
larger objects small
since the third thumb
finger and wrist since there the hand deformed less when an object was taken.
could be used like the original one, just with a larger distance to the other fingers. On the
The programs testedside,
opposite during
i.e., this study
between (cf.finger
small Appendix A) were
and wrist, found
it could to react
be used differ-
to hold a soft ball, but
ently to noise in was
the sometimes
EMG signals. The latterespecially
even disturbing, had a higher
duringdegree
grabbing oflong
freedom,
objects,making
such as a bottle of
movements morewater. Regarding
reliable, the holders,has
but nevertheless the third
to bethumb was better
developed wearable
further between
to reduce unde- small finger
and wrist
sired movements due to noise. since there the hand deformed less when an object was taken.

(a) (b) (c)

Figure Figure 9. Screenshots
9. Screenshots from animation-small-finger.avi
from animation-small-finger.avi (Supplementary
(Supplementary Material): (a)Material):
holding a (a)
softholding
ball, (b)apressing
soft it
againstball, (b) of
the side pressing
the thirditthumb,
againstand
the(c)side of the
holding third thumb,
it between real andand (c)thumb.
third holding it between real and third
thumb.
The programs tested during this study (cf. Appendix A) were found to react differently
to noise
It should be mentionedin the that
EMGthis
signals. The latter
project mainly hadaimed
a higheratdegree of freedom,
producing making movements
a customized
more reliable, but nevertheless has to be developed further to reduce undesired movements
low-cost myoelectric prosthesis, without applying brain related measured or investigat-
due to noise.
ing the effect of training, as the aforementioned studies on using a third thumb did
It should be mentioned that this project mainly aimed at producing a customized
[27,28]. Another difference to Ref. [27]
low-cost myoelectric is that here
prosthesis, not applying
without the big toe, but
brain an arm
related muscleorwas
measured investigating
used to control the
the third
effect thumb, making
of training, as the aaforementioned
comparison ofstudiestraining effectsa third
on using complicated.
thumb did [27,28].
The extra roboticAnother
thumb described inRef.
difference to Ref.[27]
[28]
is has
that ahere
different
not thedesign, but an
big toe, but similar degrees
arm muscle was used to
of freedom, and is equipped
control withthumb,
the third a force sensor;
making such a feedback
a comparison system
of training wascomplicated.
effects not yet im- The extra
robotic studies
plemented here. Other thumb described in Ref.
mentioning [28] has a different
myoelectric design, but
finger control are similar
usuallydegrees
relatedof freedom,
and is equipped with a force sensor; such a feedback system
to full hand prostheses [39,40] or evaluate myoelectric signals during finger control in was not yet implemented
here. Other studies mentioning myoelectric finger control are usually related to full hand
general for future implementation [41–44].
prostheses [39,40] or evaluate myoelectric signals during finger control in general for future
implementation [41–44].
5. Conclusions
In this project, a low-cost myoelectric thumb was developed. While tested here as
third thumb, it can also be used as prosthesis by people who lost a thumb.
This first approach works for grabbing and holding small, lightweight objects. With
the improvements pointed out, more functional prostheses or third thumbs can be de-
Technologies 2021, 9, 63 9 of 12

5. Conclusions
In this project, a low-cost myoelectric thumb was developed. While tested here as
third thumb, it can also be used as prosthesis by people who lost a thumb.
This first approach works for grabbing and holding small, lightweight objects. With
the improvements pointed out, more functional prostheses or third thumbs can be de-
veloped, based on the freely available printing models for 3D printing and programs
available as Supplementary Material. Further research is necessary to improve thumb
positioning as well as the motor strength and the material of the fishing line, actuating the
thumb. Besides, implementation of a feedback can allow for better control of the prosthesis,
especially concerning the grabbing force, which is planned to be measured in the next step
of the project.

Supplementary Materials: The following videos are available online at https://www.mdpi.com/

article/10.3390/technologies9030063/s1: animation-thumb-side.avi, showing the assembly and
function of the third thumb placed between thumb and wrist; animation-small-finger.avi, showing
the assembly and function of the third thumb placed between small finger and wrist; and third-
thumb-tests.avi, showing real tests with a third thumb, grabbing and holding a small ball with
different combinations of fingers.
Author Contributions: Conceptualization, all authors; methodology, M.A.; investigation, M.A.;
writing—original draft preparation, A.E.; writing—review and editing, M.A. and I.W.; visualization,
M.A., A.E. and I.W.; supervision, I.W. All authors have read and agreed to the published version of
the manuscript.
Funding: This research was partly funded by the HiF funds of Bielefeld University of Applied Sciences.
Data Availability Statement: All data created in this study are presented in this paper or in the
Supplementary Material.
Conflicts of Interest: The authors declare no conflict of interest. The funders had no role in the design
of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or
in the decision to publish the results.

Appendix A
Program code V1:
#include <Servo.h> // includes servo library
Servo SERVO1; // setup servo object
float VALUE; // define float VALUE
int BORDER = 80; // border defined by integer 80
void setup () {
Serial.begin(9600); // sets baud rate to 9600 bits/s
SERVO1.attach(9); // addresses servo library to pin 9
}
void loop () {
VALUE = analogRead(A3); // addresses analog input A3 as VALUE
if (VALUE > BORDER) {
SERVO1.write(0); // SERVO1 receives value 0
}
else {
SERVO1.write(120); // SERVO1 receives value 120
}
Serial.println(VALUE); // serial port receives data VALUE
}
Program code V2:
#include <Servo.h>
#define SLOPE 30
Servo SERVO1;
Technologies 2021, 9, 63 10 of 12

float DIFF; //define different float values

float NEU_VALUE;
float VALUE;
long timer = 0; // recent time
int DELAY = 100; // delay time
void setup () {
Serial.begin(9600); // sets baud rate to 9600 bits/s
SERVO1.attach(9); // addresses servo library to pin 9
}
void loop () {
if (millis () > (timer + DELAY)) {
VALUE = analogRead(A3);
DIFF= VALUE-NEU_VALUE;
NEU_VALUE = VALUE;
if (DIFF > SLOPE) {
SERVO1.write(0); // SERVO1 receives value 0
}
else if (DIFF < (-SLOPE)) {
SERVO1.write(120); // SERVO1 receives value 120
}
timer = millis ();
}
Serial.println(analogRead(A3)); // serial port receives data VALUE
}

References
1. Childress, D.S. Historical aspects of powered limb prosthesis. Clin. Prosthet. Orthot. 1985, 9, 2–13.
2. Geethanjali, P. Myoelectric control of prosthetic hands: State-of-the-art review. Med. Devices Évid. Res. 2016, 9, 247–255. [CrossRef]
[PubMed]
3. Hahne, J.M.; Biessmann, F.; Jiang, N.; Rehbaum, H.; Farina, D.; Meinecke, F.C.; Muller, K.R.; Parra, L.C. Linear and nonlinear
regression techniques for simultaneous and proportional myoelectric control. IEEE Trans. Neural. Syst. Rehabil. Eng. 2014,
22, 269–279. [CrossRef] [PubMed]
4. Cipriani, C.; Zaccone, F.; Micera, S.; Carrozza, M.C. On the shared control of an EMG-controlled prosthetic hand: Analysis of
user-prosthesis interaction. IEEE Trans. Robot. 2008, 24, 170–184. [CrossRef]
5. Weir, R.F.; Troyk, P.R.; DeMichele, G.A.; Kerns, D.A.; Schorsch, J.F.; Maas, H. Implantable myoelectric sensors (IMESs) for
intramuscular electromyogram recording. IEEE Trans. Biomed. Eng. 2009, 56, 159–171. [CrossRef] [PubMed]
6. Kuiken, T.A.; Li, G.; Lock, B.A.; Lipschutz, R.D.; Stubblefield, K.A.; Englehart, K. Targeted muscle reinnervation for real-time
myoelectric control of multifunction artificial arms. JAMA 2009, 301, 619–628. [CrossRef]
7. Miller, L.A.; Stubblefield, K.A.; Lipschutz, R.D.; Lock, B.A.; Kuiken, T.A. Improved myoelectric prosthesis control using targeted
reinnervation surgery: A case series. IEEE Trans. Neural Syst. Rehab. Eng. 2008, 16, 46–50. [CrossRef] [PubMed]
8. Hijjawi, J.B.; Kuiken, T.A.; Lipschutz, R.D.; Miller, L.A.; Stubblefield, K.A.; Dumanian, G.A. Improved myoelectric prosthesis
control accomplished using multiple nerve transfers. Plast. Reconstr. Surg. 2006, 118, 1573–1578. [CrossRef]
9. Lotze, M.; Grodd, W.; Birbaumer, N.; Erb, M.; Huse, E.; Flor, H. Does use of a myoelectric prosthesis prevent cortical reorganization
and phantom limb pain? Nat. Neurosci. 1999, 2, 501–502. [CrossRef]
10. Fougner, A.; Stavdahl, O.; Kyberd, P.J.; Losier, Y.G.; Parker, P.A. Control of upper limb prostheses: Terminology and proportional
myoelectric control—A review. IEEE Trans. Neural Syst. Rehab. Eng. 2012, 20, 663–677. [CrossRef]
11. Madusanka, D.G.K.; Wijayasingha, L.N.S.; Gopura, R.A.R.C.; Amarasinghe, Y.W.R.; Mann, G.K.I. A review on hybrid myoelectric
control systems for upper limb prosthesis. In Proceedings of the 2015 Moratuwa Engineering Research Conference (MERCon),
Moratuwa, Sri Lanka, 7–8 April 2015; pp. 136–141.
12. Carey, S.L.; Lura, D.J.; Highsith, M.J. Differences in myoelectric and body-powered upper-limb prostheses: Systematic literature
review. JRRD 2015, 52, 247–262. [CrossRef]
13. Iqbal, N.V.; Subramaniam, K.; Asmi, P.S. A review on upper-limb myoelectric prosthetic control. IETE J. Res. 2018, 64, 740–752.
[CrossRef]
14. Light, C.M.; Chappell, P.H.; Hudgins, B.; Engelhart, K. Intelligent multifunction myoelectric control of hand prostheses. J. Med.
Eng. Technol. 2002, 26, 139–146. [CrossRef]
15. Ma, J.X.; Thakor, N.V.; Matsuno, F. Hand and wrist movement control of myoelectric prosthesis based on synergy. IEEE Trans.
Hum.-Mach. Syst. 2015, 45, 74–83. [CrossRef]
Technologies 2021, 9, 63 11 of 12

16. Godfrey, S.B.; Zhao, K.d.; Theuer, A.; Catalano, M.G.; Bianchi, M.; Breighner, R.; Bhaskaran, D.; Lennon, R.; Grioli, G.; Santello,
M.; et al. The SoftHand Pro: Functional evaluation of a novel, flexible, and robust myoelectric prosthesis. PLoS ONE 2018,
13, e0205653. [CrossRef] [PubMed]
17. Krasoulis, A.; Vijayakumar, S.; Nazarpour, K. Effect of user practive on prosthetic finger control with an intuitive myoelectric
decoder. Front. Neurosci. 2019, 13, 891. [CrossRef] [PubMed]
18. Gijsberts, A.; Bohra, R.; Sierra González, D.; Werner, A.; Nowak, M.; Caputo, B.; Roa, M.A.; Castellini, C. Stable myoelectric
control of a hand prosthesis using non-linear incremental learning. Front. Neurorobot. 2014, 8, 8. [CrossRef]
19. Resnik, L.; Meucci, M.R.; Lieberman-Klinger, S.; Fantini, C.; Kelty, D.L.; Disla, R.; Sasson, N. Advanced upper limb prosthetic
devices: Implications for upper limb prosthetic rehabilitation. Arch. Phys. Med. Rehabil. 2012, 93, 710–717. [CrossRef]
20. Atique, M.D.M.; Rabbani, S. A cost-effective myoelectric prosthetic hand. J. Prosthet. Orthot. 2018, 30, 231–235. [CrossRef]
21. Ku, I.H.; Lee, G.K.; Park, C.Y.; Lee, J.H.; Jeong, E.C. Clinical outcomes of a low-cost single-channel myoelectric-interface
three-dimensional hand prosthesis. Arch. Plast. Surg. 2019, 46, 303–310. [CrossRef]
22. e-NABLE. Available online: https://enablingthefuture.org/ (accessed on 19 August 2021).
23. Lu, Z.Y.; Chen, X.; Zhang, X.; Tong, K.-Y.; Zhou, P. Real-time control of an exoskeleton hand robot with myoelectric pattern
recognition. Int. J. Neural Syst. 2017, 27, 1750009. [CrossRef]
24. Anam, K.; Rosyadi, A.A.; Sujanarko, B.B.; Al-Jumaily, A. Myoelectric control systems for hand rehabilitation device: A review.
In Proceedings of the 2017 4th International Conference on Electrical Engineering, Computer Science and Informatics (EECSI),
Yogyakarta, Indonesia, 19–21 September 2017; pp. 1–6.
25. Burns, M.K.; Pei, D.Y.; Vinjamuri, R. Myoelectric control of a soft hand exoskeleton using kinematic synergies. IEEE Trans. Biomed.
Circ. Syst. 2019, 13, 1351–1361. [CrossRef]
26. The Third Thumb. Available online: https://www.daniclodedesign.com/thethirdthumb (accessed on 19 August 2021).
27. Kieliba, P.; Clode, D.; Maimon-Mor, R.O.; Makin, T.R. Robotic hand augmentation drives changes in neural body representation.
Sci. Robot. 2021, 6, eabd7935. [CrossRef] [PubMed]
28. Meraz, N.S.; Sobajima, M.; Aoyama, T.; Hasegawa, Y. Modification of body scheme by use of extra robotic thumb. ROBOMECH J.
2018, 5, 3. [CrossRef]
29. Zhuang, K.Z.; Sommer, N.; Mendez, V.; Aryan, S.; Formento, E.; D‘Anna, E.; Artoni, F.; Petrini, F.; Granata, G.; Cannaviello,
G.; et al. Shared human-robot proportional control of a dexterous myoelectric prosthesis. Nat. Mach. Intell. 2019, 1, 400–411.
[CrossRef]
30. Markovic, M.; Schweisfurth, M.A.; Engels, L.F.; Bentz, T.; Wüstefeld, D.; Farina, D.; Dosen, S. The clinical relevance of advanced
artificial feedback in the control of a multi-functional myoelectric prosthesis. J. NeuroEng. Rehab. 2018, 15, 28. [CrossRef]
[PubMed]
31. Shehata, A.W.; Scheme, E.J.; Sensinger, J.W. Myoelectric prosthesis control strategies. IEEE Trans. Neural Syst. Rehab. Eng. 2018,
26, 1046–1055. [CrossRef]
32. Betthauser, J.L.; Hunt, C.L.; Osborn, L.E.; Masters, M.R.; Lévay, G.; Kaliki, R.R.; Thakor, N.V. Limb Position Tolerant Pattern
Recognition for Myoelectric Prosthesis Control with Adaptive Sparse Representations From Extreme Learning. IEEE Trans.
Biomed. Eng. 2018, 65, 770–778. [CrossRef] [PubMed]
33. Strbac, M.; Isakovic, M.; Belic, M.; Popovic, I.; Simanic, I.; Farina, D.; Keller, T.; Dosen, S. Short- and long-term learning of
feedforward control of a myoelectric prosthesis with sensory feedback by amputees. IEEE Trans. Neural Syst. Rehab. Eng. 2017,
25, 2133–2145. [CrossRef] [PubMed]
34. Winslow, B.D.; Ruble, M.; Huber, Z. Mobile, game-based training for myoelectric prosthesis control. Front. Bioeng. Biotechnol.
2018, 6, 94. [CrossRef] [PubMed]
35. Boschmann, A.; Neuhaus, D.; Vogt, S.; Kaltschmidt, C.; Platzner, M.; Dosen, S. Immersive augmented reality system for the
training of pattern classification control with a myoelectric prosthesis. J. NeuroEng. Rehab. 2021, 18, 25. [CrossRef] [PubMed]
36. Chalgham, A.; Wickenkamp, I.; Ehrmann, A. Mechanical properties of FDM printed PLA parts before and after thermal treatment.
Polymers 2021, 13, 1239. [CrossRef] [PubMed]
37. Ehrmann, G.; Ehrmann, A. 3D printing of shape memory polymers. J. Appl. Polym. Sci. 2021, 138, 50847. [CrossRef]
38. Tower Pro. Available online: https://www.towerpro.com.tw/product/sg90-7 (accessed on 21 August 2021).
39. Dupan, S.S.G.; Vujaklija, I.; Stachaczyk, M.K.; Hahne, J.M.; Stegeman, D.F.; Dosen, S.S. Online simultatenous myoelectric finger
control. In Converging Clinical and Engineering Research on Neurohabilitation III; Masia, L., Micera, S., Akay, M., Pons, J., Eds.; ICNR
2018. Biosystems & Biorobotics 21; Springer: Cham, Switzerland, 2018.
40. Srimaneepong, V.; Heboyan, A.; Yaqin Syed, A.U.; Anh Trinh, H.; Amornvit, P.P.; Rokaya, D. Recent advances in myoelectric
control for finger prostheses for multiple finger loss. Appl. Sci. 2021, 11, 4464. [CrossRef]
41. Mehra, P.; Dave, M.; Khan, A.; Tong, R.K.Y. Spatial mapping and feature analysis for individual finger movements using high
density electromyography: Preliminary study. J. Image Graph. 2020, 8, 75–79. [CrossRef]
42. Barsotti, M.; Dupan, S.; Vujaklija, I.; Dosen, S.; Frisoli, A.; Farina, D. Online finger control using high-density EMG and minimal
training data for robotic applications. IEEE Robot. Autom. Lett. 2019, 4, 217–223. [CrossRef]
Technologies 2021, 9, 63 12 of 12

43. Anam, K.; Al-Jumaily, A. Evaluation of extreme learning machine for classification of individual and combined finger movements
using electromyography on amputees and non-amputees. Neural Netw. 2017, 85, 51–68. [CrossRef] [PubMed]
44. Murciego, L.P.; Barsotti, M.; Frisoli, A. Synergy-based multi-fingers forces reconstruction and discrimination from forearm EMG.
In Haptics: Science, Technology, and Applications; Prattichizzo, D., Shinoda, H., Tan, H., Ruffaldi, E., Frisoli, A., Eds.; EuroHaptics
2018. Lecture Notes in Computer Science 10894; Springer: Cham, Switzerland, 2018.