Modular 3D Printed Prosthetic Hand – Amputee Tests with the Touch Hand 4

. The popularity of myoelectric prosthetic hands increases as technological advances allows for the sophistication on prosthetic devices to become more affordable. A 3D printed myoelectric prosthetic hand was designed to ensure that it is cost-effective whilst having functionality similar to commercially available prosthetic hands. A three-fingered prosthetic hand was designed to grip objects of different shapes using one grip. An electromyography control system was investigated by performing tests on an able-bodied person and observing the results. The electromyography control system for the prosthetic hand was then designed. The prosthetic hand was then tested by an amputee by gripping objects of different shapes and sizes to determine its practicality of the prosthetic hand. The prosthetic hand was capable of picking up the object in 50% of its tests and showed a grip strength of 30.4 N and a response time of 4.5 seconds.


Introduction
Commercially available myoelectric prosthetic hands have proven to meet the requirements of a myoelectric prosthetic hand since all of their designs have been medically certified and are currently being used by amputees on a large scale [1]. Amongst the market for myoelectric prosthetic hands, the four most popular products are the Bebionic by Ottobock [2], the Michelangelo by Ottobock [3], and the i-Limb by Touch Bionics [4]. Whilst these three products vary from one another regarding the operation, they all contain certain crucial design features and aspects that are useful to analyze.
Each of these hands requires a custom-made socket containing electromyography (EMG) sensors and other electronics housed inside. Each hand contains DC motors, which are controlled by a control system. The control system determines the various types of grips to perform from the signal obtained by the EMG sensors. All of the hands discussed are capable of wrist rotation and have opposable thumbs [5]. Whilst the capabilities of prosthetic hands are growing, these hands are used primarily for gripping objects and performing common tasks such as pressing buttons and operating a computer mouse [6]. Prosthetic hands [2] are yet to have the control to perform more intricate [3] and complex tasks [4]. The comparison of the Michelangelo [2], Bebionic [3], and i-Limb [4] prosthetic hands can be seen in Table  1. In Table 1, the weight of the prosthetic hands has a range since some prosthetic hands are offered in different sizes. The range in price for the i-Limb Ultra is due to optional additions, which can be purchased with the hand. The hand load limit is the maximum weight, which could be held by the hand without opening. The grip strength was evaluated by using both a power grip (the grip used to grab a suitcase handle) and a lateral grip (the grip used to pinch a credit card between the index finger and the thumb). The adaptive digits column evaluated whether the digits of the hands curled when closing or whether they remained static. Consultation with amputees in testing the Touch Hand 2 [7] and the Touch Hand 3 [7] indicated that the prosthetic hands' functionality is often compromised compared to the prosthetic hook. With repairs conducted on these specific Touch Hands, it was found to be very time-consuming, and therefore modularity and ease of repairs were crucial.
The contributions of this paper are: • Modular fingers prosthetic hand and ease of repairs.

Mechanical Design and Grip Consideration
The hand parts were developed by creating models which were then 3-D printed [8] for all the parts according to the specified dimensions. Tolerances were added to ensure that the parts can be assembled and moved [9]. The shape was adjusted in the design for aesthetic purposes and weight reduction. Fillets were added to parts after performing finite element analysis (FEA) on parts, which experience the most tensile stress. Figure 1 and Figure 2 show the CAD models of the Touch Hand 4 with static digits. Figure 3 and Figure 4 show the CAD models of the Touch Hand 4 with synchronous digits:   Three digits were required in the hand to operate: the index finger, the ring finger, and the thumb. These three digits are slotted into the palm and screwed into the palm plate to ensure they are fastened. The index and ring fingers contain a motor link, which screws into the linear actuators in the palm whilst the thumbs contain a linear actuator within them-all of the digits house force sensors, which are located at the tips of the digits. Pathways for wires are made in the digits so that the wires can run into the palm.
The static and synchronous thumbs are assembled with printed parts and a Firgelli PQ12 linear actuator [10]. The thumbs are assembled rather than being printed in one piece because it was required to fit the linear actuator into the digit, which required the part to be assembled.  Figure 5 shows the CAD models of the static and synchronous thumbs. The index and ring fingers were evaluated together since both of these digits have the same kinematic design and operate in the same manner. Two designs were created for the index and ring fingers: the static digits and the synchronous digits. The index and ring fingers were designed and tested to be printed in one piece. However, the tolerances required to print moving parts with selective laser sintering (SLS) is 0.3 mm, which was too large for the designs of the digits. Printing the parts with a tolerance of 0.3 mm caused the joints to have too much radial and axial freedom. The digits have been designed to be assembled. Figure 6 shows the CAD models of the static index and ring fingers, and Figure 7 shows the CAD models of the synchronous index and ring fingers: The asynchronous version of the index and ring fingers has been designed. However, they were not capable of operating. The digits have been designed to be printed in one piece. However, the parts contained too small links to be produced with a tolerance of 0.3 mm. The digits could not be designed to be assembled because the design did not contain sufficient space to place pins for links. Even with the configuration considered for an assembly that was not easy to repair, it was found that the synchronous finger had less grip strength than the static fingers. Figure 8 shows the CAD models of the asynchronous index and ring fingers, and Figure 9 shows a cross-section of the CAD model of the asynchronous index finger. The digits were abandoned before the shape of the digits was refined.

Silicone padding for improved grip
Silicone padding was added to the surface of the palm and digits of the Touch Hand 4 to increase the friction of the surface. The silicone pads were cut into the shape of the palm and digits and glued on with cryoacrylate adhesive. The padding used on the digits is 3 mm thick, and the padding used on the palm is 5 mm thick.
The addition of soft padding onto the digits and palm increased the friction and allowed a softer contact surface to grip objects with. This padding allowed for an increase in the grip strength of the device. The implementation of padding also allows hand arches to implement hand reshaping, therefore allowing the device to adapt to the shape of objects more suitably since the padding would sink in areas where the hand has more pressure placed.
Surface friction tests were conducted to identify the hand's friction with the objects it is holding. The test was also performed on SLS nylon to compare the change in friction when padding is added.
The test consisted of placing an object of known weight across the tested surface to determine the static friction coefficient between the surface of the silicone pad and the surface of the object. The object was required to be pulled horizontally across the surface with an increasing force until the object begins to move. The force applied just before the object begins to move is obtained with a scale and used to calculate the static coefficient of friction.
The test was performed on three surfaces. The specification of the surfaces can be seen in Table 3 [11]. The values shown in the shore hardness scale column in Table 3 refer to the scale the shore hardness value was calculated.
The following equation is used to determine the force caused by friction [1]:

= •
(1) Where: • Ffs -Static frictional force • µs -Static coefficient of friction The direction of the static frictional force opposes the applied force onto the object [12]. Newton's first law states that every object shall remain motionless or in uniform motion unless an external force acts on it [12]. Since the surface friction test is performed with the container being static, the sum of the horizontal forces equals to zero. Therefore, it can be deduced that the magnitude of the applied force onto the container is equal to the static frictional force between the object and the surface.
The coefficient of friction of the different surfaces is tabulated in Table 4 and represented in Figure 10.  It could be seen from Figure 10 that the static friction coefficients of the silicone padding are significantly greater than SLS nylon. It is assumed that pad 1 had a significantly higher static coefficient of friction than pad 2 due to its low hardness and larger thickness. Like the role of fat in the human palm, the object sinks into the surface of pad 1 and, therefore, has more resistance to movement. Therefore, the use of padding on the Touch Hand 4 benefits its functionality by reducing slippage of objects with the same grip strength.

Electronics and control
EMG sensors are used to obtain signals from the amputee to control the Touch Hand 4. However, processing is required to determine the instructions from the signals. The research was performed to investigate optimal methods to use EMG sensors on the residual limb of transradial amputees and perform signal processing on the signals to obtain as many unique instructions as possible. This signal processing was required to allow the device to perform more gestures. The research was performed in collaboration with Touch Prosthetics to investigate the possible control signals used to operate the Touch Hand 4. The research investigated the following: • The effects of placing the sensors on different muscle groups.
• The different EMG signals are obtained by performing different movements with the hand. • The different signal features of the EMG signals, could be differentiated.
The EMG sensors were placed on different muscle groups, and different hand gestures were considered. The hand gestures were performed with the arm oriented in different positions. The magnitude of the signals obtained from the different muscle groups varied according to the size of the muscle groups. Bigger muscle groups, such as the biceps and the extensor muscle group, had a greater peak voltage than the smaller muscles, such as the flexor muscle groups. Muscle groups such as the flexor carpi ulnaris seemed not to have any unique response to the hand gestures, other than crosstalk from the extensor muscle groups. The crosstalk in the signals was seen to come from the extensor carpi ulnaris, which is located directly beneath the flexor carpi ulnaris. Figure 11 shows the EMG signals for the flexor carpi ulnaris and extensor carpi ulnaris when a fist gesture was performed.  The following discoveries from the results of the research made a large impact on the design of the EMG system: • The brachioradialis and the extensor digitorum produced the most significant signals from all muscle groups in the forearm. Responses in these muscle groups showed clearer signals relative to the other muscle groups. • The muscle groups in the forearm seemed to show minimal variation in the hand gestures that they had responded to. It seems unlikely that many unique gesture identifications can be obtained from the forearm muscle groups • The noise experienced in the signals when the muscles in the forearm were fatigued was significant and may cause problems if they had to be used in the control system without filtering.
The EMG and motor control systems were separated into two parts, which communicate via a serial line. This modularity allows for the EMG and motor control of the prosthetic hand to be independent, allowing each module to be upgraded independently. Figure 12 shows a basic logic flow chart of the control system implemented in the Touch Hand 4.

Fig. 12. Logic flow chart of Tough Hand 4 Control System
The process begins by using the EMG sensors to obtain control signals from the amputee. The EMG program processes the signal by means of filtering and mapping the results to the desired range. The instructions for the motor control system are then calculated and sent via serial communication. The motor control system uses these instructions alongside the force sensors readings to calculate the control signals required to be sent to the motors. The motors, housed in the hand chassis, actuate the digits and wrists according to the EMG system and force sensors.
The motor control calculation function calculated the motion that the linear actuator would move in one loop of code. Therefore, it was essential to ensure that the time taken between each loop of code was consistent, so the calculations for the linear actuator controls are considered time-invariant. The motor control system controlled the linear actuators in this manner because there is no feedback on the position of the linear actuators to indicate how far the linear actuator had moved. It was therefore required to move the linear actuator incrementally, so its position was known. The linear actuators use a servo pulse width modulation (PWM) to be controlled. Therefore, the pulse width was required to be calculated. The function first checked if the force sensor was making contact with another object. If the force sensor were not in contact, the instructions would be checked to determine whether opening or closing the digit is required. The function thus sets the linear actuator to move only enough for one loop of code. If the digit has made contact with an object, the program will check the force being applied relative to the instructions provided by the EMG system. If force is required to be applied, the necessary change was calculated. If the instruction indicates that no force should be applied, the digit started to open.
The actuator control boards were wired with the EMG control board and integrated into the Touch Hand, as shown in Figure 13. The electronic boards were attached to the socket that the amputee would wear.

Tests and Results
Tests with the amputee were performed in collaboration with Touch Prosthetics to evaluate the practical operation of the Touch Hand 4. Figure 14 shows the amputee operating the Touch Hand 4. The tests were performed on a trans-radial amputee. However, the amputee's residual limb was short, causing the brachioradialis and extensor muscles to be small as well. The socket which the amputee was using was too tight to consider fitting EMG sensors into the socket. Therefore, an EMG sensor was placed on the bicep of the amputee. Figure 15 shows the residual limb of the amputee.

Fig. 15. Residual limb of amputee
Before the tests could be performed, the EMG system was required to be calibrated to operate according to the amputee's muscles. An EMG sensor was connected to an Arduino Uno. which displayed the readings from the EMG sensors on a computer. The EMG sensor was placed on the amputee's bicep, and the EMG signal was observed. The signal was analyzed and used to calibrate the control system. Once the calibration of the control system was performed, testing of the Touch Hand 4 proceeded.
The Touch Hand 4 was tested by attempting to grip various objects to determine the responsiveness and functionality of the Touch Hand 4 in a more practical environment. Ten objects were used in the test to be gripped by the Touch Hand 4, which were of similar dimensions as the objects used in the Cybathlon [13]. Table 5 shows the items used in the tests, which grip was used, and the number of attempts performed to grip the object.  The glass tumbler was mentioned twice because it was gripped in two different orientations to test different grip patterns. Objects which failed were not capable of being picked up.
Objects that were not attempted were objects that were not attempted to be picked up due to the improper operation of the index and ring fingers. Figure 16 shows the amputee picking up the different objects listed in table 5.
The Touch Hand 4 was tested with an amputee to explain how the device would work if used as an everyday prosthetic device. The test performed to grip objects resulted in six tests being performed successfully, three tests being performed unsuccessfully, and three not attempted. A video showing some of the tests being conducted can be seen at https://www.youtube.com/watch?v=EBXVIsVMt5M.
Difficulty could be seen by the amputee in correctly orienting the prosthetic hand to grip objects in the test. The amputee found it difficult to correctly judge the distance between the objects and the hand and therefore found it difficult to orient the device to grip the objects correctly. Further training by the amputee with the device would be required to improve the Touch Hand 4 more efficiently.
The rotation of the ring finger was not corrected when the tests were performed with the amputee. Therefore, the index and ring fingers were incapable of meeting the thumb to make a closed grip to grip small objects. This problem led to the inability to grip or pinch small objects, such as pinching the cloth or picking up a book or notepad. Since the gap between the digits could be visibly seen, it was clear that the Touch Hand 4 would not be capable of gripping a small object such as a pen, a nut, and a credit card. Therefore, these tests were not performed. After the tests were conducted, the finger orientation was corrected, and the padding on the fingers was adjusted, which allowed these objects to be gripped more successfully.

Conclusion
The analysis of the digits showed that the operation of the synchronous digits is not capable of moving with the designed kinematics of the Touch Hand 4. The kinematics of the synchronous thumb does not function as expected due to high shear stress acting on links responsible for joint coupling, causing the rotation of the joints to not operate as expected. The problems experienced by the synchronous index and ring fingers are expected to be due to the tolerance added to the joints of the links, causing them to move unexpectedly and orient themselves in a manner that causes the digits to seize. The static digits operate smoothly. However, the design of the index finger requires to be altered to accommodate for its inability to meet the thumb. The configuration of the motor control system was altered afterward to ensure that the ring finger is capable of rotating its full range. The changes, however, were made only after the tests were performed with the amputee.
To determine the capability of the Touch Hand 4 as a prosthetic device, the Touch Hand 4 was compared with other prosthetic devices in terms of performance and specification of the devices. The comparison was performed with commercial prosthetic devices and with the previous Touch Hand prosthetic devices. Table 6 shows the comparison of the prosthetic hands: The weight used for the Bebionic and i-Limb Ultra is the weight of the large-sized hands. The price of the Touch Hand devices is the cost estimations of only the components required to assemble the devices. From Table 6, it is observed that the Touch Hand 4 has a better grip than the Touch Hand 1 and Touch Hand 3. The grip strength of the Touch Hand 4 is a similar range to that of the Bebionic prosthetic hand. The closing time of the Touch Hand 4 is longer compared to the other prosthetic hands. Other actuators could increase the grip strength and decrease the closing time, yet these actuators will cause the cost of the Touch Hand 4 to increase dramatically. The Touch Hand 4 is heavier than the Michelangelo prosthetic hand but lighter than all the other prosthetic hands, except for the Touch Hand 2. The Touch Hand 4 was slightly lighter yet had some constraints with the efficiency of gripping. The Touch Hand 4 has the lowest price of all the prosthetic hands. The design of the Touch Hand 4 has allowed for the possibility of more grips to be performed, compared to the previous prosthetic hands, allowing for functionality and being still aesthetically appealing.
The research contributions of this research and paper were achieved as follows: • Modular fingers prosthetic hand, designed with synchronous digits. The Touch Hand 4 can pick up 50% of objects placed forward in its tests, can output a grip strength of 30.4 N, and has a response time of 4.5 seconds. The Touch Hand 4 is also easy to repair based on the design and easy interlinking of components. • Amputee performance tests which showed the volunteer picking up the glass tumbler spherical and cylindrical, table tennis ball, apple, grocery bag, and plastic tube. The Touch Hand 4 did not pick up the cloth, A4 book, A5 notepad, pen, M10 nut, and credit card. Hence, the Touch Hand 4 was 50% successful in picking up the objects placed forward.