Nowadays no powered prosthetic shoulders are available on the market, probably because shoulder disarticulations are much less common than other levels of amputation1: current prosthetic solutions comprise at most passive revolute joints with locking mechanisms or frictional ball-and-socket articulations. The orientation of these devices can be adjusted with the help of the sound limb, if present, or using fixed points in the surrounding environment, in the case of bilateral amputees. Because of their restricted mobility and the intricacy of their control (with the exception of humeral passive rotators), these solutions require the amputee to execute common tasks with unnatural control strategies and/or preclude the possibility to perform important activities of the daily living. For instance, the amputee must rely on to gross body movements and/or special aid-tools to compensate for the lack of sufficient functionality of the devices, above all for the activities requiring above-the- shoulder reaching tasks.
Prototypes of electrically actuated shoulder mechanisms have been proposed in the literature.2–6 Gow et al.2 proposed a shoulder joint with 1 degree of freedom (DoF) for the upper-arm flexion, integrated in the Edinburgh Modular Arm System (whose kinematic basic structure was inspired by the pioneering gas-powered Simpson Arm developed in the 1970s3). Cattaneo et al.4 developed a shoulder articulation with two coupled DoFs that – based on a differential gear mechanism – can orient in space the upper arm. An electromechanical shoulder articulation composed of a 2-DoFs articulated mechanism (for the upper-arm flexion and abduction) and a humeral rotator (with 1 DoF for the humeral intra-extra rotation) was introduced instead by Higashihara et al.5 Finally, Weir and Grahn6 developed a powered humeral rotator (developed for proximal and medial transhumeral amputees). Unfortunately, none of these prototypes seemed suitable for commercialization, for reasons ranging from excessive size, weight, cost, unpleasant appearance, or difficulty of control.
To overcome the current lack of active shoulder articulations available on the market, we developed an actuated shoulder prosthesis with 2 DoFs, suitable for interscapulothoracic and first-proximal transhumeral amputees.7 The articulation allows the elevation of the upper arm in any vertical plane passing through the mechanism's centre of rotation, whereas a further frictional passive joint allows the humeral intra-extra rotation. To assure its practical application, the new shoulder is fully compatible (both from the mechanical and electrical viewpoints) with commercially available components, such as the INAIL EMEI-20 elbow joint,8 as well as the Otto-Bock prosupination unit and hands. Moreover, the development of the new shoulder was based on a systematic methodology based on a multicriteria optimization of the design. Indeed many contrasting features of different nature are required in designing prostheses; in particular, the functional performance must be balanced with the actual wearability of the resulting artificial arm, otherwise the risk is to design a high-performance system finally refused by amputees, e.g., because of a high weight, an intricate control, or bad appearance.
This article provides an overview of the development of the mechanism prototype, illustrates its integration in the INAIL prosthetic arm, and reports the main outcomes of the tests performed during a preliminary clinical application.
DEVELOPMENT OF THE SHOULDER
The development of the actuated shoulder, in its mechanical and electronic aspects, proceeded through four steps, which are described below.
KINEMATIC MODEL OF THE PROSTHETIC SHOULDER
To determine the topology and the geometry of the shoulder mechanism, we developed9 and applied a procedure named “Determination of the Optimal Prosthesis Architecture (DOPA).” The procedure can be seen as a multicriteria optimization technique specifically developed to determine the best prosthetic solution, given the contrasting characteristics required by an upper-limb prosthesis, i.e., functionality (e.g., high performance in terms of mobility, payload, velocity), and wearability (e.g., tolerable weight, humanlike size and appearance, easy control).10,11 The core of the method consists of kinematic and dynamic analyses of alternative models of upper-limb prostheses simulating the execution of important motor tasks. For the present application, 19 motor tasks (e.g., drinking from a glass, reaching an object at the head height …) were selected as the most significant activities of the daily living for the functional autonomy of upper-limb amputees (a bilateral amputee's profile was assumed as a reference to guide the design process). The optimal prosthesis architecture is the simplest prosthetic solution (i.e., with few actuated joints working at low power) able to execute the motor tasks with a sufficient accuracy. The minimum accuracy considered acceptable was defined for each task in terms of maximum error of the hand pose with respect to a reference trajectory representing the correct execution of the activity.12
The prosthesis models that we considered differed in the shoulder mechanism architecture, i.e., the number (1 to 3), the type (actuated or passive), and the arrangement of the revolute joints that form the articulation. Since the prosthetic shoulder was intended to be introduced in the current INAIL prostheses, all the models comprised also the elbow joint, the wrist prosupination unit, and the hand, which was considered as a simple rigid body (Fig. 1).
From the results of the application of the DOPA procedure, whose details are discussed in Ref. 9, the shoulder articulation model with the simplest and lightest architecture consisted of 1) two powered revolute joints, J1 and J2 in Figure 1e, with orthogonal axes intersecting at point C and 2) a passive revolute joint JH, for the control of the humeral intra-extra rotation. The active joints, respectively, actuate the rotation θ1 about an axis fixed to the prosthesis socket (which is vertically directed when the patient is in the upright posture) and the rotation θ2 about a horizontal axis, whose orientation is determined by θ1. In other words, the rotation of the first joint selects the vertical plane passing through C along which the second joint elevates the upper arm, thus actuating a spherical motion in terms of azimuth and elevation coordinates. The humeral intra-extra rotation angle can be controlled before (after) the activation of the other joints' motors to provide a proper predetermined (adjusted) orientation of the elbow joint axis. This can be easily done with the help of the sound limb (if present) or by resting the prosthesis against a fixed point (e.g., the edge of a table).
DESIGN AND PROTOTYPING
The above mentioned kinematic and dynamic simulations (implemented in a Matlab® code) also provided some of the technical specifications needed to guide in the mechanical design of the device, e.g., the ranges of motion and the maximum powers required for J1 and J2 to perform the given motor tasks. Other important specifications were 1) the shoulder bulk should be compatible with the dimension of a human-like limb; 2) the articulation weight should be <1 kg; 3) the inverse motion of the arm due to gravitational effects when the joint motors are turned off should be prevented (possibly avoiding the use of active devices, not to drain the battery energy in the limb's stall position); and 4) accidental impacts on the prosthesis should not cause pain to the patient's residual limb.
Feasibility studies of different mechanisms (open-chain, closed-chain, and hybrid open/closed-chain architectures) implementing the model were performed. In particular, the mechanisms were tested through further simulations of a full prosthetic arm that were executed in a commercial software for multibody dynamic analyses (MSC.visualNastran 4D, MSC Software Corporation, Santa Ana, CA). The retrieved data were used to design the actuators and the power transmission chains. It is worth noting that, although the INAIL EMEI-20 elbow can lift up to 500 g object hold in the terminal device, the shoulder articulation has been designed to allow the prosthesis to manipulate 1 kg object, to take into consideration further possible developments of the elbow joint with enhanced performance. The main criteria assumed for selecting the most suitable mechanism were 1) small size motors are preferred for their low-energy draining as well as for their limited weight and bulk; 2) power transmission chains with few velocity-reduction stages are considered better due to the containment of dissipative effects; 3) lightweight solutions with a compact design and few components are globally appreciated.
The schematic representation in Figure 2 reports the layout of the design solution that was selected following these criteria. Joint J1 is a simple actuated revolute joint (Fig. 3), whereas J2 is basically composed by an inverted slider-crank mechanism (Fig. 4), where the slider is the driving elements and the J2 chassis is the follower. This solution makes it possible to obtain a high torque without the use of a speed reducer with a great transmission ratio (thus without a high bulk and many dissipative effects and clearances, which entail a low efficiency).
After the selection of the commercial components and dimensioning the links to be machined, the articulation was finally prototyped (Fig. 5).
Joint J1 is driven by a commercial DC motor (2224U006CR, Faulhaber) with nominal voltage UN = 6 V, maximum output power Pmax = 4.55 W, and stall torque MH = 21.2 N · mm. A magnetic encoder (IE2–64, Faulhaber, Faulhaber Group, Germany) is embedded in the motor to enable a closed-loop control scheme. The total reduction ratio of the kinematic chain is i1 = 1:1050, provided mostly by a commercial harmonic drive reducer (HFUC-8-100-2A-R_0-305535, Harmonic Drive® Technologies, Harmonic Drive, LLC, Peabody, MA). This component provides a reduction ratio iHD = 1:100 and is very compact, thus allowing joint J1 to have a limited axial length (46 mm, excluding the element bolted to the socket). The other contribution to speed reduction is provided by two helical spur gearings. A commercial cross roller bearing (RE2508 UU CC0 P0, THK CO., LTD, Tokyo, Japan) supports the weight of the whole prosthetic arm as well as the reaction forces due to the bending moment that loads the pivot connecting the arm to the socket during the execution of the tasks.
Joint J2 is driven by a commercial DC motor (2342S012CR, Faulhaber, with UN = 12 V, Pmax = 17.00 W, MH = 80 N · mm) with an embedded encoder (IE2–64, Faulhaber). The higher rating voltage of the J2 motor is due to the need of maintaining within practical values the current drained when generating high torques. With reference to Figures 2 and 4, the motor drives the joint by means of an actuation transmission that includes helical gearings, a commercial planetary gearbox with reduction ratio iPG = 1:14 (PK22-2Stages, IMS-Gear GmbH Donaueschingen, Germany) and a commercial ball screw with diameter 6 mm and pitch 2 mm (SH 6X2R, AB, SKF, Sweden). The slider coupled with the screw and supported by linear ball bearings (LBBR 5, SKF) drives the rotation of the J2 chassis around the joint axis, according to the scheme of Figure 2, resulting in elevation of the artificial arm. The total reduction ratio is not constant, depending on the configuration of the closed-chain mechanism: the maximum value achieved is i2 = 1:1187.
An “anti-return system,” that is a device that makes the inverse motion impossible (i.e., the undesired motion caused by the payload when the prosthesis should be still), is also present in J2. A shaft with a cone-shaped end is introduced in the kinematic chain and can come into contact with a conic hub (having a high friction coefficient) free to translate in one direction. When the motor is turned off, a solenoid is automatically activated by a small and brief impulse of current and pushes the element toward the conic shaft; here, the compression between the two bodies is maintained—without current draining—with a force sufficient to generate a friction torque that does not allow the shaft to rotate, thus locking the kinematic chain and preventing J2 from the indirect motion. Another brief current impulse, automatically commanded by the control circuit before the activation of the motor, makes the friction surfaces separate, thus allowing the transmission chain to work correctly during the direct motion. In this phase, the two bodies are kept separated by the action of two springs.
A safety device, basically composed of a steel pin and a flat spring, allows the upper arm to freely rotate about the J2 axis when the external load acting on the prosthesis is higher than a given threshold value (related to the preload of the spring). In case of a patient's fall or other accidental impacts on the prosthesis, the corresponding shock is therefore not transmitted to the residual limb preventing the patient to suffer possible pain.
Finally, the passive revolute joint is a friction disc (in Teflon®) currently used at the INAIL Prostheses Centre. It is simply integrated at the extremity of the second joint J2. Its friction torque can be regulated by adjusting the preload of the clamping force of a locknut.
A series of tests were performed with a test bench specifically developed by the authors to experimentally characterize prosthetic devices. In particular, the joints' maximum performance, in terms of velocity and payload, and their mechanical efficiency η were estimated with an original procedure detailed in Ref. 13.
A scheme of the test bench is reported in Figure 6.
The main functions of the test bench were to 1) drive the joints according to given motion laws, 2) measure the instantaneous motor voltage rating, and 3) measure the instantaneous current draining.
The main unit is a data acquisition board, namely the CompactRIO (National Instruments Corporation, Austin, TX), which is directly connected to a PC via LAN. The board hosts I/O modules for digital input, digital output, and analog input signals that cannot operate directly with sensors and motors. Therefore, specific electronic components were placed on the “Shoulder Board” for 1) driving the motor by pulse-width modulation (PWM) signals (VNH2SP30, STMicroelectronics, Geneva, Switzerland), 2) transforming the voltage of the encoder signals from the original range of 0 to 5 V to the appropriate level 0 to 12 V required by the digital input module (Photocouplers), and 3) measuring current (with the current-sense amplifier MAX472, Maxim Inc., Sunnyvale, CA).
The mechanical efficiency of the joints vary with velocity and payload; their average values, with no lubrication of any moving components, were about η1 = 0.25 and η2 = 0.45 for J1 and J2, respectively: this means that the 75% and 55%, respectively, of the energy put in the two actuated mechanisms is lost by dissipative effects. Incidentally, the mechanical efficiency of the INAIL EMEI-20 elbow joint (grease lubricated), which is a more mature product, is about ηelbow = 0.55 on average, whereas the efficiency of the shoulder mechanism of the Edinburgh Modular Arm System is estimated at about 0.4.2Table 1 reports the most significant data relative to the shoulder joints J1 and J2 and to the EMEI-20 elbow joint. For the sake of the data interpretation, Table 2 reports the correspondence between the mass of the object hold in the terminal device and the maximum resistant torque at the elbow and J2 axes, respectively (namely the maximum torque due to the elements' weight). Note that J1 axis is not loaded by gravitational contribution for a patient in the upright posture.
CONTROL UNIT DESIGN
To integrate the shoulder articulation in the current INAIL prosthesis (which comprises at most an active elbow, prosupination unit, and hand) a new on-board control unit was required. The electronic system was designed and manufactured based on 1) the data retrieved from the bench tests (e.g., current required by the motors), 2) the technical specifications relative to the control strategies to be implemented, and 3) the need to maintain the compatibility with commercially available prosthetic components.
As shown in Figure 7, the control-unit is formed by five main elements: the microcontroller, the Bluetooth module, the current sense, the motor drivers, and the antireturn system driver. The architecture is therefore centralized, with the core of the system represented by the microcontroller.
Concerning the microcontroller, different technical specifications guided the choice of a proper commercial solution:
- The ability to drive up to five DC motors (each controlled by one or two PWM signals), because the full prosthetic arm must include the actuated shoulder, an elbow, a prosupination unit, and a hand;
- The ability to acquire at least five analog input signals, because a) the control strategies typically implemented in active upper-limb prostheses are based on signals provided by up to four EMG sensors, and b) one signal must be available to control the current running through the motors;
- The availability of at least 128 byte of EEPROM memory;
- The availability of a proper interface device for the communication with an external workstation, to allow certified prosthetist/orthotist (CPO) to set functional parameters of the prosthesis (e.g., the control strategy needed by the amputee, the EMG signal thresholds, the maximum current thresholds …) and to monitor runtime signals (e.g., the EMG levels read by the sensors, the current drained by the motors …);
- The availability of at least 16 kB of program memory, because the architecture of the control unit is centralized and therefore the microcontroller firmware results rather complex due to all the functional aspects that it must manage (signal acquisition, signal processing, motors control, control of the antireturn system, communication …).
The commercial microcontroller PIC18F4431 by Microchip Technology Inc. (Chandler, AZ) meets all these requirements and was thus selected for the development of the control unit. In particular, the PIC18F4431 provides 10 PWM outputs, 9 Analog to Digital Converter channels, 256 byte of EEPROM memory, 16 kB of program memory, 1 asynchronous serial port UART and 1 synchronous serial port serial peripheral interface.
Aside from the microcontroller, a Bluetooth wireless communication module was hosted on the board. The Bluetooth module allows an easier interaction between the CPO and the control unit, for setting parameters and monitoring signals. The interaction between the CPO and the control unit is provisionally provided through the Microsoft Hyperterminal (Microsoft Corporation, Redmond, WA) software.
As current sensor, we adopted a simple MAXIM MAX1797 together with a current sense resistor. The MAX1797, based on its operational amplifier, provides to the microcontroller a voltage proportional to the current running through the motors. This allows the microcontroller to manage the maximum power supply to the motors.
The control unit also hosts six drivers:
- Three low-power VISHAY SI9986 (Vishay Intertechnology Inc., Malvern, PA), used for the wrist prosupination, the hand, and the antireturn device;
- Two middle-power INFINEON TLE 5206-2 (Infineon Technologies, AG, Germany), used for the elbow and the J1 joint of the shoulder;
- One high-power STMicroelectronics VNH3SP30, used for the joint J2 of the shoulder.
In particular, the choice of the driver for J2 was a critical point of the design of the control unit. In fact, this driver is required to provide an electrical power much higher than the drivers of the other motors (2.2 A at 12 V, Table 1). The VNH3SP30 driver can tolerate voltages up to 40 V, currents up to 30 A, and, remarkably, has a very low internal transistor resistance (Ron = 34 mΩ) that leads to low internal voltage drops. Therefore, the decrement of the joint velocity (proportional to the actual motor supply voltage) due to electrical effects consequent to high values of the driving torque (proportional to the drained current) is negligible.
The firmware for the microcontroller was specifically developed for the implementation of the control strategies currently adopted at the INAIL Prostheses Centre. As such, the activation of the powered joints can be controlled by the patient through EMG sensors, FSR sensors, or microswitches. The velocity of rotation for each motor depends on the amplitude of signal coming from these sensors according to three different modalities:
- Single threshold: if the EMG signal level is higher than a given threshold, the motor runs at a constant velocity;
- Double threshold: as stated for 1), but with two thresholds and two velocities;
- Proportional control: the velocity of the motor is proportional to the EMG signal level.
The active motor can be selected through three different strategies:
- Third electrode selection: a separate myoelectric signal or a microswitch is specifically dedicated to this operation;
- Cocontraction: the generation of two EMG signals, corresponding to the simultaneous contraction of two muscles, allows the selection of the motor to be activated;
- Voice control: regarding this last option, through the voice control a patient can directly select with his own voice the joint to be activated by pronouncing a predefined set of words (named “command vocabulary”). Specifically, a throat microphone acquires the patient's voice and transfers it to a specialized unit mounted on the control electronic board. The specialized unit recognizes the commands vocabulary. On recognition of a valid command, the voice controller sends a command to the control unit via UART serial port indicating the motor to activate. The control unit then enables the indicated motor. In the current prostheses, the VR-STAMP module (Sensory Inc., Sunnyvale, CA) was chosen as the voice controller.14 This module can implement both speaker dependent or speaker independent recognition. Presently, only the speaker dependent solution is implemented on the firmware of the VR-STAMP module.
For prostheses with more than three actuated joints the voice control can be particularly effective, because it can bypass the constraint of the other selection strategies, which impose to follow a fixed cyclic sequence of activation of the motors.
Finally, the configuration functions and the monitoring functions (accessible via Bluetooth interface) are also implemented on the firmware.
A prosthesis equipped with the shoulder prototype and the on-board control unit was tested by a patient during 2 test days. During the first day, the patient's opinions about the comfort, control, appearance, usability, and performance of the prosthesis were collected to retrieve suggestions about possible improvements of the shoulder articulation. Because the amputee complained about pain in specific regions of the residual limb, during the second day the pressure exerted by the socket on those body regions were measured.
MATERIAL AND METHODS
A 32-year-old man amputee, initials G.F., participated in the experiments, after giving his informed consent. G.F. sustained a first-proximal transhumeral amputation of the right arm in 2003 due to a work-related injury. He presents a very short residual limb without functional flexion-extension (Fig. 8). G.F. was fitted with his first prosthesis in 2004 and since then he has been using the prosthesis continuously.
In his daily life, G.F. uses a myoelectric prosthesis equipped with hand (Sensorhand™ Speed Otto-Bock, Dunderstadt, Germany), prosupination unit (Otto Bock 10S17), and elbow joint (INAIL Elbow EMEI-20); a passive ball-and-socket joint is present at the shoulder level (Fig. 9a). For the control of the prosthesis, he uses two EMG electrodes (over middle deltoid and trapezium) together with a traction-microswitch engaged with an antiretropulsion of the shoulder girdle. With the traction-microswitch he selects the motor to activate (following the cycle: hand – prosupination unit – elbow), whereas with the two EMG electrodes he activates the rotation of the motor selected either in the clockwise or in the anticlockwise direction.
The shoulder prototype and the new control unit were fitted into G.F.'s prosthesis with minimal interventions. The active shoulder substituted the passive ball-and-socket joint (Fig. 9b). The control unit and the batteries were hosted in the forearm and in a socket cavity, respectively. The harness and the socket were not modified in shape and dimension.
No changes were required at, and below, the elbow level. The prosthesis control strategy was not modified.
Test-Day 1—Training and Collection of the Amputee's Opinion
G.F. used the new prosthesis for about 3 hours. First, he executed single-joint movements to gain confidence in the prosthetic arm. Then, he executed reaching tasks mostly in the sagittal plane. Initially, the activity consisted in reaching a bottle of water of 0.5 kg, placed on a shelf at chest level, and bringing it either to the mouth or at thigh level. Once G.F. declared to be confident, the bottle was moved at shoulder- and finally at eye level (Fig. 10). At least seven successful trials were repeated for each reaching level. At the end of the test, G.F. completed a VAS questionnaire (Table 3) to collect his feelings about the prototype and to guide future developments.
Test-Day 2—Socket-Residual Limb and Socket-Trunk Pressure Measurement
At the end of the first test day, G.F. reported a painful sensation in two areas, i.e., around the coracoids process, and at the lateral thorax, at the level of the inferior border of the socket. To have a reference in the development of a different socket, we then measured the pressure suffered by those two areas. In particular, G.F. was asked to complete the following six activities, three times each, and to declare when he was suffering pain:
- Shoulder elevation-depression in the sagittal plane (sh0) with 1a) the elbow fully extended and without any additional weight on the hand (sh0el0w0), 1b) the elbow fully extended and with a 0.5 kg object on the hand (sh0el0w0.5), 1c) the elbow flexed 90° and with a 0.5 kg object on the hand (sh0el90w0.5);
- Shoulder elevation-depression in a plane 45° anterior to the frontal (sh45) with 2a) the elbow fully extended and without any additional weight on the hand (sh45el0w0), 2b) the elbow fully extended and with a 0.5 kg object on the hand (sh45el0w0.5), and 2c) the elbow flexed 90° and with a 0.5 kg object on the hand (sh45el90w0.5).
During these activities both the socket-body pressures and the correspondent shoulder elevation-depression angles were measured.
The socket-body pressures were measured with a Novel Pedar-X system (sampling frequency 100 Hz; Munich, Germany). Two Pedar-X pads were mounted on the inner socket, at the level of the two painful areas (Fig. 11a). The pads were kept in position by a thin layer of Fixomul transparent single-sided tape (BSN Medical, Hamburg, Germany). The pads were connected to the Novel data logger that stored the pressure data. The minimum sensibility of the pads was 15 kPa, whereas the sensitivity was 0.5 kPa.
The shoulder elevation-depression angle (humerus relative to the thorax) was measured with an MTx system (Xsens Technologies, Enschede, The Netherlands, Fig. 11b), using the protocol presented in Ref. 15 with only two modifications: 1) the scapula sensor was positioned over the socket; 2) the scapula anatomical coordinate system was defined parallel to that of the proximal humerus during the static calibration phase. The shoulder elevation-depression angle was measured at 100 Hz. The MTx and Pedar-X systems were synchronized via-hardware by means of the Novel SyncBox, which received the sampling clock from the MTx.
The peak and the median pressure per elevation angle was finally computed for each activity, considering all the three repetitions. The higher the median, the more consistent the pressures over the residual limb and thorax.
Test-Day 1—Collection of the Amputee's Opinion
The questionnaire main outcomes are summarized in Table 3. The prosthesis overall weight (2.8 kg) is tolerable in the resting position and well balanced with respect to the sound limb. On the contrary, the load at the residual limb during the flexion of the arm determines an excessive strain (above all when elevating the arm with the elbow completely extended). The velocity of the shoulder joints (Table 1) is good, whereas the payload should be increased to 1 kg. The noise is unacceptable. The control strategy is simple but not effective, because it is too slow due to the high number of motors to be sequentially controlled. Finally, its appearance is acceptable.
Test-Day 2—Socket-Residual Limb and Socket-Trunk Pressure Measurement
Irrespective of the elevation-depression angle, the socket-residual limb median pressures were always lesser than 15 kPa and the peak pressures below 20 kPa.
For what concerns the socket-trunk pressures, the greatest peak and median values always corresponded to elevation-angles higher than 90° (Table 4). Generally, higher peak and median values were found for the sh45 activities compared with sh0. The activity sh45el0w0.5, resulted the most critical, both considering the peak (42.5 kPa) and the median pressures (22.5 kPa).
DISCUSSION AND CONCLUSION
The shoulder model designated by the DOPA procedure ensures a good functionality (thanks to its two active and one passive DoFs) and a good wearability with respect to all the other models. The proper arrangement of the J1 and J2 axes facilitated the design of fairly small-size and lightweight joints (mainly J1), without compromising the performance of the articulation. The passive joint is extremely small, light, and easy to control.
From a technical viewpoint, the shoulder prototype showed to be consistent with expectations, both considering range of motion, payload capability, and electromechanical performances. Thanks to its compact design, the shoulder could be easily integrated in the prosthesis of G.F., who presents a prominent residual limb that reduces the space available for the components. This proves the clinical applicability of this new prosthetic device.
However, the clinical tests revealed margins for improvements.
First, the sequential control strategy via traction-microswitch did not allow G.F. to exploit the high mobility of the prosthesis. The voice control system, which allows the patient to directly select the joint to be activated, seems a better solution. This strategy was not used during the tests because we wanted to verify if the standard control strategy adopted by G.F. could have been feasible to control the new prosthesis as well. It should be noted that the voice control system is a flexible tool and allows the implementation of various “sub-control” strategies. Aside from selecting a specific joint, the amputee could also control the rotation of the selected motor in one direction or the other, through an appropriate word; alternatively, the vocal inputs could activate the execution of predetermined “complex” trajectories (requiring the simultaneous activation of different joints' motors) stored in the memory of the control unit.
Second, the noise of the mechanism will need to be reduced by reconsidering the design and/or the quality of the manufacturing of some components. The critical parts to be rethought are the couplings of some moving components, where clearance, local compliances, and narrow geometric tolerances play an important role for the correct functioning of the mechanism. Apart from reducing the noise level, these interventions are expected to improve the joints' mechanical efficiencies as well, with positive consequences on the functional performances and the energy consumption.
Finally, to guarantee the proper wearability of the prosthesis, the load on the patients' residual limb and thorax should be reduced. Since the peak socket-residual limb pressures (measured on G.F.) were associated with pain, a new socket should be conceived to cut them always below 20 kPa. Moreover, for the socket-thorax pressure, the target for the new socket will be to reduce the median and peak pressure at about 20 kPa (considered by the patient to be a tolerable value at the thorax level) during sh45 tasks. Not surprisingly the higher peak and median values were found for the sh45 activities, because the weight of the arm in that configuration and at 90° of elevation is greatly supported by the thorax.
The decrease in the pressures will be obtained both through the moderate increment of the socket-thorax contact areas and/or through the insertion of silicon pads over the painful regions.
Following G.F. comments and the measurements, a revised version of the prosthesis is under development. Once completed, the tests will be extended to bi-lateral amputees, who are expected to profit the most from the new shoulder.
On the whole, results showed that the new shoulder is really applicable in clinical practice.