A Perspective on Rehabilitation Through Open-Source Low-Cost 3D-Printed Distal to the Wrist Joint Transitional Prosthetics: Towards Autonomous Hybrid Devices

Abstract

Over the years, patients with partial hand loss have relied on expensive prosthetics to recover some of the hand functionality. Fortunately, advancements in additive manufacturing desktop solutions allow transitional prosthetics prices to decrease. Therefore, the present work focused on providing a basic overview of the field and available low-cost 3D-printed upper-limb prosthetic devices. The aim was to develop a basic frame of reference on the field of transitional partial hand prosthetics. Concomitantly, this study also highlights additive manufacturing techniques on which further research can be done whilst helping to provide a new variant for an upper limb prosthetic device. The initial stages, current practices, and future possibilities were considered. Researchers and industry can utilize these findings to develop additional variants for the benefit of patients suffering from partial hand loss.

1. Introduction

The 3D-printed (3DP) Upper Limbs Prosthetics (ULP) offer a low-cost transitional or substitute to commercial and traditional devices. Commercial solutions for various levels of Partial Hand Amputation (PHA) are available, but require a time interval from prescription to fit [1]. Alternatively, an overview of the transitional PHA options manufactured through 3D printing with high availability, is proposed in this work.

The combined development of Computer Aided Design graphics (CAD) and Additive Manufacturing (AM) technologies has led to the design and fabrication of the first 3D-printed (3DP) ULP, the Van As and Owen’s Robohand design, for a child in 2012 [2]. This, together with the evolution of hobby electronics, has resulted in quite a few low-cost ULP devices [3].

The ULP open-source design files, as well as the extra parts, are widely available. Rapid prototyping (RP) and direct digital manufacturing have led to a resurgence of interest in children’s orthoses and prostheses. The disruptive nature of 3DP has led to the development of multiple designs for 3D ULP devices [4].

The high cost of therapy for stroke patients in developing countries has led to seeking other solutions. So, 3DP alternatives are used to reduce cost and increase accessibility [5] in rehabilitation for stroke patients with motor function loss. They are used especially for the upper limb rehabilitation of the proximal shoulder vs. elbow and the distal wrist vs. hand joints [6].

The high availability also poses a problem because individual users of 3DP ULPs cannot be surveyed and studied easily. For example, paediatric studies acknowledge that 3DP is a useful tool in teaching, as well as in developing procedural materials in the medical field, but they also underline the need for more research and data generation in this subject area [7].

Although 3DP ULPs are low-cost and bespoke, their disadvantages include durability issues, insufficient grip strength, reproducibility, and appeal [8]. Also, medical applications that imply AM technology need to consider the challenges brought by mechanical anisotropy and manufacturing orientation. The work of [9] states that Fused Filament Fabrication (FFF) and Polyjet technologies may be used for RP, while Selective Laser Sintering (SLS) printing is the solution for functional medical applications like assistive devices with good mechanical stability.

The aim of the present work is to discuss relevant findings in the field of low-cost upper-limb 3D-printed prosthetics for patients with partial hand loss. Many studies have approached the design of prosthetic limbs. In the case of the present work, a narrative overview based on a methodical approach is proposed. In the following sections initial solutions, current options, and ongoing efforts for future development are discussed.

2. Methodology

The considered studies were extracted from the Scopus database using the search strings “partial hand prosthetic” and “partial hand prostheses”, respectively. As seen in Figure 1, all accessible articles written in the English language with no time restriction were scrutinized. All book chapters and reviews were excluded. After title and abstract analysis, a total of 53 articles were examined. To develop the initial body of literature, the Connected Papers online engine was used, and the relevant reviewed references were doubled. Other articles associated with the topic and considered important in the authors’ experience were included. The authors also used the Scale for the Quality Assessment of Narrative Review Articles (SANRA) methodology [10] to improve manuscript quality. The procedure allows authors and reviewers to assess if a review is done properly by answering a series of questions. The result is an improved version of the review work. The instrument was used in medical narrative reviews, but the authors considered it when writing the current work as it relates to a multitude of fields. The paper provides an overview of the partial hand prosthetic field with a focus on open source and additive manufacturing technologies. This is how the first phase of the research was concluded.

In the second stage of the study, a compiled list of recommendations of 3DP ULP for each level of amputation is presented. The contents are based on querying online databases found as described in [11] and then shortlisted with a criterion method similar to [12]. Existing projects were filtered using the authors’ proposed criteria. A total of 21 databases were screened. A set of common keywords were utilized. The results are depicted in subsequent figures; the “Link Grabber” extension for the Chrome browser was employed for data gathering. The “Open Multiple URLs” Chrome extension was used to load multiple links at once. Relevant projects were sorted manually by using the “Save My Tabs” Chrome extension.

3. Control and Signal Processing Algorithms

The ULP devices have either a body-powered mechanism or an externally powered system to control the prosthetic movement. This section highlights some of the outcomes of the literature review regarding electrical actuation, as illustrated in Figure 2. As the mechanical variants are well documented, the control systems currently available have been analysed. Some advantages and disadvantages are presented.

3.1. IMU

Inertial Measurement Unit (IMU) data are collected to help trans-humeral amputees to control prosthetic devices. People with tetraplegia use shoulder movement to control a robotic limb. The movement is measured by an IMU, and the data are processed by Principal Component Analysis (PCA). The accuracy of the system is 89%. The decision may be translated to movement of the prosthetic hand [13]. A more recent study is processing IMU data with PCA to extract kinematics. The synergy components were analysed by a Long-Short-Term-Memory (LSTM) network, which is a type of Recurrent Neural Network (RNN). Training it results in motion prediction [14]. Furthermore, motion data can power Body-Machine-Interface (BOMI). Based on autoencoders, the control of a 4 degrees of freedom (DoF) robot arm is possible. Applications include aiding reduced motor function patients [15].

3.2. EMG

Feature detection on PHA with Electromyography (EMG) is influenced by wrist position, making the pattern recognition more complex. Although there are some debates about which is better, either Linear Discriminant Analysis (LDA), or neural networks, it is agreed that LDA and linear and non-linear neural networks perform better then Quadratic Determinant Analysis (QDA) for pattern recognition (PR) with the EMG of partial hand amputees [16].

Filtered EMG signals with a Machine Learning (ML)-based Vector Support give an accuracy ranging between 80 and 90% for some of the tested movements. It is considered that the fusion of multiple methodologies can achieve higher accuracy [17].

Classifiers are used to analyse EMG signals and conduct PR. Classification accuracy may increase while considering the temporal and spatial information [18].

A longer window length offers more information [19], but at a higher computational cost. For PHA, an adaptive window size improves prosthetic performance [20], reducing classification errors to 13% [21]. Moreover, mitigating the limb position effect in particular [22] can decrease the classification error to 7% on average for PHAs [23]. Actually, training classifiers to consider wrist positions does not improve classifier performance, but the available wrist selection grasping feature is responsible for the improved control [24]. Incorporating Movement Pattern Transmission (MPT) improves pattern recognition. The post-processing by differentiating activities different from the rest position is a better alternative to multi-window-based schemes. With Force variation and absence of relaxation considered, the results of the baseline were 77% and 68%, compared to 84% and 78% with MPT [25]. Furthermore, external load applied on the residual limb negatively impacts the performance of PR. Although improvement of the limb position effect can be attenuated by dynamic training, external load proves to be more challenging to mitigate [26].

Filtering the EMG signals can introduce errors that may affect PR. A solution to the problem is using stochastic generative models to consider uncertainty in the EMG signals while performing PR with 78% accuracy [27].

Modern surface electromyography (sEMG) prosthetics systems can sometimes perform unintended gestures. This is often because of the algorithm which processes the myoelectric signals. Although in a recent study [28] the fault-tolerance was implemented in the PR process, the proposed model achieved only 87.5% accuracy. Thus, more research needs to be conducted to reach a balance between PR and the accidental actuation of the device. The EMG sensor displacement can cause prosthetic control failure. A position validation (PV) algorithm detects shifting electrode positions and prompts a readjustment warning. Consequently, this helps reduce EMG errors and involuntary prosthetic actuation [29].

The entropy-based classification [30] may constitute a path for improved PR algorithms for prosthetic designs. A study with multiple types of data has not been carried out to test the figure of merit against other data than the one used in the cited work.

The training of algorithms to identify patterns is done with data from human or animal patients [31], or online databases. The datasets provided online are usually gathered from high-density sEMG sensors placed on the intrinsic or extrinsic muscles of subjects while they perform certain tasks [32]. Training and retraining the classifier for prosthetic hands with active movements can improve performance. Data gathering is done through able-bodied subjects that perform tasks in a virtual environment [33]. A virtual study showed that combining data from intrinsic and extrinsic muscles can improve classifier accuracy with errors of 7% [34].

Myoelectric control with abstract decoders is a prudent way of commanding prosthetic devices. This strategy uses signals from intrinsic and extrinsic muscle pairs. People with wrist absence are likely to benefit from the proposed scheme. Because of the influence of the wrist, trans-radial and trans-humeral amputees have a better chance to use the devices more than PHA [35]. Abstract myoelectric control is an alternative to pattern recognition with machine learning. It maps different muscles to motors in the prosthetic, and the user can control the device by motor learning [36].

A proposed discrete action control method to guide myoelectric prosthetics was proposed. This can provide simultaneous and independent control of multiple DoFs based on an open-close-stall decision for each DoF. This is suggested to be more effective with bilateral amputees [37].

Hand posture selection can be conducted with a Brain Machine Interface (BMI). The proposed Active Dimension Selection (ADS) decoder can allow the control of a virtual hand avatar with 93% accuracy [38].

A DoF can be controlled if two EMG sensors are placed on the hand. A study has shown that two degrees of freedom can be controlled by having four EMG sensors in place. Increasing the number of electrodes does not improve performance [39]. Similarly, a more recent study suggests that it is possible to control a multi-DoF finger force with one DoF training using a Convolutional Neural Network (CNN) and RNN [40].

The study by Toro et al., 2022 [41] proposed lowering the number of EMG channels from eight or more by using a RNN with LSTM and dense layers. A gesture recognition system that uses only four EMG channels to detect five predefined gestures was proposed. The system accuracy ranges from 99% for training and validation, and decreases to 87 ± 7% for real-time testing. The accuracy and F1 score were calculated to rate the performance of the proposed systems. However, generating good results implies overcoming challenges like classification performance and error mitigation resulting from EMG arm band positioning. The number of EMG channels can be reduced further when considering rehabilitation exoskeletons. As seen in Toro et al., 2024 [42], rehabilitation exoskeletons allow passive and active rehabilitation exercises with controllers using only two sEMG channels. The system is based on model reference adaptive controllers. In

Table 1. Comparative results from the literature.

Study of	Method	Postures	Accuracy
Fonseca et al., 2018 [13]	1 IMU	3	- supervised learning 89% - unsupervised 76% - adaptive 88%
Rizzoglio et al., 2023 [15]	2 IMU	4	- unsupervised
Adewuyi et al., 2016 [16]	offline	7	-
Zuleta et al., 2016 [17]	1 IMU + 8 EMG	5	90%
Earley et al., 2014 [19]	12 EMG	7	-
Al-Timemy et al., 2018 [21]	7 pairs of EMG	-	87%
Adewuyi et al., 2017 [22]	9 EMG	4	-
Beaulieu et al., 2017 [23]	EMG + position	-	90%
Sarabia et al., 2023 [28]	-	-	Accuracy 87.5% Precision 91.87% F1 score 86.34%
Toro et al., 2022 [41]	4 EMG	5	87 ± 7%

, a comparative study of results from the literature is presented. The main aspect presented in the table is the number of electrodes versus the number of detected movements. This is not an exhaustive list as the literature is rich in work that considers a multitude of actuation solutions that becomes increasingly harder to compare due to high the number of parameters required.

3.3. EEG

Devices like the EPOC X wireless Electroencephalography (EEG) headset (Emotiv, San Francisco, CA, USA) empower the Brain Computer Interface (BCI) control of prosthetic limbs. A CNN is a Deep Learning (DL) model. It can decode and classify non-invasive EEG motor imagery (MI) signals from BCI with an accuracy of 68%. These have applications in controlling prosthetic devices and extracorporeal robots using the Mind Machine Interface (MMI) [43]. The decoding of these signals can be improved by including adversarial data augmentation algorithms. An average accuracy of 89% may be achieved on classification tasks [44]. Conversely, the impressive results of the Direct Neural Interface (DNI) come at a high computational cost. Less challenging algorithms, based on LDA, have an average accuracy of 80.5% [45]. A recent study averaged an accuracy of 84% for a Transformed Spatial Pattern (TCSP) algorithm [46]. As an example, in 3D-printed prosthetic applications, the motor imagery EEG signal is used to control three motors [47].

3.4. FMG

Force Myography (FMG) is a method of tracking volumetric change in muscles to detect contraction and relaxation during the movement of the limb. The method has no noise. It is an alternative method of controlling externally powered prosthesis [48]. The performance characteristics shown by FMG can complement or replace EMG applications. Both EMG classifier-based and regression-based schemes were outperformed by FMG in online applications. FMG should be considered for Human Machine Interface (HMI) designs [49].

3.5. MEG

Magnetoencephalography (MEG) is used to estimate sensorimotor cortical currents, with applications in neurofeedback [50]. Patients with severe movement impairment can control prosthetics through MEG-based BCIs. By decoding the slow components of the MEG signals, a neuroprosthetic can be controlled with an accuracy of 76.5% [51].

3.6. ISMS

Early work shows that Intra-Spinal Micro-Stimulation (ISMS) can be used to induce the movement of the arm and hand. This is done through the coordination of muscle contraction by sending an electric signal through an electrode. Inducing a 20 μA current in the spinal cord results in muscle activity. The generated muscle force is influenced by the increase in frequency. Furthermore, this results in the grasp, transportation, and release of objects through ISMS. The integration of cortex and spinal cord control through ISMS allows the use of advanced neuroprosthesis [52]. Similarly, Spinal Cord Stimulation (SCS) systems use electrical pulses in the ventral and dorsal surface of the cord to trigger the movement of the upper limbs. Such a system is surgically implanted, and it surrounds the spinal cord, like a cuff. In the future, such an approach can help in the treatment of patients with tetraplegia [53].

3.7. ECOG

Locked-In Syndrome (LIS) and epilepsy patients were compared with the ability to control a BCI using subdural Electrocorticography (ECOG). The attempted hand movement was detected, making the technology a candidate to use BCI and enable external robot control [54].

3.8. Ultrasonic

An A-mode ultrasound device has been proven to detect muscular deformation to control ULP, with an LDA classifier accuracy of 91%. The practical miniaturized design makes it a complementary or an alternative control strategy to EMG [55].

4. Functional Challenges of 3D-Printed Prosthetics

In this section, some challenges derived from the literature are presented, as illustrated in Figure 3. Prosthetic design should include the goals of the patient for whom the device is made [56]. The finger positioning accuracy may decrease with use. A recent study [57] shows that using a 3D-printed prosthetic for a month already introduces some deviations to the initial calibration. Although the accuracy is not affected significantly in the tested period, an extended 1-year test may show otherwise. If, however, the manufacturing material were to be titanium, the time before maintenance will grow to 3 years [58]. Moreover, if the prosthetic finger were to use similar Direct Metal laser Sintering (DMLS) technologies and be made of steel but also be powered, it may provide additional functionality with a higher complexity [59]. Truthfully, in the latter two cases, it is not about a transitional device but more a permanent solution. Although they are still in the research stage, some EMG-controlled, two-finger configurations are also available [60]. Furthermore, advances in 3D printing development will reduce the performance difference between 3D-printed and commercial prosthetic devices [61]. This is, for example, the case for transradial PHA, where the achieved functional performance of 3D-printed prosthetics is grater then that of standard traditional prosthetics [62].

In trans-metacarpal limb reduction, more advanced designs are available. A generic 3D-printed prosthetic can be customized for the specific level of amputation. The manufacturing process contains several parts made of multiple materials such as Polylactic Acid (PLA) and Thermoplastic Polyurethane (TPU) [63]. Modifying existing designs is easier than devising a new one. A case report shows how a modified talon hand helped a gymnast to improve her participation satisfaction and confidence [64]. This proved the value of having open-source medical projects.

An extensive study suggests avoiding socket encapsulation of the residual limb for distal-to-wrist amputations. A good socket design contributes to prosthetic embodiment [65]. Silicon sockets rely on suction and can accommodate additional prosthetic equipment like EMG electrodes [66]. However, a gel liner system with embedded electrodes and conductive fabric leads offers improved comfort over traditional socket designs [67].

5. Rehabilitation of Partial Hand Reduction

In the literature there is a wealth of case studies regarding the way in which rehabilitation is undertaken for adult and paediatric cases. Some of the studies are presented in the subsequent paragraphs.

Harmful devices at the workplace produce 75% of adult amputations. Children most commonly have door injuries that lead to amputation. With regard to partial hand prosthesis, challenges like the absence of scalable solutions, the high technical complexity of replacing digit function, recovering the opposition movement, relying on traditional rather than currently engineered solutions for ULP, may be considered [68], as seen in Figure 4.

Upper limb reduction presents significantly higher perceived body image disturbance than lower limb amputees because the latter have a greater adjustment to prosthetics. This is explained by the increased levels of depression and anxiety found in upper limb versus lower limb patients [69]. The dominant hand is the most exposed to trauma and the amputation of digits [70].

People with limb absence may choose not to use a functional prosthetic device. In low-income countries a silicone filler is considered enough finger rehabilitation [71]. This is a basic approach and is made using a cast of the contralateral digit [72]. Clearly, the focus is to have an aesthetically acceptable appearance. It seems that a silicone passive vacuum fit finger replacement can improve the aesthetics and psychological well-being of a patient [73]. Although adjusting to amputation can be done using a silicone digit that also improves aesthetics, the functional improvement was not as expected [74]. Silicone prosthetic fingers use vacuum suspension but must reduce discomfort [75]. As a side effect, sweating may occur. Decorative rings may dissolve colour mismatch between the natural and fitted digits [76]. Distal digit reductions can be treated with acrylic devices. The custom prosthetics made of heat-cured acrylic material offer a rehabilitation alternative to silicone [77]. Some variations of this design use acrylic and wrought orthodontic wire [78]. Although it may constitute an alternative, Polyvinylchloride (PVC) prosthetic users are less satisfied with their prosthetic than are those with silicone prosthetics [79].

After PHA, the regain of function is important. Ten-year-old children have an average flexion force of 15 N and an extension of 10 N, while adults have 33 N and 8 N, respectively [80]. A functional alternative to silicone fingers has been proposed. A 3D-printed articulated device has been proven to be more effective than cosmetic fingers at a comparable manufacturing price [81], thus proposing a functional solution for people with finger disarticulation. Although some researchers consider that powered prosthetic devices seem to be better suited to patients with a higher percentage of limb loss than just a digit [82], more recent studies have proven otherwise. Functional improvement has been seen in objective hand functions for patients with 4–5-digit absence [83].

In adults’ home intervention training for people with upper limb absence (ULA), cortical activation increases, which results in an improved performance in handling a prosthetic device [84]. Furthermore, this type of training may invoke an increase in dexterity and achieve independence in children and adolescents [85]. Unfortunately, studies following children with ULA into adulthood are absent [86].

An important objective of rehabilitation is to provide patients with the functionality of doing as many Activities of Daily Living (ADLs) as possible. Better joint and locking mechanism designs may lead to achieving more ADLs with ease [87].

In cases of congenital limb deficiency in children, rehabilitation training should start as soon as possible. A study shows that children with limb absence can benefit from rehabilitation by developing better motor skills if they start therapy at an early age [88]. Initially, a six-month-old can be fitted with a passive prosthetic and, as he develops, more advanced devices should be fitted and training should be conducted through play to reduce abandonment risk [89]. While this is a solution to the problem of rehabilitation, it may pose an economic problem. Luckily, transitional 3D-printed prosthetics have shown good results in children’s hand function improvement at a low cost [90]. Similarly, the early adoption of occupational therapy for persons with accidental trauma can decrease the time necessary for reintegration into the workforce [91].

Influencing the plasticity in the sensorimotor cortex of people with upper limb disarticulation (ULD) may reduce pain. By using a BMI training scheme to dissociate phantom hand and prosthesis, phantom pain reduction can be achieved [92]. The MEG signals were used to estimate the movements of a hand. The visual analogue scale for pain (VAS) survey showed that neurofeedback training demonstrated pain modulation without hand visual feedback [50].

A particular kind of prosthetic is the supernumerary Robotic Finger. This is a complementary prosthetic to enhance remaining or healthy limbs. Movement prediction accuracy using bioartificial synergies is 65% to achieve motion [93]. Control of these robots can be done using redundant DoF of human operator. This results in an aid in task completion at the workplace [94].

Rehabilitation using BCI offers access to advanced prosthetic control. In general, patients like the independence of BCI but an increase in device dependence is bothersome [95]. Interacting with patients and users of BCI has an efficacious impact on User Centred Design (UCD) of prosthetics. The effect of engagement of researchers with persons with disabilities benefits the UCD by helping with end user perspectives [96]. For patients, BCI is exciting and interesting at first and as they moved on, the BCI experience becomes normal. If BCI always does the intended tasks, the users incorporate the feeling of agency over the BCI generated actions. Emotional control in BCI is required to avoid mishandling the devices due to external events [97].

The treatment of the effects of cardiovascular accidents with Functional Electrical Stimulation (FES) has beneficial outcomes. Stroke rehabilitation with the daily use of a neuroprosthesis enhances recovery [98]. Unilateral upper extremity paralysis improves hand performance after FES treatment. The hemiplegic patients that participated in home rehabilitation used a powered assisted FES system for electrical stimulation. Power-assisted FES detects EMG signals of monitored muscles and establishes if the signals are too weak and provides stimulation. It has been proven that patients have seen hand function improvement after five months of therapy [99].

The commercial multi-DoF robots for stroke rehabilitation are not accessible in low-income countries. Patients’ therapy includes repetitive hand movements with devices for regaining mobility. Lowering hand stroke therapy cost can be done by manufacturing affordable rehabilitation robots. A solution may be building a 1 DoF device, with standardized parts and using 3D printing as a means to reduce cost and weight of rehabilitation devices [6]. Furthermore, a design of a 3D-printed, easy to assemble, low-cost, soft robotic exoskeleton can help with lowering therapy costs [5].

A two-part device with a shoulder harness and an electric end effector can be used, similar to EMG control in some cases. Hybrid ULPs are more robust than myoelectric devices and can be considered as an alternative in demanding situations [100].

A 3D-printed orthosis for injured limb immobilization proved to be better in terms of function, comfort, and satisfaction according to this study [101].

6. Surgical Intervention for Optimal Prosthetic Embodiment

In the case of ULP devices, some form of care is required. As seen in the highlighted literature in this section, patients require various levels of treatment to benefit the most out of the situation.

The surgeon and prosthetist combined experience in a multidisciplinary team that can decide the best outcome for limb amputation. Reconstructive surgery in PHA is done to preserve phalanx length or lengthen or shorten the phalanx depending on the considered ULP as seen in Figure 5. This helps rehabilitation with prosthetic devices [102]. In other cases, maybe an implant is the better alternative. Trapeziometacarpal joint Osteoarthritis of the thumb can be treated with an uncemented total joint replacement prosthesis [103], resulting in 95% satisfaction among patients [104].

Through surgery, a direct muscle link can be established between the human and a gripper. Pseudo-cineplasty is a surgery to connect a human muscle to a gripper. In basic terms, the surgeon accesses the palmaris longus tendon and adds a device to measure contractions. With muscle contraction, i.e., shortening and lengthening, the gripper would open or close. Overall accuracy was 80% due to an error in installation. Cineplasty may also permit shape information sensing [105].

More advanced residual peripheral nerve surgery facilitates sense and motor functions to be regained. Microsurgery allows Fascicle-specific targeting of longitudinal interfascicular electrodes (FAST-LIFE) electrodes interfacing. This provides PHAs with the ability to control a 15 DoF robotic hand [106].

Osteointegration surgery is conducted for improved upper limb motor function. This type of attachment for a prosthetic does not require a socket. The two-stage procedure consists of (1) a Titanium implant fixture and, (2) a skin penetration abutment. After the operation, the patient can be fitted with a full prosthetic device. Risks involve overloading and infection for the case of osteointegration. Benefits of the bone-anchored prosthetic are experiencing motion similar to able body people [107]. This type of implant offers a survival rate of 80% and 38% skin infections after five years [108].

The nature of the injured limb sometimes requires special surgery. Regenerative Peripheral Nerve Interfaces (RPNI), which rudimentarily involves implanting a peripheral nerve into a muscle, not only facilitates prosthetic limb control but also has a tendency to reduce pain [109]. This type of surgical procedure has also shown potential in restoring sensory feedback [110].

Targeted muscle reinnervation (TMR) is often used to improve control of prosthetic devices [111]. The pain associated with such surgery and the amputation pain itself can be limited with phantom limb therapy. A current case study described the improved cortical efficiency and reduced training time for myoelectric device use [112].

The unfortunate occurrence of brachial plexus injury leads to the poor motor function of the human hand. Unless biological reconstruction is possible, the functionless hand is amputated and replaced with a hand prosthetic actuated via EMG sensors [113]. Patients undergoing bionic reconstruction perceived function, body-image, and quality of life after bionic reconstruction seem to improve over time [114].

Chronic ischemia that leads to upper extremity requires surgical revascularization. The arterial reconstruction of upper extremity for treatment of chronic ischemia is an infrequently performed surgery, and when not done may lead to amputation [115]. Similarly, the occurrence of Raynaud’s phenomena requires a bypass between the axillary and brachial arteries through a autologous saphenous vein graft as a treatment to ischemia [116].

7. Proprioceptive Feedback Enhancements

The proprioceptive system of recognition can offer a sense of the position of the limbs without having a visual cue. When an amputation occurs, the system is damaged, and the literature offers some suggestions on how the system may be used onwards to adapt.

Touch feedback with prosthetic devices requires sensors. Therefore, solutions of standard embedded pressure capacitive sensors can be used with prosthetic devices, but they need to be calibrated [117]. To reduce the cost of a prosthetic device, a paper-based touch mode capacitive gauge pressure sensor has been proposed. The sensor can be integrated into a skin-like structure to provide touch feedback [118]. Moreover, they have a rectangular and circular diaphragm with trapezoidal cantilevers for acoustic pressure measuring [119]. A solution to calibration errors might be the use of an autocalibration algorithm. A fingertip soft sensor can be used to sense touch. The force sensor can be 3D-printed out of TPU material. A neural network was proposed to correct calibration errors [120]. Although the applied force can be detected thanks to the sensors described, slippage can occur, and the prosthetic may fail to properly grip the object and as a result drop it. A fast solution is a hardware-based automatic slippage prediction and prevention with a tactile sensor [121]. The DESC-finger is a cosmetic low-cost and low-energy digit incorporating such technology. The digit’s touch sensor is a force sensitive resistor (FSR). All components are encapsulated [122]. While touch sensors enable feedback, another method to estimate grasping force is through EMG signal estimation of muscle force [123].

Sound feedback has also been considered to control myoelectric ULP. A 3D-printed myoelectric hand prosthetic with auditory biofeedback for children signals the open prosthetic with a friendly animal sound [124]. A more advanced, non-back driveable solution is the S finger, which is based on a synergetic prehension scheme [125].

Visual feedback prevails over other types of feedback. In its absence, artificial proprioceptive feedback improves the control of myoelectric devices. Artificial feedback may be provided to the residual limb or to another healthy body part [126].

Feedback information overload can decrease control performance. It has been demonstrated that the best initial feedback results come from audio-visual feedback loops. This is superior to providing visual and tactile feedback or visual-audio-tactile [127].

Soft robotic devices with haptic feedback are developed to realize advanced grasping. Advanced grasping refers to holding two or more objects at the same time. Haptic feedback absence impacts the prosthetic with success rates dropping from 90% to 20% when disabling the haptic feedback function [128].

A recent study compares a soft and conventional rigid neuroprosthetic hand. The design offers simultaneous myoelectric control and tactile feedback employing the use of pneumatic actuation, EMG, and Elastomer capacitive sensors [129].

8. Initial Fitting and Training

Following the emergence of mass customization, the new industrial revolution of 3D printing surfaced. Then came the first ever 3D-printed artificial limbs [130]. The development of 3D printing has seen some challenges with low quality prints and advanced CAD skills requirements that have kept home users at a distance. To combat this, the introduction of online platforms that enable users to customize existing designs with little to no CAD knowledge increased the role of open innovation with customers and the strategies used with crowd innovation for the mutual benefit of consumers and companies [11].

Prompt fitting and training with ULP enhances motor coordination for children with ULD [131]. Remote fitting represents an advantage of 3D-printed prosthetic devices. Early introduction of EMG based ULP for children can now be feasible to enable the patient to get used to EMG control. On the other hand, 3DP ULP should not be fitted to children less than 4 years old, due to unmonitored development and their inability to express discomfort [4], as also mentioned in Figure 6. Patients in low-income countries and with limited access to health services can get fitted with a prosthetic by providing some photos of the injured limb and follow the remote instructions [132]. A reduction in coactivation has been seen in children with PHA that used a ULP. Thus, indicating that timely adoption of ULP after limb reduction is beneficial for prosthetic control [133].

When choosing a 3D prosthetic device life-level must be considered. Children may prefer a robotic like device which resemble film props, while adolescents may prefer a more subtle prosthetic, that does not attract that much attention [134]. In case of adults with ULA a positive attitude is encouraged to improve Quality of Life (QoL) with support from family and the amputee community [135].

9. Device Testing

Device testing can be done either through quantitative or qualitative methods as seen in Figure 7. Testing of ULPs can be done through questionnaires. Quebec User Evaluation of Satisfaction with assistive Technology (QUEST) is a methodology to assess the level of user satisfaction with assistive technology devices. Understanding the satisfaction level can lead to improvement of devices to suit an individual’s needs [136]. The Box and Block Test (BBT) is an experiment to measure hand dexterity [137]. Similarly, the Jebsen Taylor Hand Function Test (JTHFT) establishes the level of ADL progress [138].

Proprioceptive testing of prosthetics through a systematic evaluation device is needed to compare performance. A proposed force wall device permits testing of different grasping performance and feedback types of prosthetic devices; this is preferable to questionnaires [139].

Prosthetic simulators have been designed to test new prosthetic devices without the initial implication of people with limb reduction. Prosthesis hand simulators (PHS) may aid in research for the benefit of people with ULA. Simulators may be useful for amputees because they may train with the intact hand while the injured side heals. Thus, improving the overall handling of the ULP device after healing. Although simulators placed on the dorsal or palmar side exist, most are positioned axially. Albeit the ULP devices may have a weight like the intact hand, the simulator adds some more weight to the body of the person undergoing the trials. Thus, results of studies of such simulators are questionable [140].

10. Implementation of FFF 3D Printing Technology

A presence of additive manufacturing technology has been established in the medical device sector. For proof of concept and beyond, a 3D-printed prosthetic design begins with a simple mechatronics implementation and builds functionality based on set goals [141]. The hardware implementation may consist of the beginning of open source hardware like the Arduino and Raspberry Pi systems [142], but FFF 3D-printed devices are still in infancy with regards to adoption due to limited clinical assessment [2]. Most FFF 3D-printed ULPs are lacking in both durability and functionality. Furthermore, they have reduced user experience [3]. A high experience level of design is needed to design ULP devices. Some designs are available online to be customized but they are not medically approved and may pose safety concerns. Although cost is reduced with the manufacturing technology, functionality of 3D ULP is lacking and limited knowledge about these devices makes them hard to recommend [143]. Given these disadvantages 3D printing is useful in paediatric care to educate on surgery, developing intervention strategies, generate procedures and material manufacturing of devices [7]. Three-dimensional-printed prosthetics are said to be cheap, but when computing the costs usually not all parameters are considered. Costs neglected of warranty, design, certification, assembly are usually not considered. Also, 3D printing has not yet fount its place in prosthetic applications. With a low complexity and high need of customization, the partial hand prosthetics development is the most certain to profit from 3D printing [8].

11. Device Additive Manufacturing

Procedurally, the selection of design, customization, printing, assembly, fitting, and follow up compose the 3D printing process of ULP [12]. Moreover, printing parameters like cooling time and part orientation strongly influence the mechanical properties of the manufactured devices [9,144]. During the design stage different compartments must be present to accommodate the electronics [145]. For the best results, the parts need to be designed with the 3D printing manufacturing process in mind. The tuning of key process parameters in desktop 3D printers for optimal outcomes in part fabrication is key [146]. As the process evolves so do the materials and production towards industry 4.0 and beyond.

A practical 3D-printed ULP design must consider appropriate manufacturing materials. Materials in ULP need to be biocompatible [147] and biodegradable that enable biomimicry [148]. An antibacterial material used for 3D-printed transitional prosthetics, prevents skin infection [149]. In addition to FFF, which is mainly used for low-cost prosthetics there are other additive manufacturing technologies used in medical applications such as Stereolithography (SLA), Direct Ink Writing (DIW), Laser Guided Direct Ink Writing (LGDIW), and thermal inkjet bioprinting [150]. More advanced material manufacturing processes acquire and filter data with neural network algorithms to improve control of the production line [151]. The more advanced technologies are beyond desktop 3D printing and their applications involve permanent solutions like Bone drug delivery scaffold for Bone fractures to form pseudo bone and induce healing [152].

Forequarter amputations require a highly customized prosthetic. The maker must use an Iterative UCD approach to fulfil the set goals [153].

A low-cost wearable sEMG sensor design enables development of a 3D-printed myoelectric prosthetic. The single channel sEMG sensor with included power supply showed comparable performance to commercial solutions [154]. Trans-radial limb reduction patients can benefit from this implementation [155].

Some prototypes of 3D-printed ULP for trans-metacarpal amputees use angular position sensors. The actuation is done with DC motors. The total cost of parts was given at USD 785 [156].

In modular multi-sensory ULP, the design includes a self-locking mechanism with a worm and wheel set [157].

In the design and development of the Tact hand, a 3D-printed myoelectric ULP device, the Distal Interphalangeal Joint (DIP) is fused with relative movement between the distal and middle phalanx. The ULP provides adaptive grip capability to people with trans-radial amputation [158].

Adaptive grip prosthetic hands usually use a selectively lockable differential mechanism [159].

Multi-material printing allows embedded parts for ease of assembly. A Bio-inspired ULP device built with multi-material printing weighs 92 g costs USD 12. The sliding joints make this design more advanced than other 3D-printed ULP [160].

The x-limb is a soft robotic hand prosthetic printed with TPU90. It has 1 EMG sensor and 1 button to switch between the 3 grasp types [161]. Clinical evaluation shows comparable performance to existing devices [162].

Softhand Pro D is an advanced, dynamic synergy ULP control based on posture synergies. The hand has a motor, and differential and damping systems, that can change the posture of the hand based on the synergies extracted from frequency signals analysis [163]. Patients with bilateral amputation seam to perform better and have a higher acceptance for the Adaptable Poly-Articulated SoftHand Pro [164].

12. Paediatric Prosthetics

The Cyborg Beast is an Open-source paediatric prosthetic. It has been demonstrated to have a good influence on the quality of life in upper-limb reduction cases [165]. As a low-cost transitional device, it improves range of motion (ROM) [166]. Another example is the Unlimbited Arm that can be adapted to be used by children with below the elbow amputation [167].

13. Future Development

A wearable source of energy will benefit myoelectric prosthetic users. Powering wearable devices with the biomechanical energy harvesting backpack can boost prosthetic development. The harvesting system has two components: the energy generating component, the human, and the energy harvesting component. These systems are knee joint-based, limb swinging-based, and the movement of waist as an inverted pendulum model. They produce 1W of output power [168].

The 4D printing process is like 3D printing with smart materials. The 4D materials change shape or properties over time in response to external stimuli. As the additive manufacturing field develops, its applications also benefit further. The invention of smart materials that react to external stimuli has a promising future, leading to applications of nontoxic and non-immunogenic to be biocompatible in implants that can adapt as the patient heals [169].

An alternative to sensor-controlled ULP is a sensorless control strategy based on state observers in under-tendon-driven mechanisms [170]. Conversely, Linear Quadratic Gaussian (LQG) integral control models seem to have good results in coordination of function in Central Nervous System (CNS) with prosthetic digit rehabilitation [171]. Recent research has focused on rejection systems for disturbance and uncertainty in prosthetic control. This can be done by using the mixed mu-synthesis controller to control a virtual 21 DoF prosthetic device [172].

Further evolution of collaborative work based on open-source and file sharing online platforms with version control can improve open-source prosthetic development [11].

14. Results

There is a multitude of model databases for printable 3D models due to the hobby level of 3D printing industry. Although there may be several hundreds of online 3D printable model databases, the authors considered the most relevant and up-to-date ones and presented them in

Table 2. Online platforms queried for partial hand prosthetics projects.

No.	Database	Web Address	Access Date
DB1	thingiverse	https://www.thingiverse.com/	30 June 2023
DB 2	myminifactory	https://www.myminifactory.com/	30 June 2023
DB 3	pinshape	https://pinshape.com/	30 June 2023
DB 4	printables	https://www.printables.com/	30 June 2023
DB 5	NIH 3D print exchange	https://3d.nih.gov/	30 June 2023
DB 6	youmagine	https://www.youmagine.com/	30 June 2023
DB 7	grabcad	https://grabcad.com/	30 June 2023
DB 8	instructables	https://www.instructables.com/	30 July 2023
DB 9	Sketchfab	https://sketchfab.com/	30 July 2023
DB 10	Cults	https://cults3d.com/	30 July 2023
DB 11	CGtrader	https://www.cgtrader.com/	30 July 2023
DB 12	TurboSquid	https://www.turbosquid.com/	30 July 2023
DB 13	3D export	https://3dexport.com/	30 July 2023
DB 14	Free3D	https://free3d.com/	30 July 2023
DB 15	Redpah	https://www.redpah.com/	30 July 2023
DB 16	Zortrax	https://zortrax.com/	30 July 2023
DB 17	Creality cloud	https://www.crealitycloud.com/	20 September 2023
DB 18	Thangs	https://thangs.com/	20 September 2023
DB 19	sketchup 3D warehouse	https://3dwarehouse.sketchup.com/	20 September 2023
DB 20	Creazilla	https://creazilla.com/	20 September 2023
DB 21	Threeding	https://www.threeding.com/	20 September 2023

. Several online searches have allowed the compilation of the list in

Table 3. Selected prosthetic projects.

No	Prosthetic Device Name	Link
01	Knick’s Prosthetic Finger v3.5.5	https://www.thingiverse.com/thing:1340624, accessed on 30 July 2023
02	Flexy Hand 2	https://www.thingiverse.com/thing:380665, accessed on 30 July 2023
03	Raptor Reloaded by e-NABLE	https://www.thingiverse.com/thing:596966, accessed on 30 July 2023
04	AH- Partial Finger Prosthetic	https://www.thingiverse.com/thing:471755, accessed on 30 July 2023
05	PROSTHETIC THUMB H.A.	https://www.thingiverse.com/thing:2565563, accessed on 30 July 2023
06	e-NABLE Phoenix Hand v2	https://www.thingiverse.com/thing:1453190, accessed on 30 July 2023
07	Cathy’s Lucky Fin V3	https://www.thingiverse.com/thing:4902137, accessed on 30 July 2023
08	Flexibone Prosthetic Hand 2019	https://www.thingiverse.com/thing:3962905, accessed on 30 July 2023
09	Cyborg Beast	https://www.thingiverse.com/thing:261462, accessed on 30 July 2023
10	Roth Hand (Progressive and Independent Finger Movement)	https://www.thingiverse.com/thing:220942, accessed on 30 July 2023
11	Robohand	https://www.thingiverse.com/thing:305160, accessed on 30 July 2023
12	Ody Hand 2.1	https://www.thingiverse.com/thing:262930, accessed on 30 July 2023
13	Flexy-Finger Prosthesis	https://www.thingiverse.com/thing:693429, accessed on 30 July 2023
14	The Osprey Hand by Alderhand and e-Nable	https://www.thingiverse.com/thing:910465, accessed on 30 July 2023
15	Talon Hand 3.0	https://www.thingiverse.com/thing:229620, accessed on 30 July 2023
16	Kinetic Finger	https://www.thingiverse.com/thing:1737001, accessed on 30 July 2023
17	BioMech Finger Prosthesis (Index Left Hand)	https://www.thingiverse.com/thing:6202389, accessed on 30 July 2023
18	Second Degree Hand	https://www.thingiverse.com/thing:300499, accessed on 30 July 2023
19	Flex Lok Finger	https://www.thingiverse.com/thing:5499514, accessed on 30 July 2023
20	Dynamic finger prosthesis	https://www.thingiverse.com/thing:498335, accessed on 30 July 2023
21	Hollies hand V5	https://www.thingiverse.com/thing:794079, accessed on 30 July 2023
22	e-Nable Raptor Hand Lock	https://www.thingiverse.com/thing:1750858, accessed on 30 July 2023
23	Aline’s Index	https://www.thingiverse.com/thing:4757371, accessed on 30 July 2023
24	Flexy Index	https://www.thingiverse.com/thing:4845022, accessed on 30 July 2023
25	Rainbow Phoenix	https://www.thingiverse.com/thing:4632992, accessed on 30 July 2023
26	Falcon Hand V2	https://www.thingiverse.com/thing:603039, accessed on 30 July 2023
27	BIOT hand prosthesis	https://www.thingiverse.com/thing:1388216, accessed on 30 July 2023
28	Solo Finger Pen	https://www.thingiverse.com/thing:2122752, accessed on 30 July 2023
29	Kinetic Hand	https://www.thingiverse.com/thing:4618922, accessed on 30 July 2023
30	“Spock” Basketball Prosthetic Hand	https://pinshape.com/items/24569-3d-printed-spock-basketball-prosthetic-hand, accessed on 30 July 2023
31	V.2 Flex Fingers, Swivel Thumb Prosthetic Hand	https://pinshape.com/items/16478-3d-printed-v2-flex-fingers-swivel-thumb-prosthetic-hand, accessed on 30 July 2023
32	BioMech Finger Prosthesis (Index Right Hand)	https://www.printables.com/model/572873-biomech-finger-prosthesis-index-right-hand, accessed on 30 July 2023
33	Ody Hand	https://3d.nih.gov/entries/3DPX-001010, accessed on 30 July 2023
34	Raptor Hand	https://3d.nih.gov/entries/3DPX-000996, accessed on 30 July 2023
35	K1 Hand	https://3d.nih.gov/entries/3DPX-020271, accessed on 30 July 2023
36	Ody Hand	https://3d.nih.gov/entries/3DPX-020274, accessed on 30 July 2023
37	Talon Hand	https://3d.nih.gov/entries/3DPX-020273, accessed on 30 July 2023
38	Kinetic Hand	https://3d.nih.gov/entries/3DPX-020261, accessed on 30 July 2023
39	Osprey Hand	https://3d.nih.gov/entries/3DPX-020262, accessed on 30 July 2023
40	e-NABLE Phoenix Hand v3	https://3d.nih.gov/entries/3DPX-020260, accessed on 30 July 2023
41	Cyborg Beast	https://3d.nih.gov/entries/3DPX-020267, accessed on 30 July 2023
42	e-Nable—YuLia Custom hand	https://www.youmagine.com/designs/e-nable-yulia-custom-hand, accessed on 30 July 2023
43	MULTIHAND 3.0	https://www.youmagine.com/designs/multihand-3-0, accessed on 30 July 2023
44	MCP Driver (Naked Prosthetics)	https://grabcad.com/library/mcp-driver-naked-prosthetics-1, accessed on 30 July 2023
45	Hero Hand UPDATE—Bionic prosthetic hand	https://grabcad.com/library/hero-hand-update-bionic-prosthetic-hand-1, accessed on 30 July 2023
46	T-hook: Prosthetic Design for 3D Printing	https://www.instructables.com/T-hook-prosthetic-design-for-3D-printing/, accessed on 30 July 2023
47	Empower: The Aquatic 3-D Printed Prosthetic	https://www.instructables.com/EMPOWER-the-Aquatic-Prosthetic/, accessed on 30 July 2023
48	The Paraglider	https://hub.e-nable.org/s/e-nable-devices/wiki/The+Paraglider, accessed on 30 July 2023
49	Flex Finger	https://hub.e-nable.org/s/e-nable-devices/wiki/Flex+Finger, accessed on 30 July 2023

. Data extraction from the internet databases was done between June and September 2023.

The initial search of these databases brought tens of thousands of results. In this situation, common words like hand and finger were removed and the following keywords were considered (i) prosthetic, (ii) prosthesis, (iii) prostheses, (iv) partial hand, (v) partial finger, and (vi) partial limb.

After the initial database query, a number of 50 thousand links were generated. By removing the common keywords (a) hand, (b) finger, and (c) limb, 5000 links remained. The distribution of these results can be seen in Figure 8.

The remaining links were hand-sorted to filter prosthetics from other models. A total of 865 prosthetics were included in the list. The distribution of the results can be seen in Figure 9 with databases on the X–axis and percentage of total on the Y–axis. As seen in Figure 9, the higher number of models comes from the first database while other, smaller databases have no relevant results. The results from the eNable community have not been included in Figure 9, but the results are preset in the body of Table 3.

The results were subjected to a set of criteria that included: part availability, availability of pictures of patients wearing the devices, a unique name (not just “prosthetic hand”), if the device had multiple versions, the most recent was included and if the device had remixes, the original project was considered only. In Table 3, the 49 projects that comprise the final list of prosthetic devices for partial hand amputees are presented.

On consideration of the process of selecting some prosthetic devices out of the thousands of objects found during the search, a table of criteria to compare against the candidates was devised, as seen in

Table 4. Main features of prosthetic for partial hand reduction.

Category	Criteria	Description
Presentation	Name	A proper name, not just a generic one like prosthetic hand
	Graphics	Provides detailed graphics
	Description	Detail description
Availability	Licensing	Proof of open-source licensing
	Print files	User must be able to download the files
	Design files	Possibility to edit the design files
Documentation	Manufacturing guides	Present at least in part the print settings
	Assembly guides	Present at least in part the assembly process
	Tutorials	Present a guided text and image instruction or provide video tutorials
	Community forum	Have a community bulletin board or a means of communicating feedback and exchange messages with the interested groups.
Technical	A sound design	Must adhere to engineering common sense in terms of assembly techniques
	Version control history	Must have prior stable versions
	Practical application	Must be suitable to wear and provide realistic capability
Research	Literature	Be present in research studies
	Case studies	Must be present in case studies
	Research stage	Must have surpassed the stage of technical testing and moved on to human testing
	Validation through Online presence	Appear in results of reputable databases

. The authors considered the criteria presented in the early work of Ten Kate [3] and the most recent work of Wendo [12], but also reflected on their own experience.

15. Discussion

In the search for 3D-printed low-cost upper-limb prosthetics, many devices were found to have similar properties, shape, and function. A set of devices is proposed for different levels of amputation. These are examples of usable low-cost 3D-printed transitional upper-limb prosthetics.

15.1. Finger Prosthetics

In the case of patients with partial or full finger amputation, there are silicone passive devices described in the literature. A viable solution for mechanical body-powered finger prosthetics is Knicks Finger 3.5, seen in Figure 10, which can be found in online databases, in the Enable catalogue, and mentioned in the literature with good results.

This prosthetic device is 3D-printable and has good documentation coverage. The design has seen multiple versions over the years and the history can be checked on the database providing the files. The available files are provided on terms of the license CC BY-NC-SA. The project is truly open source as it provides access to both STL and SCAD design files. The documentation provides detailed guidance on manufacturing and assembly with text, image, and video tutorials. The device is present in the literature [81] with good results. This solution is 3D-printable so in comparison to a silicone variant it is much practical. If the first iteration does not fit, the adjusted version can be easily printed.

Although there are some notable efforts in the field, to our knowledge there are no practical transitional externally powered finger prosthetics.

15.2. Partial Hand Prosthetic with All Fingers Amputated

Patients who had undergone partial hand amputation but with an intact palm and functioning wrist can use a body-powered prosthetic. This kind of prosthetic permits a low number of grasps but nevertheless allows getting used to the device.

A device like the Phoenix Hand V3, as seen in Figure 11, is a suitable candidate for a transitional prosthetic for such patients. Even though the design files are not provided, the print files are licensed under the CC BY license and can be used freely. Like the Knicks finger, this device has good documentation and is supported by a community. The design has evolved throughout the versions available and is now present as a generic base line for the prosthetic community.

The device is preferable due to its maturity as a design. The nature of the design also allows it to be scaled to fit paediatric cases.

15.3. Partial Hand Prosthetic with Weak Wrist

A useful device for such patients is the Flexy Hand 2, seen in Figure 12. A weak wrist means a low mechanical actuation capability.

A device like the Phoenix Hand V3 is not recommendable because it requires a considerable amount of force to be actuated. Although they are not as good as the other two solutions proposed, the documentation guidelines are available and can provide instruction for manufacturing and assembly of the available files. They are licensed under the CC BY-NC-SA license for the STL files provided. Although it is present in the literature [12], it has a smaller footprint compared to the others.

15.4. Paediatric Partial Hand Prosthetic

A device present in literature and on the online databases is the Cyborg Beast, as depicted in Figure 13.

This device can be used with children and according to several previously cited studies, and the device can be remotely prescribed and fitted to children of all ages. The production STL files are provided under the CC BY-NC license. The design is undergoing thorough research conducted by Zuniga et al. [165], and represents a considerable item to compare other prosthetics in the paediatric cases.

Roads ahead should include:

-

Developing other means of performance comparisons between prosthetics that may lead to a unified testing procedure for medical devices. Although some criteria have been proposed by this work and others, there is still a need for further development.

-

Multidisciplinary teams in clinical studies may unite to give a prescription prosthetic based on both open source and commercial solutions while thinking about both engineering and medical effects on the patient’s wellbeing, including both surgeons and prosthetists, as well as biomedical engineers in the discussion may lead to better outcomes.

-

A procedure to filter or certify platforms to provide medical equipment should be considered. This may provide patients with files and instructions but under the guidance and supervision of a specialist.

-

Technology development in the field of additive manufacturing may open new capabilities for healing of patients with trauma and helping them adapt to their new situation. The 4D printing process is promising to revolutionize the field of manufacturing once again. The medical applications are still under development but have promising results.

-

Proprioceptive feedback techniques development may lead to a better handling of prosthetic devices in the future. Better control will also mean more people will be able to use a prosthetic. Nowadays, there are problems with device abandonment due to poor control strategies that cause frustration.

Energy-efficient prosthetic devices that are powered through the motion of the other body parts relinquishing the need for charging. Although the more advanced prosthetics promise a dexterous human-like hand device, the functionality ends with battery depletion. Further research in the field of body generated electricity may help sustain all day autonomy for prosthetics.

The small number of results in clinical testing as well as the lack of literature in the field of testing such devices makes them harder to recommend. The ability to differentiate between these types of devices is currently based on personal experience and not on experimental results. Such limitations along with others mentioned by similar work, pave the way for future research and development.

16. Conclusions

The development of commercial prosthetics and additive manufacturing has given rise to the transitional low-cost partial hand prosthetics. As a result of sensor development, we may start to see innovative solutions for sensing technology. Consequently, more research projects will be developed into practical applications of feedback enabled transitional prosthetics. One of the challenges of finding low-cost 3D printable transitional partial hand prosthetic devices is the multitude of databases available. Moreover, the high number of results and the duplicates found can insert a higher level of difficulty on the choice. Consequently, specialists may rely on research like the present work to narrow the search for their patients.