Enhancing Phantom Limb Pain Rehabilitation Technology with User-Centered Improvements Improving a CE-Marked Medical Device in Collaboration with Integrum AB Master’s thesis in Biomedical Engineering OLIVER MOBERG DEPARTMENT OF ELECTRICAL ENGINEERING CHALMERS UNIVERSITY OF TECHNOLOGY Gothenburg, Sweden 2024 www.chalmers.se www.chalmers.se Master’s thesis 2024 Enhancing Phantom Limb Pain Rehabilitation Technology with User-Centered Improvements Improving a CE-Marked Medical Device in Collaboration with Integrum AB OLIVER MOBERG Department of Electrical Engineering Division of Signal Processing and Biomedical Engineering Chalmers University of Technology Gothenburg, Sweden 2024 Enhancing Phantom Limb Pain Rehabilitation Technology with User-Centered Im- provements Improving a CE-Marked Medical Device in Collaboration with Integrum AB OLIVER MOBERG © OLIVER MOBERG, 2024. Supervisor: Morten B. Kristoffersen Examiner: Petter Falkman Master’s Thesis 2024 Department of Electrical Engineering Division of Signal Processing and Biomedical engineering Chalmers University of Technology SE-412 96 Gothenburg Telephone +46 31 772 1000 Cover: Neuromotus device used with the textrode band, playing the improved Breakout game in joystick mode. Typeset in LATEX, template by Kyriaki Antoniadou-Plytaria Printed by Chalmers Reproservice Gothenburg, Sweden 2024 iv Enhancing Phantom Limb Pain Rehabilitation Technology with User-Centered Im- provements Improving a CE-Marked Medical Device in Collaboration with Integrum AB OLIVER MOBERG Department of Electrical Engineering Chalmers University of Technology Abstract Phantom limb pain (PLP) is a common and distressing sensation experienced by 50−85% of individuals with limb loss, often manifesting as pain in the removed body part without any physical cause. Rehabilitation methods such as mirror therapy, motor imagery, and phantom motor execution (PME) are used to alleviate this pain. PME, in particular, engages the central and peripheral nervous systems to activate muscles in the residual limb. Integrum AB has developed Neuromotus, a medical device for PLP rehabilitation that leverages PME principles through virtual reality (VR), augmented reality (AR), and serious gaming. This project aims to identify and implement user-centered improvements to Neuromotus. The project began with a qualitative pre-study involving semi-structured interviews with experienced therapists who have used Neuromotus in clinical practice. Based on their feedback, several improvements were identified and implemented. Including two new open-source games, enhanced functionalities in existing games, the addi- tion of a proportional controller, and the ability to customize the AR limb skin tone. Stakeholders subsequently tested these enhancements and validation tests were made for all then new functionalities. The results indicate that the proportional controller significantly enhanced game- play, while the customizable limb color in the AR environment promoted inclu- siveness and ensured alignment with ethical standards. These improvements could collectively contribute to a more effective and user-friendly PLP rehabilitation ex- perience. Keywords: Phantom Limb Pain, Phantom Motor Execution, Neuromotus, Aug- mented Reality, Serious Gaming. v Acknowledgements I would like to express my sincere gratitude to Integrum AB for granting me the opportunity to be part of this enriching experience. I am deeply thankful to my examiner, professor Petter Falkman, and my supervisor, Morten B. Kristoffersen, for their invaluable guidance and continuous support. I am also grateful to the physiotherapists who agreed to be interviewed and provided feedback and informa- tion that guided this project forward. Additionally, I extend my appreciation to the representatives from the HOPE study for their valuable insights and contributions to this project. I am appreciative of all my peers and colleagues at Integrum for their support, col- laboration, and shared experiences throughout my academic journey. In particular, I am grateful to, Magnus Lindholm, Elin Sjöqvist, and Björn Davidsson, whose exper- tise, support, and camaraderie have been invaluable throughout my thesis project. Oliver Moberg, Gothenburg, 07 2024 vii List of Acronyms Below is the list of acronyms that have been used throughout this thesis listed in alphabetical order: AR Augmented Reality HID Human Interface Device MDCG Medical Device Coordination Group MDD Medical Device Directive MDR Medical Device Regulation MPR Myoelectric Pattern Recognition PLP Phantom Limb Pain PME Phantom Motor Execution RMS Root Mean Square sEMG surface Electromyography TAC Target Achievement Control VR Virtual Reality ix Contents List of Acronyms ix List of Figures xiii List of Tables xv 1 Introduction 1 1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1.2 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1.2.1 Cortical mapping . . . . . . . . . . . . . . . . . . . . . . . . . 3 1.2.2 Current therapy methods . . . . . . . . . . . . . . . . . . . . . 3 1.3 Neuromotus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 1.3.1 Hardware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 1.3.2 Software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 1.4 Previous related Neuromotus studies . . . . . . . . . . . . . . . . . . 10 1.4.1 The first case study . . . . . . . . . . . . . . . . . . . . . . . . 10 1.4.2 The second study . . . . . . . . . . . . . . . . . . . . . . . . . 10 1.4.3 The third study . . . . . . . . . . . . . . . . . . . . . . . . . . 11 1.5 Medical device regulations . . . . . . . . . . . . . . . . . . . . . . . . 12 1.5.1 Regulatory implications of changing a MDD classed medical device . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 1.6 Purpose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 1.7 Delimitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 2 Pre-Study 15 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 2.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 2.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 2.4 Final improvements of Neuromotus . . . . . . . . . . . . . . . . . . . 19 3 Methodology 21 3.1 Improvements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 3.1.1 Improvement 1.(i) Finding suitable open-source games . . . . 21 3.1.1.1 Game evaluation and elimination . . . . . . . . . . . 23 3.1.2 Improvement 1.(ii) Improving the games . . . . . . . . . . . . 25 3.1.2.1 Racing game . . . . . . . . . . . . . . . . . . . . . . 25 3.1.2.2 Breakout . . . . . . . . . . . . . . . . . . . . . . . . 25 xi Contents 3.1.2.3 2048 . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 3.1.2.4 Snake . . . . . . . . . . . . . . . . . . . . . . . . . . 26 3.1.3 Improvement 1.(iii) Integrate a proportional controller . . . . 27 3.1.4 Improvement 2.(i) Customize AR limb color . . . . . . . . . . 30 3.2 Tests of the new improvements . . . . . . . . . . . . . . . . . . . . . 31 3.2.1 Stakeholder feedback . . . . . . . . . . . . . . . . . . . . . . . 31 3.2.2 Validation tests . . . . . . . . . . . . . . . . . . . . . . . . . . 31 4 Results 33 4.1 Feedback from stakeholders . . . . . . . . . . . . . . . . . . . . . . . 33 4.1.1 HOPE representatives . . . . . . . . . . . . . . . . . . . . . . 33 4.1.2 Integrum employees . . . . . . . . . . . . . . . . . . . . . . . . 34 4.2 Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 4.2.1 General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 4.2.2 AR environment limb color customization . . . . . . . . . . . 36 4.2.3 Games . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 4.3 Regulatory aspects . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 5 Discussion 41 6 Conclusion 45 Bibliography 47 A Interview Questionnaire I B Interviews III B.1 Interview 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . III B.2 Interview 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XI C Charts guiding the assessment whether changes are significant XIX C.1 Chart A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XIX C.2 Chart B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XX C.3 Chart C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XXI C.4 Chart D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XXII C.5 Chart E . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XXIII D Question form for stakeholders XXV xii List of Figures 1.1 The cortical mapping of the somatosensory cortex, highlighting how different regions of the cortex are linked to specific body parts. The image was adapted from Nursing Hero and licensed under CC BY 3.0 US. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 1.2 Neuromotus user kit. . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 1.3 Neuromotus user interface. . . . . . . . . . . . . . . . . . . . . . . . . 7 1.4 Neuromotus settings for a user with upper limb loss. . . . . . . . . . 8 1.5 Current games included in the Neuromotus package. . . . . . . . . . 9 1.6 Neuromotus Questionnaire interface. . . . . . . . . . . . . . . . . . . 9 1.7 Significant changes regarding the transitional provision under Article 120 of the MDR (Source: MDCG 2020-3 Rev.1) [31]. . . . . . . . . . 14 3.1 Comparison of the old and new version of the racing game. . . . . . . 25 3.2 Comparison of the old and new version of Breakout. . . . . . . . . . . 26 3.3 Comparison of the old and new version of 2048. . . . . . . . . . . . . 27 3.4 Comparison of the old and new version of Snake. . . . . . . . . . . . 27 3.5 Illustration of simultaneous movement limitations. . . . . . . . . . . . 29 3.6 Monk skin tone scale. . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 3.7 Comparison between the 1st and 10th skin tone on the Monk skin tone scale. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 4.1 Feedback form, question two. . . . . . . . . . . . . . . . . . . . . . . 34 4.2 Feedback form, question three. . . . . . . . . . . . . . . . . . . . . . . 35 4.3 Feedback form, question five. . . . . . . . . . . . . . . . . . . . . . . . 35 4.4 Decision route for significant changes regarding the transitional provi- sion under Article 120 of the MDR (Modified version from the source: MDCG 2020-3 Rev.1) [31]. . . . . . . . . . . . . . . . . . . . . . . . . 39 C.1 Guiding changes related to the intended purpose. Source: MDCG 2020-3 Rev.1 [31]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XIX C.2 Guiding changes related to the design where the following principles should apply. Source: MDCG 2020-3 Rev.1 [31]. . . . . . . . . . . . . XX C.3 Guiding changes related to software. Source: MDCG 2020-3 Rev.1 [31].XXI C.4 Guiding changes related to a substance or material. Source: MDCG 2020-3 Rev.1 [31]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XXII C.5 Guiding changes related to sterilisation. Source: MDCG 2020-3 Rev.1 [31]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XXIII xiii List of Figures D.1 First two questions of the form. . . . . . . . . . . . . . . . . . . . . . XXV D.2 Last four questions of the form. . . . . . . . . . . . . . . . . . . . . . XXVI xiv List of Tables 3.1 Game Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 3.2 Game evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 3.3 Game elimination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 4.1 Tests of improvements for general use. . . . . . . . . . . . . . . . . . 36 4.2 Tests of improvements in the AR environment . . . . . . . . . . . . . 36 4.3 Tests of improvements of the games. . . . . . . . . . . . . . . . . . . 37 xv List of Tables xvi 1 Introduction The human body perceives information about the surroundings by using receptors that sense vibration, temperature, touch, and noxious stimuli [1]. A part of the neu- rological system that processes this information is called the somatosensory system, which facilitates feedback from the surroundings through the stimulated receptors. It has several purposes including working as a warning system, alerting our body of harmful situations perceived as pain. Amputation can inflict a sensation of pain without any physical strain, called neu- ropathic pain [2]. The origin of this pain is yet to be conclusively determined. A different type of pain that can emerge after amputation is nociceptive pain which is associated with physical pain receptors producing afferent discharges, sometimes resulting in neuroma pain. A pain inflicted from the growth of nerve tissue that occurs due to nerve injury, often presenting as a painful, disorganized mass of nerve fibers [3]. Both of these two mechanisms inflict pain with varying sensations and are referred to as phantom limb pain (PLP). They are different and it is therefore important to make a distinction between these, Professor Ortiz Catalan defined it as [2]: "Neuropathic Phantom Limb Pain is pain perceived as arising from the missing limb due to sources other than stimulation of nociceptive fibers that used to innervate the missing limb." "Nociceptive Phantom Limb Pain is pain perceived as arising from the missing limb deterministically by stimulation of nociceptive fibers." This distinction is important for the treatment being adapted to fit the appropriate pain profile such that the clinicians can offer personalized therapy to reduce PLP [4]. Bacterial or viral infections could stimulate the nociceptive receivers and cause nociceptive PLP[[5], [6]]. Treating this as neuropathic PLP with cognitive therapy would be unsuccessful, finding the correct cause is therefore of significant impor- tance. The phenomenon of PLP has been recognized for centuries. The term "phantom limb" was first coined by American neurologist Silas Weir Mitchell in the late 19th century [7]. In his observations of Civil War persons with limb loss, Mitchell de- scribed expressive sensations of lost limbs, including severe pain. Prior to Mitchell’s work, French military surgeon Ambroise Paré, in the 16th century, documented cases 1 1. Introduction of persons with limb loss who felt pain in their absent limbs [8]. These historical accounts laid the foundation for understanding PLP, which has since evolved signif- icantly with advancements in medical science. The relevance of studying PLP and its underlying causes is emphasized by the in- creasing number of amputations due to diabetes, vascular diseases, trauma, and cancer [9]. The perception of a phantom limb can result in sensations such as numb- ness, tickling, and cramps. When these sensations become painful, they are defined as PLP [10]. This painful sensation appears in 50-85 % of all persons with limb loss, making it the most common condition after amputation. It affects daily life during work, sleep, and exercise which reduces enjoyment and life quality [11]. In 2017, 57,7 million people worldwide were living with limb loss due to traumatic causes [12]. With advancements in surgical techniques and prosthetic technology, amputations are becoming more common, making PLP a growing concern in both civilian and military populations [13]. Addressing PLP effectively requires a multi- faceted approach, incorporating medical, psychological, and rehabilitative strategies to improve patient outcomes [14]. 1.1 Background The company Integrum AB will in collaboration with University of Borås partici- pate in a new study, HOPE, founded by the KK-stiftelsen (Stiftelsen för kunskaps- och kompetensutveckling). The study will focus on how textile-based electrodes can be combined with Neuromotus, a medical device for treating PLP, to improve adherence to home-based rehabilitation. The clinicians and therapists included in the study will use Neuromotus for the PLP rehabilitation treatment. Leading up to this project, two research projects have been conducted. The first study was conducted from 2016 to 2019 to introduce the textrode-band for Surface Electromyography (sEMG) recording using advanced flat knitting tech- nology [15]. This band integrates electrodes and connectors into the fabric, provid- ing a flexible, wearable, and washable alternative to conventional single-use elec- trodes. The study evaluated the band’s feasibility in a myoelectric pattern recogni- tion (MPR) task involving decoding movements of the lower limb. Results showed promising signal-to-noise ratio, offline MPR accuracy, and real-time performance metrics, particularly for proximal limb movements like knee flexion/extension. The second study was conducted from 2020 to 2023 and focused on advancing smart textile technology to improve healthcare, particularly in neuromuscular rehabilita- tion [16]. By embedding sensors in textiles, patients can monitor their health easily by wearing sensor-equipped clothing and using a smartphone app. The project aims to develop NeuRehab@home, a smart textile platform for neuromuscular rehab at home. It utilizes sEMG for biofeedback-based therapy, including treating PLP us- ing sEMG signals to control VR hands. Traditional sEMG methods with disposable electrodes have limitations, so the project explores smart textile solutions with im- 2 1. Introduction proved comfort and usability. 1.2 Theory This chapter presents the theory behind PLP and the current therapy methods used for rehabilitation. It also provides a detailed description of Neuromotus, including its key features and previous studies related to the device. Furthermore, an overview of the regulatory framework for medical devices and the process of changing these devices is included. 1.2.1 Cortical mapping Understanding the neural mechanisms underlying PLP is crucial for developing effec- tive therapeutic interventions. One key aspect of this understanding lies in cortical mapping and cortical reorganization [17]. Cortical mapping refers to the organization of the somatosensory cortex, the area of the brain responsible for processing sensory information, including touch and propri- oception [18]. Figure 1.1 illustrates how different parts of the somatosensory cortex are linked to specific body parts. In individuals with intact limbs, cortical mapping is highly organized, with specific regions dedicated to different body parts. However, following amputation, the cortical mapping can experience significant changes [17]. This phenomenon is called cortical reorganization, occurring due to the brain’s abil- ity to reorganize its structure and function in response to sensory input or damage. In the context of PLP, cortical reorganization involves remapping the somatosensory cortex as the brain attempts to repurpose areas previously used by the now-missing limb. This can become maladaptive if it begins to contribute to the development and persistence of PLP. The therapeutic methods of managing PLP are closely linked to cortical reorganiza- tion [19]. The methods can promote neuroplasticity and facilitate the reorganization of cortical maps by engaging the brain’s motor and sensory networks. Studies have shown that the therapy method mental imagery, can modulate neural activity in the sensorimotor cortex and reduce PLP intensity [18]. This therapeutic effect may be attributed to the restoration of cortical organization, where the brain learns to adapt to the absence of the limb and alleviate pain sensations. 1.2.2 Current therapy methods There is currently no therapeutic method that is effective for all patients in reduc- ing PLP. Attempts to reduce this pain have therefore been researched to find the underlying cause for both the neuropathic and nociceptive pain as well as find effec- tive rehabilitation therapy methods. Clinical trials and studies have found potential treatments that work for some but not all patients by using mirror therapy, virtual mirror therapy, motor imagery, or motor execution [[10], [20]]. Mental imagery is 3 1. Introduction Figure 1.1: The cortical mapping of the somatosensory cortex, highlighting how different regions of the cortex are linked to specific body parts. The image was adapted from Nursing Hero and licensed under CC BY 3.0 US. included in these methods and has been established to facilitate cortical reorganiza- tion, alleviating PLP [[18], [21]]. Mirror therapy was introduced in the 1990s as a therapy treatment to reduce PLP [10]. It works by having the amputated limb concealed with a mirror which dis- plays the healthy limb in the visual space of the phantom limb. The patient then performs the intended movements with the phantom limb while simultaneously mov- ing the healthy limb with the equivalent movement. This provides visual feedback through the mirror that the phantom limb is moving as intended and is thought to 4 1. Introduction result in changes in the somatosensory and primary motor cortex, relieving pain [11]. Motor imagery involves mentally simulating a movement without physically per- forming it, engaging areas in the brain similar to those activated during actual movement [22]. There are two types of motor imagery, implicit and explicit. The implicit motor imagery, also called laterality recognition, is a therapy form where the patient must differentiate between images of the left or right side of the body. Persons with limb loss are generally slower and less accurate at determining whether the body part is on the left or right side, especially if their loss was on their pre- dominant side [23]. This therapy improves performance in lateral recognition tasks and activates somatosensory and motor areas opposite to the phantom limb [24]. The explicit part of the motor imagery involves the patient actively thinking about a movement that is being displayed on a screen and performing this movement with the phantom limb [22]. Combining therapy methods and performing them in series can reduce the intensity of PLP [11]. The Graded Motor Imagery is a method that includes implicit and ex- plicit motor imagery together with mirror therapy. Contradicting conclusions have been made about the effectiveness of this sequential therapy method and a study including a larger sample of people is needed [22]. Phantom Motor Execution (PME) involves generating phantom movements by en- gaging both central and peripheral circuits, leading to muscular activation at the residual limb [2]. While mirror therapy is a common approach to facilitate PME, it lacks the ability to measure actual motor output. This introduces uncertainty, as individuals may focus solely on visual feedback from the contralateral limb, neglect- ing physical execution in the residual limb which might play a role in pain relief. However, employing myoelectric decoding of motor volition at the residual limb ver- ifies movement execution. This approach relies on detecting muscular contractions as a reliable physiological response to motor execution. Obtaining phantom motor intention from the muscular activity at the residual limb confirms the activation of the relevant parts of the central and peripheral nervous system. This phantom motor intention can then be utilized to provide feedback to the user through various media such as Virtual Reality (VR), Augmented Reality (AR), and serious gaming by incorporating these elements into the therapy. 5 1. Introduction 1.3 Neuromotus Integrum AB developed Neuromotus, a CE-marked and medical device directive (MDD) Class 1 medical device, used in rehabilitation to reduce PLP for people with limb loss effectively. It uses VR, AR, and serious gaming focused on PME with the help of pattern recognition. 1.3.1 Hardware The Neuromotus kit consists of six components illustrated in Figure 1.2. Their func- tion is the following: • 1: The Neuromotus main device with the function of sending the electrode signals to the dongle. • 2: The dongle connects the main device to a computer. • 3: Electric leads with eight channels connecting the main device to the elec- trodes. • 4: Fiducial marker used for the AR function to detect the residual limb loca- tion. • 5: Charger for the main device. • 6: Leg band with belt clip. Figure 1.2: Neuromotus user kit. 6 1. Introduction 1.3.2 Software When starting up the Neuromotus program, a user profile must either be created or selected, providing the program with the patient’s personal information. The user interface is then divided into three segments illustrated in Figure 1.3. It starts with a signal check, followed by a movement selection and recording phase, and limb control can then be used with the built-in functionalities. Figure 1.3: Neuromotus user interface. Before entering any of these sections, the settings tab can be opened and the layout is illustrated in Figure 1.4. Here you can define the movements to record, the left or right limb, which movements are used for the game control, and advanced settings 7 1. Introduction for more control over the calibration sequence. The user can then move on to the first section. Figure 1.4: Neuromotus settings for a user with upper limb loss. The first section is the connection part where the system ensures the dongle has established a connection with the main device before the user can move on to the next step. All of the electrode channels can be viewed in the signals window to confirm that all channels produce the expected output. Secondly, in the recording sequence, the user selects the desired movements such as open/close hand, flex/extend hand, and supination/pronation, from the Movements button if not previously selected in the settings tab. Subsequently, the user initiates a Record session and performs the selected movements. The therapist or user can select a movement duration and rest time which is three seconds by default. The contraction/rest sequence is performed three times and the resulting data is fed to a pattern recognition algorithm capable of classifying phantom limb movements. Thirdly, the main functions of the treatment, consist of VR, AR, serious gaming, and Target Achievement Control (TAC) test that can be initialized. They are all based on pattern recognition for control and provide different objectives for the user. In VR, the user controls a virtual arm for easy visualization with the freedom to perform the selected movements at their own pace and desired order. In AR, a fidu- cial marker is placed on the user’s residual limb to display an AR limb in its original place in the AR space. The AR limb will then follow the user’s position through the marker allowing back/forth and rotational positional changes while still performing the selected movements. Another function is the TAC test which measures how well the user can match defined limb positions and receives feedback on how well it matched. 8 1. Introduction For the serious gaming function, specific phantom limb movements, such as opening and closing the hand, are translated into keyboard inputs for controlling a rac- ing game and Breakout, as illustrated in Figures 1.5a and 1.5b, respectively. The racing game features four maps of varying difficulty and user-changeable driving parameters, while Breakout includes features such as a score, a high score, and user- changeable ball and paddle parameters. (a) Racing game. (b) Breakout. Figure 1.5: Current games included in the Neuromotus package. Figure 1.6: Neuromotus Questionnaire interface. A Questionnaire for PLP (Q- PLP) tracking can be filled in after a session to track recent progress and pain development over time, illustrated in Figure 1.6. It contains a questionnaire regarding pain intensity, loca- tion, type of pain, and change over time. It is possible to only use the visual analogue scale for a quicker evaluation, this is also included in the questionnaire. The pain distribution is used to evaluate how the pain dif- fers throughout the day. The results are then displayed in the Results section. This is used as a metric for the therapist, user, and researchers to track progress and evaluate the result of the treatment. 9 1. Introduction 1.4 Previous related Neuromotus studies Studies have examined the effects of using PME while utilizing pattern recognition, VR, AR, and serious gaming [26]. Three main studies have been completed, the first was a case study that used the principles of PME, but the actual Neuromotus device was not yet created. The second and third studies were clinical trials that used Neuromotus for the PLP treatment. 1.4.1 The first case study The first study was published in 2014 and introduces a system that combines myo- electric signals from the patient’s residual limb with AR, allowing direct volitional control of a virtual limb [26]. The case study involved a 72-year-old male person with limb loss who experienced PLP for 49 years and has unsuccessfully tried var- ious treatments, including mirror therapy, drugs, acupuncture, and self-hypnosis. The intervention utilizes BioPatRec, an open-source myoelectric control platform, to predict the patient’s intended movements. The AR environment is created using a webcam capturing the patient’s surroundings. Pain levels were monitored using the short-form McGill pain questionnaire, revealing a gradual reduction in pain intensity and the emergence of pain-free periods after multiple sessions. The patient gained the ability to voluntarily control the virtual limb, even outside the laboratory. Changes in phantom limb perception were also observed, with the patient transitioning from a closed fist position to a mid-open hand position. The article suggests that the combination of myoelectric control, AR, and gaming holds promise for PLP treatment. The improved quality of life and pain reduction in the patient indicate the effectiveness of the proposed system, calling for further exploration in a broader PLP population. The study underscores the non-invasive, portable, and open-source nature of this innovative technology. 1.4.2 The second study The researchers of this clinical trial introduced a novel therapy for chronic PLP, fo- cusing on promoting PME [20]. This non-invasive approach, grounded in principles of brain plasticity, aimed to address the limitations of existing treatments for pa- tients who had experienced upper limb amputations. The clinical trial highlighted the potential efficacy of the therapy and its application in cases where conventional methods had previously failed. Over a period from September 2014 to April 2015, 14 patients with refractory chronic PLP participated. The average age was 50.3 years, with a mean PLP duration of 10.3 years. Previous treatments had been unsuccessful for an average of 5.6 years before enrollment. The outcomes demonstrated a consistent reduction in PLP across 10 1. Introduction various metrics, with sustained improvements noted in pain intensity, quality, and interference with daily activities and sleep. The proposed technology, requiring vo- litional control of musculature at the residual limb, exhibited potential benefits but acknowledged limitations for certain patient groups. While the clinical trial revealed promising results, including a 50% reduction in PLP for many participants, it underscored the need for further research. Lim- itations, such as the absence of a control group and potential bias in follow-up interviews, were acknowledged. Technological feasibility and future applications of the proposed therapy beyond PLP were discussed, emphasizing its integration into a treatment–evaluation system. Overall, the findings suggested a novel and potentially effective avenue for addressing chronic PLP, though additional clinical evidence and exploration of underlying mechanisms are warranted. 1.4.3 The third study This clinical trial explores the process of identifying user types for a technological therapy PME, designed to address PLP [27]. The methodology involves conducting ethnographic unstructured interviews and employing the KJ methodology for anal- ysis. The information gathered from interviews is categorized into themes, leading to the recognition of two distinct user types: Goal-Oriented Users and Experiential Users. The Goal-Oriented Users, categorized as User Type I, are individuals who measure their progress based on the completion of specific goals. They exhibit a preference for goal-oriented activities, such as the TAC test, and seek feedback related to goal achievement. On the other hand, the Experiential Users, identified as User Type II, perceive the therapy through the AR experience. They prioritize the realistic qualities of the virtual limb and tend to base their practice on time markers rather than specific movement goals. The discussion highlights the adaptability of PME for PLP treatment at home. Ad- herence to the therapy is dynamic and influenced by factors such as pain levels, patient needs, motivations, and daily life contexts. The clinical trial emphasizes that the collected pain data may not solely reflect the effectiveness of the treatment but rather how pain influences therapy usage patterns. Barriers to adherence are identified, including the time demands of the therapy ses- sions and the necessity for reusable electrodes. It suggests recommendations for designing a device compatible with both user types, such as incorporating a time display for Experiential Users and providing feedback on results for Goal-Oriented Users. Additionally, enhancements in AR realism and compatibility with diverse user populations are proposed. 11 1. Introduction The clinical trial acknowledges limitations, such as a small sample size (n=4), af- fecting generalizability. Future research is encouraged to conduct a more systematic investigation over an extended period. In conclusion, the methodological relevance of the clinical trial extends beyond PLP, offering insights into the home use of health technologies. The combined expertise of medical anthropology and human-machine interface design provides a holistic understanding of patient-led in-home care, in- forming the design of interventions that enhance technology’s capacity and relevance. 1.5 Medical device regulations The MDD, formally known as Council Directive 93/42/EEC, was the foundational regulatory framework established by the European Union to ensure the safety, ef- ficacy, and quality of medical devices within the EU market [28]. This directive required devices to meet essential safety and performance requirements, undergo conformity assessments by notified bodies, and be classified into risk categories with corresponding regulatory obligations. The MDD required CE marking, which signi- fied a device’s compliance with EU regulations and allowed its free movement within the European Economic Area. However, the MDD had limitations that necessitated a more robust framework. The Medical Device Regulation (MDR), which came into effect as EU Regulation 2017/745, replaced the MDD to address these limitations [29]. The transition from MDD to MDR was driven by the need for stricter controls, prioritizing the safety and performance of medical devices, prompted by incidents involving faulty devices that underscored the shortcomings of the MDD [30]. The MDR introduced en- hanced transparency and traceability measures, including the Eudamed database and Unique Device Identification system, to ensure better oversight. Additionally, the MDR placed greater emphasis on clinical evidence and post-market surveillance, requiring continuous monitoring of device safety and performance. The scope of the MDR was expanded to cover a broader range of products, including certain software and devices with non-medical purposes, and aimed to harmonize regulatory require- ments across the EU, improving compliance and enforcement mechanisms. 1.5.1 Regulatory implications of changing a MDD classed medical device Article 120 of the MDR addresses the transitional provisions for devices that were already on the market under the MDD or the Active Implantable Medical Devices Directive 90/385/EEC [29]. This article ensures a smooth transition from the old directives to the new regulation, providing clarity and continuity for manufacturers and other stakeholders. When a change is made to a medical device, it can be regarded as either significant 12 1. Introduction or non-significant. According to MDR Article 120(3c), point (b), there can be no significant changes in the design and intended purpose [29]. To determine whether a change is significant or not, the Medical Device Coordination Group (MDCG) has created a document providing guidance on significant changes related to Article 120 of the MDR [31]. A flow chart, as illustrated in Figure 1.7, is used to guide the classification process. Additionally, connecting child flowcharts (A-E) are provided in Appendix C. 1.6 Purpose This study aims to identify and define specific improvements to the Neuromotus system, focusing on improving usability and functionality. A qualitative pre-study aims to address and define existing challenges and limitations. The insights gained will guide the selection of targeted improvements which will be implemented and validated in preparation for the clinical study HOPE, to advance the quality of the treatment. 1.7 Delimitations To narrow the scope of this project and maintain conciseness, several delimitations were established. These limitations were crucial for shaping the design and method- ology of the study, ensuring focused research. The specific delimitations are outlined below: • Pre-study: The pre-study will only involve therapists. • Neuromotus: The improvements on Neuromotus will solely focus on the software. • Regulatory: From a regulatory standpoint, the changes made to the system will be non-significant. • Testing: User tests will not be performed. 13 1. Introduction Figure 1.7: Significant changes regarding the transitional provision under Article 120 of the MDR (Source: MDCG 2020-3 Rev.1) [31]. 14 2 Pre-Study This chapter introduces the pre-study conducted through meetings and semi-structured interviews with a focus on data acquisition and feedback collection. The results from this pre-study will guide the remainder of the project. 2.1 Introduction The Neuromotus device is utilized in clinics across Europe. Due to the stringent laws governing the handling of personal information, patients were excluded from being contacted. Therefore, the focus is placed on reaching out to physiotherapists. Semi-structured interviews were chosen as the primary method for gathering in- formation. This approach is advantageous because it combines the flexibility of open-ended questions with the focus of structured interviews, allowing for in-depth exploration of specific topics while still providing room for respondents to express their thoughts and experiences freely. This method is particularly effective for ob- taining detailed and nuanced insights from experts, such as physiotherapists, who can provide valuable information about the practical use and impact of the Neuro- motus device in clinical settings. 2.2 Methods The pre-study was divided into two parts, the first one was a meeting together with representatives from the upcoming HOPE study where they presented feed- back from previous patients using the Neuromotus device. The second part was semi-structured interviews where the interview questionnaire found in Appendix A, acted as a guideline for the interviews. The interviews were conducted with two physiotherapists having worked with Neuromotous to alleviate PLP for amputees. Transcribed interviews are found in Appendix B. From the meeting, the following feedback was provided by the HOPE representa- tives regarding patient experiences: • Games ▶ The games are not intuitive for people with lower limb amputations. Allowing games to be controlled up/down instead of left/right would be 15 2. Pre-Study more suitable for knee/foot movements since they inherently move along that axis. ▶ The games are too monotonous, leading to decreased adherence over time. ▶ Patients desire more advanced features and progress tracking. • VR ▶ Generally enjoyed by patients. • AR ▶ The limb in the Augmented Reality environment appears unnatural and the skin color is not adjustable. ▶ The camera position is difficult to adjust for lower limb amputees. • TAC Test ▶ The TAC Test and its feedback features were well-received. • Hardware ▶ The lanyard around the neck does not fit all body types. • General ▶ It should be communicated that pain levels will likely increase at the beginning of therapy. The questions in the semi-structured interviews were divided into two parts. The first part focused on patient feedback, where therapists provided insights from treat- ing and conversing with patients during rehabilitation sessions. The second part focused on therapist feedback, where the therapists described their own experiences with the Neuromotus device. Data from these two parts were then combined and categorized into sections cov- ering various aspects, including games, VR, AR, TAC tests, hardware, and general feedback. Below is a compilation of the combined feedback from both sets of inter- views. • Games ▶ The car game lacks a defined objective, resulting in a sense of aimlessness and diminished engagement. ▶ Breakout proved more enjoyable when a clear objective was established. ▶ The games provided a refreshing and engaging alternative to other con- ventional treatment methods. ▶ The synchronization between the contractions and movements of the bar or car was suboptimal for certain participants, leading to frustration. This issue primarily stemmed from patients becoming overly excited and inadvertently disrupting the cables causing movement artefacts which lowered the precision of the movement decoding. ▶ Incorporating more engaging games could enhance patient involvement, as some found the current selection to become monotonous over time. ▶ Including additional goal-oriented games with feedback mechanisms could enhance patient engagement and satisfaction. ▶ The current design appears outdated, lacking the desired level of excite- ment and engagement. Additional fun and interactive games are desired. 16 2. Pre-Study • VR ▶ The inclusion of a more realistic limb would enhance the overall experi- ence. ▶ Some movements extend beyond the screen, rendering them invisible to the patients. This poses a challenge as it becomes difficult for them to discern their actions, particularly with regard to finger and foot move- ments. ▶ Some movements appear anatomically implausible, resulting in a visually awkward appearance. ▶ Improving the graphics for a more realistic representation would enhance the overall experience. • AR ▶ The skin tone representation on the AR limb appeared unnatural and did not accurately match the diverse skin tones of some users, causing a sense of discomfort or disconnect. ▶ The graphics appeared outdated. ▶ A more realistic AR limb would enhance the experience. ▶ A portion of the patients expressed positive feedback, perceiving it as a valuable tool. ▶ Some therapists, particularly those who are older, may find it challenging to accurately position the limb. ▶ This was a popular and highly useful tool for treatment purposes. • TAC Test ▶ The guidance on movements and the feedback offered were helpful and appreciated. ▶ The patients frequently expressed appreciation for and enjoyment of the experience. ▶ The effectiveness stemmed from its goal-oriented nature. ▶ Incorporating features to display the number of repetitions, range, and time would be beneficial. ▶ This was a popular and highly useful tool for treatment purposes. • Hardware ▶ Generally, the experience was satisfactory, though occasional challenges were encountered regarding the placement of the main device ▶ The connection between the main device and the dongle occasionally became unstable or disappeared. ▶ Improved guidelines for electrode placement would be beneficial. • General ▶ A minor delay between contraction and the corresponding movement on the screen was noted. ▶ The addition of VR glasses could enhance the enjoyment of the experi- ence. ▶ The arm’s mirrored presentation at times reduced intuitiveness for pa- tients. Adjusting the perspective could improve user experience. ▶ Simplifying medical terminology and providing illustrated guides for pa- tients could improve comprehension and usage. 17 2. Pre-Study ▶ Including sound cues as auditory feedback upon task completion. ▶ Providing a final report detailing the number of sessions, session lengths, utilized movements, and progress over time. ▶ Ensure clarity regarding potential initial pain increases at the beginning of treatment, possibly by including this information in the manual. ▶ Enhancing the range of thumb movements. From the data acquisition, potential improvements to the system were identified from both the therapists’ and indirectly the patients’ perspectives. A list of criteria was established, including aspects about regulatory implications, project feasibility, and stakeholder priorities, to determine what to implement and improve in Neuromotus. The following list of criteria was used indiscriminately to guide the decision-making. • User Feedback: The indirect input provided by previous users will be of great importance, as their experiences are invaluable in pinpointing areas for enhancement. • Therapist feedback: The potential impact of suggested improvements on patient outcomes and clinical workflows will be assessed to prioritize changes that yield the greatest benefits. • Regulatory Implications: Consideration will be given to any new regulatory requirements introduced by the proposed modifications, ensuring compliance with relevant standards. This will minimize potential complications in device deployment and ensure that the changes are classified as non-significant. • Project Feasibility: The feasibility of implementing each improvement within the allocated time frame and resource constraints will be evaluated to ensure realistic project planning and execution. • Stakeholder Priorities: The viewpoints of Integrum, as well as other key stakeholders connected to the HOPE study, will be solicited to ascertain their perspectives on prioritizing improvement initiatives. 2.3 Results The information acquired from the meeting with the HOPE representatives and interview feedback was compiled into concrete improvement suggestions presented below: 1. Make changes for the serious gaming function i Conduct a thorough investigation of other open-source games, compar- ing them to previous findings to identify any recent developments or im- provements. This exploration aims to discover games that offer enhanced features or gameplay experiences, which could be considered for inclusion in Neuromotus. ii Enhance the existing games by incorporating user feedback mechanisms to provide real-time feedback and increase user engagement. Introduce competitive elements to foster a sense of competition among users and 18 2. Pre-Study make the experience more dynamic. Additionally, implement progress tracking features to allow users to monitor their improvement over time. iii Incorporate a proportional controller to provide users with more control over their interactions. 2. Enhance the augmented reality environment i Implement the ability to customize colors based on the patient’s skin tone to enhance inclusivity and personalization of the experience. ii Improve the tracking system of the fiducial marker in the AR environ- ment. iii Identify movements that extend beyond the screen boundaries or exhibit anatomical inaccuracies. 3. Create a final report detailing the number of sessions, session lengths, utilized movements, and progress over time. 2.4 Final improvements of Neuromotus The improvement suggestions presented in the results were discussed with stake- holders to create a prioritization list. Five main improvements were defined, the first two (1.(i), 1.(ii)) being mandatory and the remaining three to be addressed in accordance with the list as time permits. The first three are related to the serious gaming functionality, the fourth is related to the AR environment, and the last one is an extension of the Q-PLP result section: 1. (i) Conduct a thorough investigation of other open-source games, comparing them to previous findings to identify any recent developments or improvements. This exploration aims to discover games that offer enhanced features or gameplay experiences, which could be considered for inclusion in Neuromotus. 1. (ii) Enhance the existing games by incorporating user feedback mechanisms to provide real-time feedback and increase user engagement. Introduce compet- itive elements to foster a sense of competition among users and make the experience more dynamic. Additionally, implement progress tracking features to allow users to monitor their improvement over time. 1. (iii) Integrate proportional control to offer users an alternative controller that en- hances precision during gameplay and simultaneously trains muscle contrac- tion strength. 2. (i) Implement the ability to customize colors based on the patient’s skin tone to enhance inclusivity and personalization of the experience. 3. Create a final report detailing the number of sessions, session lengths, utilized movements, and progress over time. 19 2. Pre-Study 20 3 Methodology This chapter describes the methodological approach employed to implement and validate the improvements to the Neuromotus device defined in the pre-study. 3.1 Improvements Based on the results from the pre-study, an improvement prioritization list was created. The first four items on that list were implemented, while the fifth item, which involved creating a final report, was not implemented due to lack of time. 3.1.1 Improvement 1.(i) Finding suitable open-source games The two games accompanying Neuromotus were found and implemented before the second clinical trial in 2014. It was considered likely that now there would be new or updated open-source games available. To find suitable games, selection criteria had to be defined. The criteria were composed of both objective and subjective criteria defined as: 1. Appropriate difficulty level for myoelectric control i Games without mouse movements ii six or fewer controlling keys iii Easy to control 2. Have a license allowing use, modification, publishing and distribution rights, and software selling, such as the MIT and GNU General Public License. 3. Compatible with modern operating systems 4. Improved graphical interface or new features compared to the current Neuro- motus games 5. Ability to play as a single-player The open-source games market was investigated by using the search terms Open- source games, and Open-source racing games in Google. This produced four primary sources of information to explore the available games. The first one was Wikipedia which provided a list of 106 notable open-source games with open engine and free data [32]. The second one was SourceForge and provided 9701 open-source games [33]. The third and fourth were two Github profiles providing lists with 144 and 77 21 3. Methodology open-source games respectively [[34], [35]]. From this, seven games fit the criteria enough to be further examined by playing the games and exploring features. A short description of each game is presented in Table 3.1. Table 3.1: Game Information Game Description Genre License Stepmania Stepmania is a rhythm-based music game where the goal is to match arrows in rhythm with a song with the aid of the arrow keys. Music MIT License Speed Dreams Speed Dreams is a racing car simulator with accurate driving behavior, multiple physics engines, and high adaptability. Racing GNU General Public License version 3.0 (GPLv3) HexGL HexGL is a futuristic racing game where the vehicle is a spaceship. Racing MIT License 2048 2048 is a puzzle game where you move numbered tiles to match and merge them into higher numbers with the help of the arrow keys. The numbers are powers of two and the goal is to reach 2048. Puzzle MIT License Pong Pong is an old game featuring two paddles and a ball with the objective of getting the ball past the opponent’s paddle. The paddle is controlled with the up/down keys. Arcade MIT License Fluid Table Tennis Fluid Table Tennis is a game based on the traditional pong game with the addition of fluid dynamics where each player can shoot plasma to steer the ball without it having to touch a paddle. It is controlled by the up/down arrow keys to steer the paddle and the right/left arrow keys to control the fluid plasma. Arcade MIT License Snake Snake is a game where a growing snake is controlled within a confined space without steering it into a wall or itself. The snake grows when it eats a fruit that is spawning in a random location on the map. Arcade MIT License 22 3. Methodology 3.1.1.1 Game evaluation and elimination After further examination, the games were reevaluated against the criteria to deter- mine their suitability for Neuromotus. The seven games and their alignment with the five criteria are shown in Table 3.2. If a game fulfills all the criteria, it should proceed to the next stage and be highly considered for inclusion in the Neuromotus package. Table 3.2: Game evaluation Criterion Stepmania Speed Dreams HexGL 2048 Pong Fluid Table Tennis Snake 1 ✓ X X ✓ ✓ X ✓ 2 X ✓ ✓ ✓ ✓ ✓ ✓ 3 ✓ ✓ ✓ ✓ ✓ ✓ ✓ 4 ✓ X X ✓ ✓ X ✓ 5 ✓ ✓ ✓ ✓ X ✓ ✓ Below is a summary of the games under consideration, along with an explanation of how each game meets the specified criteria: • Stepmania: The Stepmania music game’s objective is to match key presses to keys appearing on the screen in rhythm with the music. There are adaptability features, multiple game modes, and difficulty levels for the user to choose from. However, the three songs that were included in the download did not fall under the same MIT License. This meant new songs had to be created or imported with all keys manually inserted. Hence, it does not meet the 2nd criterion and is thus excluded from the list. • Speed Dreams: The Speed Dreams racing game provides realistic driving behavior and has high adaptability regarding the physics engine, controller, and game modes. There are four track categories with multiple tracks in each category, as well as a variety of cars. When driving, multiple angles can be chosen to get the user’s preferred point of view. The steering of the car can be too realistic which makes it difficult to stay on track and control the car with myoelectric control. It is therefore eliminated from the list for being too difficult to control and does not fulfill the 1st criterion. • HexGL: The HexGl racing game is a fast-paced game where the objective is to drive three laps around a track. It has few features and adaptability possibilities regarding handling and speed control which makes it challenging to adjust the difficulty level. There is also only one track which can make it monotonous and less motivating over time. It does not fulfill the 1st and 4th criteria because of the difficulty level and lack of features resulting in elimination from the list. • 2048: 2048 is a turn-based puzzle game where the player each turn has the possibility of sliding numbered tiles to match and merge the numbers increas- 23 3. Methodology ing them as powers of two. E.g. when a four slides into a four it merges and becomes an eighth and when a 1024 slides into a 1024 it merges and becomes a 2048 which makes you win the game. The game is easy to understand yet requires strategy. It displays scores and high scores to keep track of your progress. It fulfills all of the criteria and remains on the list. • Pong: In Pong, the objective is to bounce the ball from your paddle such that the opponent misses the ball with their paddle. It keeps the score from the ongoing round but not after the game has ended. This version of Pong requires a second player as the opponent and playing against the “computer” is not an option, this was the case for all of the open-source Pong games found during the game searching phase. It does not fulfill the 5th criterion and is therefore eliminated from the list. • Fluid Table Tennis: The Fluid Table Tennis arcade game is a more ad- vanced version of Pong where the objective is the same, get the ball past the opponent’s paddle. However, a new element is now introduced, a plasma can- non that both players shoot out from their paddle that interacts with the ball and the other player’s plasma according to fluid dynamics laws. It creates a high difficulty level and makes it difficult to control. It does therefore not fulfill the 1st criterion and is eliminated from the list. • Snake: In the arcade game Snake, the objective is to control the head of a snake-like object and prevent it from colliding with itself or the walls. Points are rewarded by gathering fruits located in a confined space, when fruits are gathered, the snake grows one block longer. It has score and high score display features in combination with easy control and a modern graphical interface. All of the criteria are fulfilled and it remains on the list. From Table 3.2, it’s evident that five games have been eliminated due to their in- ability to meet all criteria outlined for inclusion. However, the puzzle game 2048 and the arcade game Snake have successfully navigated this evaluation process, as demonstrated in Table 3.3. These two games are now considered to be included in Neuromotus. It’s worth noting that they may undergo further modifications to align them more closely with the requirements of the PLP rehabilitation process facilitated by Neuromotus. Table 3.3: Game elimination Criterion Stepmania Speed Dreams HexGL 2048 Pong Fluid Table Tennis Snake 1 ✓ X X ✓ ✓ X ✓ 2 X ✓ ✓ ✓ ✓ ✓ ✓ 3 ✓ ✓ ✓ ✓ ✓ ✓ ✓ 4 ✓ X X ✓ ✓ X ✓ 5 ✓ ✓ ✓ ✓ X ✓ ✓ 24 3. Methodology 3.1.2 Improvement 1.(ii) Improving the games Improvement number two could then commence, and in-game implementations such as speedometers, leaderboards, and the possibility of adjusting the difficulty were implemented. This was done by accessing the games’ source code written in HTML for the structure and content, CSS for the control and styling of the HTLM element, and JavaScript for adding interactivity and dynamic behavior to the game. 3.1.2.1 Racing game The racing game, one of the previously included games in the Neuromotus package, needed updates to make it more interactive and engaging to increase user adher- ence. The game currently features four different maps with varying levels of diffi- culty. Three key changes were made to enhance the game, first, a top 5 leaderboard was added for all four maps, displaying rank, time, and date. Second, a speedome- ter was introduced to track the car’s speed, allowing for easier adjustments to the difficulty level. Lastly, a lap timer to observe how well the lap is going. A compari- son of the improvements before and after implementation is illustrated in Figure 3.1. (a) Old racing game version (b) New racing game version Figure 3.1: Comparison of the old and new version of the racing game. 3.1.2.2 Breakout Breakout is another game included in the Neuromotus package that required up- dates to make it more intuitive for lower-limb amputees to move the paddle. This was achieved by adding a rotate button, allowing the user to rotate the playing field by 90 degrees. Consequently, the controller was updated so that the up and down arrow keys were used instead of using the left and right arrow keys. The comparison between the old and new versions of Breakout is illustrated in Figure 3.2 In preparation for improvement number 1(iii) (Integrate a proportional controller), two new checkboxes were created to choose the controller mode. The Joystick op- tion is a proportional controller that allows the user to move the paddle with higher precision. 25 3. Methodology Smaller improvements were also made, such as changing the location of the score and lives and updating the icon representing lives from a paddle to a ball. (a) Old Breakout version (b) New Breakout version Figure 3.2: Comparison of the old and new version of Breakout. 3.1.2.3 2048 2048 is one of the new games in the Neuromotus package and already includes desir- able features for increasing adherence, necessitating only two minor changes. First, the Try again button illustrated in Figure 3.3a that appears when the player reaches 2048 on a tile was removed. Second, feedback timing was adjusted to accommodate the longer time estimated for users to reach 2048 using myoelectric control compared to manual control. It was decided that positive feedback should be given for every new power of 2 starting from 28, as illustrated in Figure 3.3b where feedback is given upon reaching 512 on a tile. 3.1.2.4 Snake The second new game introduced was Snake, which held promise due to its simplicity and high recognizability. However, several adjustments were necessary to enhance its adaptability and competitiveness, with the aim of increasing user adherence. A comparison of the old and new versions of Snake is Illustrated in Figure 3.4 The first change involved adding a top 5 leaderboard displaying rank, score, and date, akin to the racing game. The second change was the inclusion of a speed controller, allowing users to adjust the snake’s speed and thereby tailor the game’s difficulty. Snake is traditionally played using four arrow keys, and the user is then required to accurately generate four movements for Neuromotus to classify. The new version offers an alternative control scheme. Since the snake continually turns left or right, users can effectively navigate with two keys, requiring only two movements. Buttons enabling this simplified control were integrated into the game interface. 26 3. Methodology (a) Old 2048 version (b) New 2048 version Figure 3.3: Comparison of the old and new version of 2048. (a) Old Snake version (b) New Snake version Figure 3.4: Comparison of the old and new version of Snake. 3.1.3 Improvement 1.(iii) Integrate a proportional controller The third improvement involved integrating a human interface device (HID), such as a joystick, to serve as a proportional controller. This allows users to exert finer control over games by modulating the force with which they contract their muscles. This functionality was implemented using vJoy, an open-source virtual joystick de- vice licensed under the MIT license. vJoy is recognized by the operating system as a physical joystick, with state information about its axes and buttons being provided by a feeder application. By manipulating this feeder application, users can play 27 3. Methodology games that support joystick input, thus increasing the versatility and compatibility with third-party games. Unlike binary key presses, the system transmits an amplitude value derived from the root mean square (RMS) of the EMG signals for the movements. Each move- ment undergoes a calibration phase, during which the user performs the movement repeatedly for r iterations. Post-calibration, the system enters a pre-processing phase, where the RMS is computed for n window samples across all electrode chan- nels (Ci = {c1i, c2i, . . . , cni}). To mitigate the influence of noise and artifacts, the 75th percentile RMS is calculated for each channel. The channel with the maximum 75th percentile RMS is then identified, and this value serves as a reference in the real-time session. The relevant equations are presented below: Ci = {c1i, c2i, . . . , cni} (Channel i with n windows) P75i = cki, where k = ⌈0.75 · n⌉ (75th percentile RMS) imax = arg max i P75i (Channel index) P75 max = P75imax (Max 75th percentile RMS) This reference value is then used to calculate the ratio between the reference RMS and the real-time RMS value derived from m window samples, representing the amplitude of the muscle contraction. The resulting amplitude is normalized to a range of 0 to 1 where 0 is the floor noise threshold. Any values exceeding 1 (i.e. exceeding the maximum 75th percentile) are being capped at 1, as described by the following equations: RMSinow = √√√√ 1 m m∑ j=1 c2 jinow P75inow = RMSinow A = P75imax P75inow , where A ∈ [0, 1] if P75inow ≤ P75imax A = 1 if P75inow > P75imax In addition to allowing users to play more third-party games, this functionality was also implemented to work with Breakout. This was achieved by creating a web socket as the communication channel, enabling the Neuromotus application to send amplitude data that would otherwise be sent to the joystick feeder. Since the web- based game cannot receive real joystick data from an HID, this method provided a straightforward implementation The proportional controller was extended to support simultaneous movements which is a feature allowing combinations of movements to be classified as one. These simul- taneous movements are classified in the same way as individual movements, rather 28 3. Methodology than as combinations of two classified movements. Consequently, it is impossible to differentiate between the amplitudes of the two movements in a simultaneous action. Instead, it is assumed that the movements have equal amplitude. This assumption is illustrated in Figure 3.5, which shows the four movements: Flex hand, Extend hand, Open hand, and Close hand. When contradictory movement combinations, such as Open hand + Close hand, are excluded, four possible com- binations of simultaneous movements remain. Contradictory movements are repre- sented in opposite directions along an axis. The allowed simultaneous movements fall along the lines defined by the linear functions f(x) = x and f(x) = −x, causing the two movements in any simultaneous action to have the same amplitude. The empty square in the middle represents the area where the amplitude is too low for the algorithm to classify the movement as anything other than rest. Figure 3.5: Illustration of simultaneous movement limitations. 29 3. Methodology 3.1.4 Improvement 2.(i) Customize AR limb color The fourth improvement involved adding different colored limbs in the AR environ- ment for the user to choose from. The monk skin tone scale consists of 10 skin tones and claims to balance diversity with ease of use, the scale is illustrated in Figure 3.6 [36]. It was developed to be used in the field of computer vision to reduce skin tone bias in machine learning. Figure 3.6: Monk skin tone scale. The AR environment is built using the Ogre3D 3D graphics engine and employs material files to define the texture of the limb. New materials were created program- matically using Matlab, streamlining the process and allowing for easy adjustments. This method enables the creation of new textures for both the arm and leg. Users can then select the desired limb color through the Neuromotus settings interface. It is also possible to change the color of the limb using keyboard keys 1-10, where each key number corresponds to a skin tone number on the Monk skin tone scale. Figure 3.7 illustrates the difference between skin tone 1 and skin tone 10. (a) Skin tone 1 on the Monk skin tone scale. (b) Skin tone 10 on the Monk skin tone scale. Figure 3.7: Comparison between the 1st and 10th skin tone on the Monk skin tone scale. 30 3. Methodology 3.2 Tests of the new improvements The tests consisted of two parts, the stakeholder feedback phase and the validation tests. 3.2.1 Stakeholder feedback The stakeholders included representatives from the HOPE study and employees at Integrum who were previously familiar with Neuromotus. They used the upper arm textrode band, developed for the HOPE study, to test the new improvements. Feedback from the HOPE representatives was obtained by asking their opinions on the changes and whether these could improve user compliance. At the time of the meeting with them, the ability to change the skin tone of the AR limb was not yet implemented. Participants from Integrum also used the textrode band and tested the new func- tionalities and changes. The evaluation did not follow a strict protocol, and some participants did not have the time to test all the improvements. During this test, two sizes of the textrode band were available: one smaller and one larger. Given the larger size of the group, participants were asked to fill out an online form con- sisting of multiple-choice questions after their interaction with Neuromotus. This form aimed to gather their opinions on how the changes impacted the device, if they had any preferences among the functionalities, and if they were satisfied with the changes. Since this was the first time most participants had used the device, and there was a learning curve associated with myoelectric control, only features requiring 1 DOF were tested. Consequently, the 2048 game and the four-key control scheme in Snake were excluded from testing. However, participants were shown the 2048 game and how it works. Appendix D presents the questions asked in the form. 3.2.2 Validation tests To validate the changes and ensure that the final release of the updated Neuromotus device works as intended, a test protocol was created with requirements for the de- vice to meet. The following requirements had to be validated during the test phase: • General □ Neuromotus shall launch. □ All Neuromotus functionalities shall work with the keyboard or joystick controller selected. • Augmented Reality □ It shall change the AR limb color according to what’s chosen in the set- tings. □ It shall be possible to change the AR limb color in the AR environment by using key presses from 1-10. □ It shall work for both the left and right limbs. □ It shall work for both the arm and leg. 31 3. Methodology • Serious Gaming □ The racing game shall have a top 5 leaderboard and a speedometer that updates correctly for all four maps. □ Breakout shall have a button to rotate the playing field and change the control from left/right to up/down. □ Breakout shall be able to accept joystick input for proportional control of the game. □ It shall be possible to open and close the game mode in the Neuromotus interface and still connect and send information to games through the web socket. □ The 2048 game shall give positive feedback for every new power of 2 from 28 and the restart button shall be removed. □ The Snake game shall have a top 5 leaderboard, speed controller, and the possibility to choose between a controller with one or two DOFs. 32 4 Results This chapter presents the results from the feedback collected from HOPE represen- tatives and Integrum employees, along with the validation tests of the new improve- ments. 4.1 Feedback from stakeholders Feedback from HOPE representatives was gathered by asking for their opinions and taking notes, while Integrum employees provided their feedback by filling out an online form. 4.1.1 HOPE representatives Feedback from HOPE representatives is presented in the bullet points below where they gave their thoughts on the improvements. They only tried the joystick with Breakout and changing the skin tone of the AR limb was not yet implemented. Therefore, they only provided feedback on the four games. • Racing game: The addition of a leaderboard made the game more competi- tive. The lap time feature provided immediate performance feedback, allowing players to see their progress in real-time. • Breakout: They enjoyed playing Breakout with joystick control, finding it easier to maneuver the paddle. This also challenged users to be more conscious of the strength of their muscle contractions. • Snake: The ability to switch between 1 and 2 DOF and adjust the snake’s speed was well-received. However, it was somewhat easy to run into oneself and lose when using the 1 DOF. The leaderboard provided good motivation for improving and achieving a high score. • 2048: They enjoyed the 2048 game, which introduced a different type of gameplay where users could take their time and make strategic moves. This allowed users to rest their muscles between turns, enabling longer play sessions without fatigue. 33 4. Results 4.1.2 Integrum employees Thirteen participants from Integrum took part in the testing and subsequently an- swered a form. The form consisted of five mandatory questions and one optional question, covering the changes that were made. Question one asked the participants which games they played. Breakout and the Racing game were played the most with twelve and eleven participants respectively. Snake was played by three and 2048 by zero. 2048 requires a control scheme using 2 DOF and was therefore not tested since the participants only used 1 DOF. Question two, illustrated in Figure 4.1, a follow-up to the first, inquired about their game preference. Resulting in 69.2% (nine people) preferring Breakout and 30, 8% (four people) preferring the Racing game. Figure 4.1: Feedback form, question two. Question three, illustrated in Figure 4.2, is concerned with the customization of the skin tone of the AR environment limb and how well it matched the users’ actual skin tone. It ranges from "not even close" (1) to a "perfect match" (5). The responses were as follows: 30.8% thought it looked okay, 30.8% thought it looked good, and 38.5% thought it was a perfect match. Question four asked about controller preference when playing Breakout, specifically whether users preferred the joystick or keyboard controller. Among the respondents, 30.8% (four people) did not play with both controllers due to lack of time. Out of the remaining 69.2% (nine people) who tried both the joystick and keyboard con- troller, all preferred using the joystick controller. The fifth question, illustrated in Figure 4.3, was about their general experience with Neuromotus, ranging from "I strongly dislike it" (1) to "I strongly like it" (5). The results were 38.5% liked it and 61.5% strongly liked Neuromotus. 34 4. Results Figure 4.2: Feedback form, question three. Figure 4.3: Feedback form, question five. The final question was an optional open-ended question where participants could express their specific preferences or dislikes. Seven people responded, with summa- rized feedback indicating that the size of the textrode bands was not optimal for all body types, the device felt natural surprisingly quickly, and it was easy to run into oneself in Snake. 35 4. Results 4.2 Tests The validation tests are divided into three categories based on the functionality being tested. Each test consists of a requirement and a qualification method, resulting in a YES/NO outcome depending on the test’s result. 4.2.1 General General tests focus on ensuring that Neuromotus functions without any errors and assess the overall system performance. These tests also cover all the different func- tionalities within Neuromotus with no specific change being tested. Table 4.1: Tests of improvements for general use. Requirement Qualification method Approved YES/NO Neuromotus shall launch without any errors. Launch the new Neuromotus version. YES All Neuromotus functionalities shall work with the keyboard or joystick controller selected. Test all functionalities with both controllers. YES 4.2.2 AR environment limb color customization These tests focus on the AR environment and the option of limb color customiza- tion, ensuring that users can choose colors both in the settings and within the AR environment, with the correct outcome. Table 4.2: Tests of improvements in the AR environment Requirement Qualification method Approved YES/NO It shall change the AR limb color according to what’s chosen in the settings. Launch the AR environment. YES It shall be possible to change the AR limb color in the AR environment by using key presses from 1-10. Launch the AR environment and press the keys 1-10. YES Changing the AR limb color shall work for both the left and right limbs. Launch the AR environment with the left and then the right limb selected. YES Changing the AR limb color shall work for both the arm and leg. Launch the AR environment from lower and upper limb Neuromotus applications. YES 36 4. Results 4.2.3 Games Game tests focus on evaluating all the newly implemented features in the games, ensuring they work without any bugs and perform as intended. Table 4.3: Tests of improvements of the games. Requirement Qualification method Approved YES/NO It shall work to play third-party, joystick-compatible games with the joystick controller selected. Select joystick control and test to play Trackmania in joystick mode. YES The racing game shall have a top 5 leaderboard and a speedometer that updates correctly for all four maps. Launch the racing game and test drive the four maps. YES Breakout shall have a button to rotate the playing field and change the control from left/right to up/down. Launch Breakout and rotate the field. YES Breakout shall be able to accept joystick input for proportional control of the game. Select the joystick control and play Breakout. YES It shall be possible to open and close the game mode in the Neuromotus interface and still connect and send information to games through the web socket. Repeatedly open and close the game mode. YES The 2048 game shall give positive feedback for every new power of 2 from 28 and the restart button shall be removed. Launch 2048 and observe feedback. YES The Snake game shall have a top 5 leaderboard, speed controller, and the possibility to choose between a controller with one or two DOFs. Launch Snake and test functionalities. YES It shall be possible to have the simultaneous setting active and use it in the game mode. Activate the simultaneous setting and test the games. YES 37 4. Results 4.3 Regulatory aspects Neuromotus is a class 1 medical device according to the MDD. The significance of the change has therefore been assessed in accordance with MDCG 2020-3, using the guide charts found in Appendix C. It is important to note that these changes are limited to software enhancements and do not affect the core aspects of the Neuro- motus device. It does not affect: • Main Functionalities: The primary functionalities of the Neuromotus device remain unchanged. The device will continue to perform its core functions as intended, ensuring consistent operation and reliability. • Intended Use: The intended use of the Neuromotus device remains the same. It will be utilized as a tool to facilitate rehabilitation by providing interactive feedback and engagement for users, without altering its original purpose or application. • Hardware: No modifications will be made to the hardware components of the Neuromotus device. The physical design, structure, and technical specifi- cations of the hardware will remain identical to the existing configuration. • User Interface The user interface of the Neuromotus device will not undergo any major changes. Users will interact with the device in the same manner as before, ensuring a familiar and consistent user experience. This information helped guide the answers to the charts as illustrated in Figure 4.4, resulting in the determination that the changes are considered non-significant according to MDCG 2020-3. 38 4. Results Figure 4.4: Decision route for significant changes regarding the transitional pro- vision under Article 120 of the MDR (Modified version from the source: MDCG 2020-3 Rev.1) [31]. 39 4. Results 40 5 Discussion During the data acquisition phase, two interviews were conducted with physiother- apists who have used Neuromotus in practice. The information accumulated over these interviews was pivotal for guiding the direction of the improvements. While the interviews provided cohesive feedback in some areas, there were some discrepancies in others. The limited availability of therapists with appropriate knowledge made finding suitable candidates and conducting the interviews a time-consuming process. Conducting additional interviews could have offered a more comprehensive under- standing of both therapists’ and patients’ opinions on the device. However, this would have resulted in a delay in the improvement process and consequently a reduced number of improvements. It therefore became a question of either imple- menting more improvements even though they might not be the most critical ones or fewer with higher certainty that these improvements are the most critical ones. Given these options, and considering that the interviews were conducted with highly experienced therapists, it was concluded that two interviews would provide sufficient information. Feedback from the HOPE representatives provided valuable insights into the games, indicating that Breakout was the most enjoyable and that the joystick controller was a beneficial change. However, despite their familiarity with the system, the HOPE representatives still found the Snake game difficult and too easy to run into oneself and lose. It was therefore decided that Integrum employees would not be required to test Snake unless they specifically wanted to. This led to only three out of thir- teen Integrum employees testing Snake during the second part of the stakeholder feedback. In response to feedback indicating that it was too easy for the snake to run into itself and lose, a small delay was implemented to prevent this behavior. This delay ensures the snake cannot make two consecutive turns too quickly, resolving the issue. How- ever, this modification also removes the option of making rapid turns. Given that the snake moves relatively slowly, it is believed that this change will not negatively impact performance while still removing the problem of the snake running into itself. The Integrum employees never tested the game 2048 because it requires 2 DOFs, and it was decided that keeping it simple with only 1 DOF would facilitate their learning and operation of the functions. This is unfortunate, as feedback on 2048 would have been valuable to understand its reception and identify any necessary future modifi- 41 5. Discussion cations. Nevertheless, the HOPE representatives provided positive feedback on 2048. Breakout was the most played game among Integrum employees, with twelve partic- ipants, closely followed by the Racing game, with eleven participants. Despite this, Breakout was significantly preferred, with 69.2% favoring it. This preference can be partly attributed to the new feature of playing with either a keyboard or joystick controller, with 100% of the users who tried both preferring the joystick controller. The implemented changes were evaluated through a series of validation tests and by gathering feedback from stakeholders who interacted with the modified system. The results from these validation tests indicated that all changes were successfully approved and passed without any detected bugs, confirming that the modifications functioned as intended. Among the various changes, the updates to the Breakout feature and the intro- duction of customizable limb skin tones in the AR environment were particularly well-received by the stakeholders. These modifications were praised for their poten- tial to enhance user experience and satisfaction. Despite these positive responses, determining whether these changes significantly improve user compliance and enhance the overall treatment of Neuromotus remains challenging. Comprehensive user testing is necessary to draw more definitive conclu- sions about the practical benefits and effectiveness of these changes in a real-world setting. Unfortunately, conducting user tests was beyond this project’s scope due to the ethical procedures required to obtain approval for user testing. The stakeholders only tried the new improvements with the upper limb textrode band, not the lower limb, due to practical reasons when placing the band. The games function the same regardless of whether the lower limb is used for myoelec- tric control. However, Breakout’s new feature of rotating the playing field 90 degrees allows for more intuitive control when using the lower limb with the up and down movements instead of left and right. In the AR environment, the same skin color customization is available for the lower limb as for the upper limb. Therefore, even though the stakeholders did not use the lower limb control, the potential for its use was evident through their experiences with the upper limb, given the similarities between the two. The racing game received the updates: the addition of a top 5 leaderboard displaying rank, time, and date, a speedometer, and a lap timer. These changes could enhance user compliance by making the game more competitive and engaging, provided there are no fundamental issues with the graphics or similar elements. Additionally, the integration of joystick functionality allows users to play more modern games that require a joystick, giving them the freedom to choose their preferred games. This increased personalization option is believed to boost engagement and result in an improved user experience. 42 5. Discussion Simultaneous movements are currently classified as a single discrete movement, and the system is trained to recognize that specific combination. This approach has limitations, as the two movements must be contracted with similar strength during both the calibration phase and real-time use to produce the correct combination of signals for the classifying algorithm. This makes using the joystick mode with simultaneous movements impractical, as the two movements would always have the same amplitude, which is not ideal for gaming scenarios. If the algorithm could detect and separate simultaneous movements, identifying their individual signal strengths, it would simplify the calibration phase by eliminating the need to calibrate each movement combination. Additionally, this would make the joystick mode more beneficial by allowing greater flexibility in the signal strength of combined movements. However, this approach would increase the computational complexity of the classification system and introduce greater latency. This tradeoff should be considered in the future development of the system. There is no universal scale for defining skin tone due to the high variation of peo- ple’s skin tones worldwide. To address this diversity, the Monk Skin Tone Scale, created by Ellis Monk in collaboration with Google, was adopted in this project. Originally designed for the computer vision field, this scale aims to capture demo- graphic representation with high precision while ensuring that the selection process remains user-friendly and straightforward. Its applicability extends to both VR and AR environments, where it can be used to model bodies or body parts accurately. From an ethical standpoint, the inclusion of a diverse range of skin tones in the AR limb model is crucial. It recognizes and respects the wide spectrum of human diversity, promoting inclusiveness and ensuring that users of various backgrounds feel represented. This inclusivity helps to avoid continuing biases and stereotypes that can arise from the underrepresentation or misrepresentation of different skin tones in digital environments. Moreover, incorporating the Monk Skin Tone Scale aligns with ethical principles of equality and fairness. It supports the creation of technology that serves all users equitably, which is particularly important in healthcare applications. By enabling users to choose skin tones that closely match their own, we enhance the psycholog- ical comfort and acceptance of the AR limb, which can positively impact the user experience and therapeutic outcomes. In terms of regulatory standards, the use of a comprehensive skin tone scale can be connected to the principles outlined in various guidelines and frameworks that advocate for diversity and inclusivity in technology. For instance, the World Health Organization (WHO) emphasizes the importance of cultural sensitivity and appro- priateness in health technology design. Similarly, the ISO 9241-171:2008 standard on ergonomics of human-system interaction, particularly in software accessibility, underscores the need for inclusive design that accommodates a wide range of human characteristics and preferences. 43 5. Discussion The result of question three in the form showed that 30.8% thought the skin tone of the limb looked okay, 30.8% thought it looked good, and 38.5% thought it was a perfect match. This indicates that people are generally satisfied with the ability to select a skin tone resembling their own. However, even with the introduction of new skin tones, there remains room for improvement in the realism of the AR limbs. Enhancing the realism of the arms would create a more convincing illusion of the AR limb being part of the user’s own body. Specifically, the shading and contrast of the limbs could be improved, particularly for darker skin tones. The contrast and shading of the skin appear less realistic for the darkest skin tones be- cause there is, for example, a greater difference in the nuances between the forearm skin tone and the hand palm for darker-skinned compared to lighter-skinned individ- uals. This variation was not fully considered during the development of the new skin tone textures and represents an area for future improvement in the AR environment. Improvement number five involves creating a final report that details the number of sessions, session lengths, utilized movements, and progress over time. Due to time constraints, this improvement was not implemented during this project. However, it represents a valuable future enhancement that could provide therapists and users with comprehensive information about their rehabilitation progress. Such a report could offer a better overview of their rehabilitation journey and serve as motivation to continue using the Neuromotus device, potentially improving adherence and out- comes. Modifying an MDD-certified medical device requires a thorough understanding of the regulatory framework governing such changes. According to Article 120 of the MDR, significant modifications to devices under the MDD are restricted. Therefore, it was crucial to navigate the project in compliance with these regulations. To ensure adherence, the project utilized the MDCG 2020-3 decision tree, which is designed to assess the significance of changes made to medical devices. This as- sessment confirmed that the changes implemented were classified as non-significant. As a result, the modifications did not require a new conformity assessment or re- certification under the MDR, thus maintaining regulatory compliance and ensuring that the device continued to meet all required standards. 44 6 Conclusion The implementation of a user-centered design approach was highly effective in iden- tifying and addressing the needs of therapists using the device. However, incorporat- ing feedback from amputated patients would have further enriched the development process and ensured that the device meets the needs of all end-users. Identifying a suitable open-source game for myoelectric control with high-quality graphics and single-player capability presented significant challenges. Even the games selected for the Neuromotus package required adjustments to ensure that the game was on an appropriate level. Feedback from users indicates that these improvements were beneficial and introduced valuable enhancements to the device, potentially leading to higher user compliance. The joystick controller for playing Breakout was particularly appreciated, with all users preferring it over the key- board controller. Nonetheless, definitive conclusions regarding the impact of these improvements on compliance and long-term adherence can not yet be drawn due to the limited scope of user testing. Furthermore, the ability to customize the skin tone of the limb in the AR environ- ment carries significant ethical considerations. By utilizing the Monk skin tone scale, the device promotes inclusivity and ensures that users from diverse backgrounds feel represented. This feature not only enhances user experience but also aligns with eth- ical standards in medical settings. Overall, while the advancements made through this project show promise, further research, and extensive user testing are required to validate the long-term benefits and ensure the device meets the needs of all potential users. 45 6. Conclusion 46 Bibliography [1] Kaas, Jon H. “ Somatosensory system.” The human nervous system (2004): 1059–1092. Elsevier New York, NY. [2] Ortiz-Catalan, Max. “ The stochastic entanglement and phantom motor ex- ecution hypotheses: a theoretical framework for the origin and treatment of phantom limb pain.” Frontiers in neurology 9 (2018): 748. Frontiers Media SA. [3] Foltán, René, Karel Klíma, Jana Špačková, and Jiří Šedỳ. “Mechanism of trau- matic neuroma development.” Medical Hypotheses 71, no. 4 (2008): 572–576. Elsevier. [4] Griffin, Sarah C., and Jack W. Tsao. “A mechanism-based classification of phan- tom limb pain.” Pain 155.11 (2014): 2236–2242. LWW. [5] Chiu, Isaac M. et al. “Bacteria activate sensory neurons that modulate pain and inflammation.” Nature 501.7465 (2013): 52–57. Nature Publishing Group UK London. [6] Weinstein, Sharon M. “Phantom limb pain and related disorders.” Neurologic clinics 16.4 (1998): 919–935. Elsevier. [7] Collins, Kassondra L., Hannah G. Russell, Patrick J. Schumacher, Katherine E. Robinson-Freeman, Ellen C. O’Conor, Kyla D. Gibney, Olivia Yambem, Robert W. Dykes, Robert S. Waters, Jack W. Tsao, et al. “A review of current theories and treatments for phantom limb pain.” The Journal of Clinical Investigation 128, no. 6 (2018): 2168–2176. Am Soc Clin Investig. [8] Finger, Stanley, and Meredith P. Hustwit. “Five early accounts of phantom limb in context: Paré, Descartes, Lemos, Bell, and Mitchell.” Neurosurgery 52, no. 3 (2003): 675–686. LWW. [9] Kalbaugh, Corey A., Paula D. Strassle, Nicole J. Paul, Katharine L. McGinigle, Melina R. Kibbe, and William A. Marston. “Trends in surgical indications for major lower limb amputation in the USA from 2000 to 2016.” European Journal of Vascular and Endovascular Surgery 60, no. 1 (2020): 88–96. Elsevier. [10] Laura Herrador Colmenero, Jose Manuel Perez Marmol, Celia Martí-García, María de los Ángeles Querol Zaldivar, Rosa María Tapia Haro, Adelaida María Castro Sánchez, and María Encarnación Aguilar-Ferrándiz, Effectiveness of mirror therapy, motor imagery, and virtual feedback on phantom limb pain following amputation: A systematic review, Prosthetics and orthotics interna- tional, vol. 42, no. 3, pp. 288–298, 2018, SAGE Publications Sage UK: London, England. [11] Katleho Limakatso, Victoria J Madden, Shamila Manie, and Romy Parker, The effectiveness of graded motor imagery for reducing phantom limb pain in 47 Bibliography amputees: a randomised controlled trial, Physiotherapy, vol. 109, pp. 65–74, 2020, Elsevier. [12] McDonald, Cody L., Sarah Westcott-McCoy, Marcia R. Weaver, Juanita Haagsma, and Deborah Kartin. “Global prevalence of traumatic non-fatal limb amputation.” Prosthetics and Orthotics International 45, no. 2 (2021): 105–114. LWW. [13] Horgan, Olga, and Malcolm MacLachlan. “Psychosocial adjustment to lower- limb amputation: a review.” Disability and Rehabilitation 26, no. 14-15 (2004): 837–850. Taylor & Francis. [14] Esquenazi, Alberto. “Amputation rehabilitation and prosthetic restoration. From surgery to community reintegration.” Disability and Rehabilitation 26, no. 14-15 (2004): 831–836. Taylor & Francis. [15] Lendaro, Eva et al. “ Seamless Integrated Textrode-Band for Real-time Lower Limb Movements Classification to Facilitate Self-Administrated Phantom Limb Pain Treatment.” In 2019 41st Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), pp. 1753–1756. IEEE, 2019. [16] NeuRehab@home - En smart textilplattform för neuromuskulär rehabilitering in hemmiljö. Högskolan Borås. Accessed: 2024-03-13. Available online: https: //www.hb.se/forskning/forskningsportal/projekt/neurehabhome/. Oc- tober 4, 2023. [17] MacIver, K., D.M. Lloyd, S. Kelly, N. Roberts, and T. Nurmikko. “Phantom limb pain, cortical reorganization and the therapeutic effect of mental imagery.” Brain 131, no. 8 (2008): 2181–2191. Oxford University Press. [18] Dehaene, Stanislas, and Laurent Cohen. “Cultural recycling of cortical maps.” Neuron 56, no. 2 (2007): 384–398. Elsevier. [19] Flor, Herta. “Phantom-limb pain: characteristics, causes, and treatment.” The Lancet Neurology 1, no. 3 (2002): 182–189. Elsevier. [20] Ortiz-Catalan, Max et al. “ Phantom motor execution facilitated by machine learning and augmented reality as treatment for phantom limb pain: a single group, clinical trial in patients with chronic intractable phantom limb pain.” The Lancet 388.10062 (2016): 2885–2894. Elsevier. [21] Flor, Herta, Thomas Elbert, Stefan Knecht, Christian Wienbruch, Christo Pan- tev, N. Birbaumers, Wolfgang Larbig, and Edward Taub. “Phantom-limb pain as a perceptual correlate of cortical reorganization following arm amputation.” Nature 375, no. 6531 (1995): 482–484. Nature Publishing Group UK London. [22] Katleho Limakatso, Lieselotte Corten, and Romy Parker, The effects of graded motor imagery and its components on phantom limb pain and disability in upper and lower limb amputees: a systematic review protocol, Systematic reviews, vol. 5, pp. 1–6, 2016, Springer. [23] Daniele Nico, Elena Daprati, François Rigal, Lawrence Parsons, and Angela Sirigu, Left and right hand recognition in upper limb amputees, Brain, vol. 127, no. 1, pp. 120–132, 2004, Oxford University Press. [24] Lawrence M Parsons and Peter T Fox, The neural basis of implicit movements used in recognising hand shape, Cognitive Neuropsychology, vol. 15, pp. 583–616, 1998, LAWRENCE ERLBAUM ASSOCIATES. 48 https://www.hb.se/forskning/forskningsportal/projekt/neurehabhome/ https://www.hb.se/forskning/forskningsportal/projekt/neurehabhome/ Bibliography [25] Max Ortiz-Catalan, Bo Håkansson, and Rickard Brånemark, An osseointegrated human-machine gateway for long-term sensory feedback and motor control of artificial limbs, Science translational medicine, vol. 6, no. 257, pp. 257re6– 257re6, 2014, American Association for the Advancement of Science. [26] Ortiz-Catalan, Max et al. “ Treatment of phantom limb pain (PLP) based on augmented reality and gaming controlled by myoelectric pattern recognition: a case study of a chronic PLP patient.” Frontiers in neuroscience (2014): 24. Frontiers. [27] Lendaro, Eva et al. “ Out of the clinic, into the home: the in-home use of phantom motor execution aided by machine learning and augmented reality for the treatment of phantom limb pain.” Journal of Pain Research (2020): 195–209. Taylor & Francis. [28] European Union. “Council Directive 93/42/EEC of 14 June 1993 concern- ing medical devices.” 1993. Available online: https://eur-lex.europa.eu/ legal-con