Applications of Augmented Reality in Neurology: Architectural Model and Guidelines

Neurology is considered as one of the most challenging fields of medical science due to the complexity of the nervous system in human body. The treatment processes of the patients suffering from neurological disorders are greatly influenced by the advancements in technology. In this regard, the infusion of augmented reality (AR) has immensely improved patients’ and healthcare workers’ experiences in the field of neurology. Several efforts are being made by the community to introduce various applications of AR for patients as well as for healthcare workers in neurology. This study systematically examines the applications of AR in neurology and synthesizes their uses and impacts in the treatment and assistance of the patients suffering from neurological disorders as well as in the assistance of healthcare workers. The results reveal that most of the AR applications are used for motor disorders while lesser systems are designed for patients with degenerative diseases, most of them being task assistance systems. Moreover, the use of AR applications leaves both long-term and short-term impacts in improving the quality of patient’s life. Furthermore, there is a need of introducing more interactive AR systems for patients with degenerative diseases as well as exploring the possibility of using such systems for mental rehabilitation. It has also been identified that there is a potential to develop AR applications for healthcare workers, to train and assist them through simulations. In addition to reviewing the AR applications, this work proposes a taxonomy of AR applications in neurology posing different dimensions including scope of AR, type of neurological disorder, and impact of AR assistance on patient’s life. Furthermore, this article also highlights challenges of using AR applications in neurology. Based on the identified challenges, guidelines to design AR systems are suggested for application designers and developers to help them design and develop more useful AR systems. Lastly, this work proposes an architectural model to design the AR systems for patients suffering specifically from degenerative or movement neuro-disorders in order to guide the future research of AR usage in neuroscience.


I. INTRODUCTION
Neurology is a term used in medical science to describe 24 the functionality of the central nervous system (CNS) and 25 brain [92]. Neurological disorders are caused by physical 26 impairment of the brain, nerve, or spinal cord [93]. Other 27 causes to have these disorders could be because of changes 28 in certain biochemical aspects. Even though the cause may 29 The associate editor coordinating the review of this manuscript and approving it for publication was Michele Nappi . be unidentified but the effects or symptoms on CNS can be 30 perceived [38]. Neurological disorders may cause symptoms 31 that are related to brain activity such as thinking and ana-32 lytics. These disorders may also cause problems in muscle 33 functionality and balance. Certain chronic conditions such 34 as hormonal imbalance may also result due to these dis-35 orders. However, temporary conditions induced due to the 36 environment or certain events may cause hurdles in brain 37 functionality but are not classified as neurological disorders. 38 These are psychiatric illnesses that are reversible, such as 39 The main goal of this research is to examine various appli-114 cations of AR in neurology and find the types of neurological 115 disorders being treated using AR. To the best of our knowl-116 edge, no study in past posed a review of AR applications' 117 usage in neurology. Based on the findings of this review, 118 this study suggests guidelines for designing AR systems to 119 enhance the user experience. Moreover, a taxonomy for the 120 use of AR technology in the field of neurology has also been 121 presented in this article. Furthermore, this work proposes a 122 dedicated AR apps structure to overcome the challenges faced 123 by patients with recognized neurological disorders. A user-124 friendly design model and layered architecture for neurolog-125 ical disorder apps based on the identified gaps has also been 126 presented in this study. 127 The rest of this article has been organized into the fol-128 lowing sections. Section II describes the AR systems that 129 are being used in the field of neurology. Section III defines 130 the research methodology along with data analysis exhibit-131 ing the responses to the research questions. Section IV 132 describes the discussion and implications, which poses the 133 taxonomy, and gaps and challenges. The sub-system division 134 of physical therapy apps and an AR system architecture 135 for neurological disorders are also proposed in this section. 136 Lastly, the Section V provides the conclusion by highlighting 137 the major findings of this work. 139 This section presents the work done in neurology using AR 140 applications specifically related to the patient's treatment and 141 assistance journey. AR applications may also be used for 142 healthcare worker's training or facilitations who are treating 143 patients with neurological disorders. The use of AR, as rec-144 ommended in [83] and [84] is a technique to improve motion 145 recovery through action observation. The majority of the AR 146 systems proposed in the literature for rehabilitation activities 147 require a person to wear external devices or sensors. 148 Rehabilitation for stroke patients involves regaining 149 motion in the afflicted body parts, primarily the limbs, 150 VOLUME 10, 2022 through intense, skilled training exercises. Muscular tasks 151 such as grasping, stretching, and holding objects of every-152 day usage (such as coffee making, cellphones, TV remotes, 153 etc.) are used in various rehabilitation procedures [9]. 154 Task-specific rehabilitation, like treadmill exercise with frac-155 tional body weight support and hip and knee extension 156 workouts, is all part of lower limb rehabilitation [24]. The rehabilitation methods because it improves user engagement 207 and exercises performance outcomes. But the target popula-208 tion of these systems is only the patients after stroke and mas-209 tectomy for their arm movement exercises. However, due to 210 the complexity and high cost of the AR Sleeve system, it can 211 be used in rehabilitation gyms with multiple and concurrent 212 therapeutic sessions. 213 Dementia is an incurable neurological illness that causes 214 memory loss, anxiety, agitation, depression, and other neu-215 ropsychiatric symptoms, as well as a detrimental influence on 216 the cognitive, emotional, and physical capacities of those who 217 suffer from it [85]. Daily activities are tough for caregivers 218 and people with Alzheimer's disease. AR approaches have 219 been used to generate important studies. Quintana and Favela 220 [25] used AR markers to insert reminders for completing a 221 task [30]. Using HoloLens, an AR system was created to 222 evaluate the beginning of Alzheimer's disease at an early 223 stage. The symptoms that appear to be linked to early memory 224 loss, particularly spatial memory, are the most critical features 225 of early AD diagnoses. A game environment incorporates the 226 capacity to store and recover the memory of a specific event 227 involving an association between things such as the location 228 and the object attributes [32]. However, such apps challenge 229 the spatial intelligence of the patients with already impaired 230 brains. Moreover, as such patients have compromised mem-231 ory, they may not recall the steps and process to move forward 232 in the application.

233
According to the researchers, a variety of physical activ-234 ity therapy, such as aerobic exercise, resistance training, 235 yoga, and dancing, may help patients with Parkinson's dis-236 ease (PD). Dance's pleasurable aspect also elicits positive 237 emotional responses, gives a means of communication and 238 expression, and encourages frequent attendance at dance 239 courses. MTG (Moving Through Glass) is a Google Glass 240 application that acts as a portable, continuous extension of 241 the device. It projects MTG films of various activities into the 242 wearer's natural environment [19]. People with Parkinson's 243 illness may benefit from AR-based therapies like MTG 244 because they make dancing more accessible and adherent. 245 Although these researchers digitalized the MTG and tested 246 it, but the patient's sample size is comparatively small and no 247 significant improvement has been observed as compared to 248 the traditional therapy. 249 This research has been conducted to provide the synthe-250 sis of AR applications in the field of neurology. It poses 251 a solution to the challenges faced by the patients suffering 252 from neurological disorders. The novelty of this study is 253 that it provides a guideline to design and develop appli-254 cations used in neurology based on the challenges and 255 gaps identified from the previous work. A taxonomy has 256 also been proposed to present the classification of various 257 aspects of AR applications in neurology. Moreover, this work 258 presents a sub-system division of the physical therapy based 259 applications and provides a layered architecture especially 260 designed for neurological disorders considering the identified 261 challenges.  The fundamental part of an SLR is the formulation of a search 303 plan to effectively determine and gather relevant articles in the 304 selected field. This process involves the narrative of a search 305 string, all the literature resources that are used to apply the 306 search, and the segregation (inclusion/exclusion) plan to get 307 the most related papers out of the identified records. To collect relevant research, and work it is important to search 310 the right way. To find related work, the searching methodol-311 ogy has been divided into three parts: disorder search, string 312 search with disorder keyword, and general string search with-313 out the specific disorder. The first part includes finding the 314 most common neurological disorders with either a diagnosis 315 or treatment plans. This includes searching for the recognized 316 medical disorder that is common in all parts of the world. This 317 search has been done on search engines and online medical 318 databases and websites like WebMD. This provided a list of 319 neurological disorders that are used in the second part using 320 a citation search engine to find previous literature ranging 321 from January 2009 (initial phases of AR/VR technology) to 322 April 2022. 323 The journals have been picked from IEEE Xplore, Google 324 Scholar, Science Direct, ACM, PubMed, Springer, and sim-325 ilar. The search string was built in a similar format that 326 works on a typical search bar where a neurological disorder 327 was placed with Augmented Reality in the double quotation. 328 For some search strings with a disorder that has no cure 329 therapy keywords such as ''Activity'', ''rehabilitation'', and 330 ''therapy'' were also searched. In the third part, a standalone 331 VOLUME 10, 2022

388
EC 2) Paper in which augmented reality and recognized 389 neurological disorder are not mentioned explicitly.

390
EC 3) Paper that is not written in English.

391
EC 4) Paper that is re-published.

393
This SLR has been conducted based on the papers published 394 between January 2009 and April 2022. Numerous articles 395 were gathered during the primary search but some of them 396 were related to the study and some were not. The papers were 397 then filtered following the process shown in figure 3 [58].

398
Initially, the papers were filtered out based on the titles 399 and abstracts. The abstracts have been analyzed carefully to 400 determine relevance to the research questions and the main 401 objective of this research.

402
The literature screening was a two-step process. In the first 403 step, all the literature has been scanned to see if keywords 404 such as neurological disorder and AR are present in the title. 405 The journal/conference in which papers were published to 406 ensure the literature title matches context-wise. Out of around 407 135 selected papers, the first step of screening resulted in 408 around 47 papers. After the initial round, the abstract of the 409 paper has been carefully read and detailed the paper using the 410 technique defined by Keshav [4].

411
In figure 4, the PRISMA technique used for literature 412 searching and selection is presented, together with the num-413 ber of articles found at each stage. The screening has been 414 done twice, in the first cycle, 82 related papers were gath-415 ered that were shortlisted to 21 papers after passing through 416 the inclusion and exclusion criteria. In the second round, 417 29 papers were shortlisted from a total of 108 articles. So, this 418 study provides the review of 47 papers in total. As the search 419 results were based on keywords that may have alternate 420 meanings in different scenarios; hence, a careful screening 421 of papers was carried to shortlist the finalized set of papers 422 for review. To establish a ranking system, we look at the different goals 425 each paper presented. For that purpose, point system has been 426 used, which is a known metric for ranking literature. Three 427 basic ideas have been developed and if the ideas have been 428 accomplished in the said literature, then a (+1) point was 429 awarded to the paper. Moreover, if the paper was published 430 in a CORE A or greater conference journal with a significant 431 impact factor then it would be awarded an additional (+1). 432 Here are the general ideas 1) Does the study claim to present a novel idea of using AR 435 for treating or diagnosing neurological disorders? If so, 436 the paper is given a +1 Rank.

437
2) Does the paper pose significant results in achieving the 438 goals set at the start of the research? If so, the paper is 439 given a +1 rank.   The articles were assessed and scored using the previously 474 stated scoring criteria (internal and external). Figure 5 shows 475 a visual depiction of the rankings of the articles based on 476 internal and external scores.

477
Each question to assess the internal criteria has been 478 assigned 1 score. And question for the external criteria has 479 been assigned 2 score, so the score has been calculated out of 480 the 5. Papers having a total score of more than 2.5 have been 481 ranked as high level, less than 2.5 have been ranked as low 482 level, and 2.5 score is considered average ranking.

483
Only 14.8 % are scored as low, and 12.7% as average, 484 according to the results of the whole review. The remaining 485 72.3% are rated higher than average.  AR technology has introduced a newer and interactive dimen-496 sion to the treatment and therapy process by introducing 497 senses that are often difficult to simulate. This allows the 498 simulation of the brain and nervous system, which may not be 499 possible with medication or traditional physical therapy. This 500 also allows users to experience situations virtually, which 501   mounted device for movements such as walking in research 518 done in [24]. This allowed the researcher to identify different 519 gait positions than expected, which may result in degradation 520 of gait movement treatment overall for users. In some cases 521 of MS patients, physical therapy might also be needed to 522 slow down the effect of the disease. The application has been 523 designed to treat balance and rehabilitate MS patients who 524 experience balance issues using a home-based VR/AR system 525 and infused daily tasks such as multi-tasking and obstacle 526 negotiation [22].

527
Mental health therapy such as Neurofeedback Therapy 528 for the patients suffering from ADHD has shown a sig-529 nificant improvement in the treatment's efficacy. This dis-530 ease usually effects the younger individuals and cause them 531 to make impulsive decisions or decrease their success rate 532 in all the aspects of life. The usual treatment process of 533 this disease is deadly that causes patients to lose motiva-534 tion but with the help of AR technology, the whole process 535 has been made interesting by designing interactive graphical 536 VOLUME 10, 2022  been observed [26].

548
AR systems help patients to perform their everyday tasks 549 to improve their living [65]. In case of diseases like dementia, 550 where patient's memory is worsen day-by-day, AR applica-551 tions assist them to perform an activity/task like making tea 552 [25], or to recognize objects, for example, AR systems have 553 been observed where intellectually impaired individuals get 554 an alert signal to stay away from hot objects [22].

562
The researchers developed an application [16] where AR 563 was used to transform medical records into 3D organs, which 564 gave an in-depth idea of medical history to patients allowing 565 them to visualize their organs.  observed in which the AR applications have been provided 577 as a treatment alternative to traditional medicine processes. 578 AR applications are either mostly used for treating move-579 ment disorders where patients are struggling with limbic 580 movement, balance, or coordination of hand-eye movements. 581 Another pattern is observed where AR application is being 582 used to maintain the capabilities of the brain to think and 583 memorize, especially in parts of the cerebrum. Disorder 584 or degenerative diseases where a person loses capability 585 and gets slower while using interactive games/activities 586 using AR.

587
AR has been used for the treatment of motor diseases such 588 as CP using AR applications that help do motion coupled 589 with a physical device such as a treadmill [15]. Level 3 590 patients of GMFCS (Gross Motor Function Classification 591 System) have been rehabbed with AR-mounted headsets for 592 movement disorders [24]. Research in [17] developed an 593 AR application for hand movements such as finger tapping 594 gestures with healthy people as a control group and patients 595 with PD and cerebrovascular accident (CVA). MS and PD 596 already treated with AR applications has been observed such 597 as in [29]. Similarly, in [18] AR being used to treat patients 598 who have suffered paralysis due to stroke and PD by asking 599 them to perform hand movements. Similarly, PD patients 600 were provided with paths such as circular paths using AR [31] 601 to improve the motor function of the nervous system. Around 602 20 patients with idiopathic PD were part of research that 603 helped do physical activity or 'class' where participants were 604 required to do movement such as boxing/swimming using 605 Google glass in research conducted in [19]. More movement 606 exercises were done in [27] for patients with stroke disabil-607 ities. While some applications focused on only movement 608 disorders some applications also focused on the disorder that 609 affected brain functionality. In [14] research was done where 610 FIGURE 11. Statistics of the neurological diseases treated by AR.
[30] provided a patient and caregiver with an AR-based 637 application to keep track of progress and perform the daily 638 task using AR assistance coupled with smartphone technol-639 ogy. It was tested with patients with Alzheimer's. 640 Figure 10 shows the statistics of the types of neurological 641 disorders that are mostly being catered by AR technology. 642 By carefully evaluating the research papers, it has been 643 observed that AR technology has been used for degenerative 644 disorders more than motor disorders. 645 Figure 11 presents the frequency of diseases that are being 646 treated by AR technology. It has been observed after evalu-647 ating the selected papers that mostly AR systems have been 648 used for the rehabilitation of the stroke patients.

649
In all those disorders, AR has been infused in various 650 capacities. From the results, it is clear that AR systems are 651 very helpful for stroke patients due to the of use physical 652 health therapy for their treatment. Although AR has enhanced 653 the quality of life of the patients suffering from neurological 654 disorders, yet there is a need to increase the use of AR in for 655 the diseases like pituitary tumors and ADHD. AR applications have a wide variety of use cases and most 660 of the applications focus on rehab of the users. As shown 661 in RQ2 the two different targets are motor movements and 662 mental ability that AR application tries to improve on. 663 Different research is done to examine the effectiveness of 664 AR applications. A study [28] presented the use of social 665 robots and AR to assist the display for patients with dementia. 666 The research showed an increase in contextual interactions 667 and usage. Similarly, a work [9] showed AR-based patient 668 support system called 'cARe' was introduced that helped 669   This allows healthcare workers to maintain a low malprac-707 tice rate and improve patient care. In [23] better surgical 708 process has been provided to reduce stress for patient and 709 surgeon while performing neurosurgery by allowing better 710 visualization. This greatly improves the surgery success rate 711 and decreases complications. A research [36] provides smart 712 visualization for tumors in the brain using smart that gave 713 better accuracy than medical 2D imaging disorders, which 714 often need AR application assistance. This has also been 715 experienced by [20] where the 2D model was considered 716 better by participants than the 3D model.

718
A taxonomy has been proposed in figure 13 after carefully 719 evaluating the collected papers, and concluded that AR sys-720 tems in neurology are based on three main aspects. The first 721 aspect is the scope of an AR system that describes a purpose 722 and the user system is intended. Two users are predominantly 723 using AR systems in neurology. The first category of the user 724 is patients or people suffering from a neurological disorder. 725 To support their treatment, there are two types of activities 726 that AR application performs. The first one is rehabilitation 727 that includes all types of therapies of the patients. Therapies 728 may be physical or mental. These applications try to directly 729 impact the users' progress during the patient's life cycle and 730 mitigate the effects of disease. Physical therapy refers to the 731 movement and exercises that a patient does with the help of 732 AR system to improve his deterioration of disease or gain 733 back the abilities of their muscles, strength, balance, and 734 movement. This is the most common type of AR system 735 found in neurology where interactive exergames are used to 736 treat patients. The second type of therapy is mental therapy, 737 which targets the user's cognitive ability to improve his think-738 ing, understanding, and reflexes.

739
This may include memory exercises, attention exercises, 740 and cognitive ability testing. These types of AR systems are 741 rare in neurology as observed that such patients often have 742 limited cognitive ability to understand the setup and operate 743 AR systems.

744
While many systems are trying to rehabilitate users. There 745 is also a category of task assistance systems designed for 746 patients to perform daily activities that are hampered by their 747 disease. Those activities may not directly relate to symptoms 748 caused by the disorder but have an influence on the patient's 749 quality of life.

750
Some AR systems help users to perform menial tasks 751 such as making food, taking medicine, and assisting them 752 in performing day-to-day activities while some systems help 753 users to navigate social conditions such as cautioning them 754 of dangerous objects in the environment (e.g. Alzheimer's 755 patients being cautioned to stay away from the oven or hot 756 objects).

757
The other type of users of an AR system in neurology 758 is healthcare workers such as doctors, paramedical staff, 759 surgeons, and personal healthcare providers. These systems 760 help doctors in the field of neurology and improve their 761 ability to diagnose and treat patients. This category includes 762 general-purpose systems that help healthcare workers from 763 training to diagnose disorders using AR-based 3D models 764 and help them visualize a patient's medical report in a more 765 interactive environment. There are also surgery simulation 766 eries before performing them on real patients. There are also a 768 few AR systems that help healthcare workers through passive 769 medical record gathering and helping them visualize patients' 770 progress.

771
The second aspect is the classification of neurological 772 disorders, which are being treated using the AR system. First  patient's system. This includes muscle activities/exergames 800 that last for weeks up to months. This is the most com-801 mon type of goal setting in the AR system observed in our 802 research.  Table 5 summarizes the AR-based system for neurological 806 disorders. These systems contain different types of meaning-807 ful plays to rehab the patients, such as apps to help in the ther-808 apy or exercise and also for assisting the patients/healthcare 809 workers. As AR is a field that augments the real-world expe-810 rience accessible to all, the implementation also requires AR 811 to be universally available, which means AR must not require 812 specific hardware to run. Whatever device runs, the AR must 813 be powerful enough to render 3D models and have enough 814 sensors to be interactive. This means that today, common 815 smartphones, tablets, and handheld devices can run AR appli-816 cations [39]. The most common devices used in AR sys-817 tems include handheld devices such as smartphones/tablets, 818 smart glasses [33], HMD [13], and desktop PCs as shown 819 in figure 14. 820 VOLUME 10, 2022   So, it will depend on the requirement and usage of the system 857 that which technology will work better. With excessive use of 3D some patients with neurological 870 disorders often do more spatial computing while using AR 871 applications. This may create more cognitive load on the 872 already stressed brain and nervous system. Research in [10] 873 shows that AR HoloLens for Alzheimer's patients may often 874 create a cognitive barrier for the user. Moreover, according to 875 [6] a larger device size is needed for spatial representation, 876 which may not always be available. This may cause the user 877 to misperceive the space around them causing harm. Most applications of AR require hardware to run, which is 880 a standard problem for the audience, which contains aged 881 or impaired people. However, this is a huge challenge for 882 people with neurological disorders. If an AR application 883 requires some time for the user to learn and requires the user 884 to remember it may be a problem as this target group has 885 difficulty maintaining information. People with Alzheimer's 886 disease/dementia cannot be educated over and over on how 887 to execute said goal using AR. Research like [11] shows that 888 often AR equipment needs a technician on hand and normal 889 people without any disability need constant training on how 890 to use applications.

892
Applications of AR that rely on movement may have a higher 893 error percentage due to omission errors caused by missed 894 objects/vectors. This was experienced by LeapMotion in [17] 895 where 5% of motions were incorrectly recognized due to 896 fingers being omitted from the line of sight. This also was 897 detailed by [7] where one size of AR hardware may cause an 898 error for patients with different physical attributes.

900
Based on the challenges identified in the section above, 901 a design requirement guidelines have been proposed as 902 original work in this research. The goal of this design 903 requirement is to overcome such challenges and provide 904 designer/developer guidelines to make AR applications and 905 systems that are more suited to people suffering from neu-906 rological disorders. Based on all three challenges identified, 907 a design idea has been proposed to overcome that challenge 908 as well hardware/software improvement that may lead to 909 overcoming that challenge. The most challenging thing about cognitive overload is that 912 it requires a reduction in the system in terms of modeling, 913 it also requires the system to have less text and less spatial 914 input. This means that system needs to tune to lower quality, 915 VOLUME 10, 2022 improvement. It is identified that often the application that targets patients with neurological disorders does not consider 918 cognitive load as the main criteria. After considering each 919 neurological disorder, a matrix has been provided where cer-920 tain input and output should be avoided.  this has been observed that lack of spatial intelligence is a big 942 concern, which may also require a larger input/output device  3D modeling in output is also preferred as patients have 972 better spatial intelligence than Neuro-Degenerative patients. 973 Moreover, the output may require future input as the next 974 steps, which means that real-life-based 3D modeling with 975 accurate sizes is a must for pin-point movement training. 976 Table 6 enlists the input and output preferences of AR 977 applications for patients suffering from different neurological 978 disorders according to the identified classification of neuro-979 logical disorders. The literacy problem was observed when AR applications 982 have been tested with patients who suffer from long-term 983 degenerative disorders. This was due to fact that a young 984 person is less likely to develop a neurological disorder. Most 985 patients are in a more advanced age bracket. This means 986 that not only their memory may be compromised due to 987 disorder but generally their ability to memorize, remember 988 and navigate digital systems may already be very low.

989
This means the AR system treating such patients needs to 990 be more in line with being elder-friendly technology, which 991 most systems lacked. In this research, design rules have been 992 provided as guides to designers along with the reasoning that 993 can be incorporated to reduce the literacy gap for the AR 994 system. The system must be easy to start and should take less input 998 from user to reach main state of application. This will also 999 smooth user experience for patient to start application ever 1000 time and will counter difficulty patients face during start-1001 ing the application such as remembering steps or navigating 1002 through screens. The System should auto-sense for getting started and should 1006 be mostly audio based with the ability to accept multiple 1007 words/phrases for the same step. some problems. It is observed some AR applications that use 1059 both a 2D fiducial tracker as well as two calibrated fiducials 1060 for 3D models such as in [2]. This can improve tracking and 1061 identification. Other algorithms such as the head-leveling algorithm and 1080 the optical tracking system may also be used to reduce the 1081 overall errors in movement.

1082
A model on the basis of the presented design rules has been 1083 provided as shown in figure 16. This model is a guide for 1084 the developers to design a better and more user-friendly AR 1085 application for patients with neurological disorders.

1086
By using chronological order in which normal applications 1087 are executed, presented model also runs in similar order with 1088 additional steps such as device setup and launching guide-1089 lines as well as correction and session saving layers. This 1090 allows the model to behave like a traditional AR system with 1091 improvements to facilitate the patients suffering from neuro-1092 logical disorders resulting in enhancing the patient experience 1093 significantly.  It has been identified that the countries with growing 1101 percentage of population over 65 years of age have more 1102 proportion of patients suffering from neurological disorders 1103 [93]. This inherently causes the technology literacy issues, 1104 as learning to use AR applications could be difficult for 1105 the specified age group. The proposed sub-system enhances 1106 the user experience and reduces this technology literacy gap 1107 through automation.

1108
It consists of the patient and therapist's subsystem. Each 1109 exercise that will be designed by the therapist for each reg-1110 istered patient will be rendered by this subsystem. The user 1111 VOLUME 10, 2022  To facilitate developers to make successful and result-1135 yielding AR applications in the field of neurology, an archi-1136 tectural model has been proposed that is shown in figure 18 1137 based on the gaps identified earlier. The input/output (I/O) 1138 processes in the architecture system of AR, presented in the 1139 previous studies, were general for all fields [101], [102] but in 1140 our model, we further categorized the interaction and render-1141 ing layers for different types of neurological disorders based 1142 on the reasoning defined in Table 6. Moreover, the model has 1143 start-up rules in the processing layer to minimize technology 1144 literacy and cognitive overloading. As this model is supposed 1145 to provide precise feedback to the therapist, so this model 1146 makes sure that the movement done by the patient is accu-1147 rately tracked. For this purpose, we have added a correction 1148 sensors including monochrome or RGB cameras. Specific The physical layer also houses the dedicated processing unit 1202 such as CPU, GPU, and VPUs, which are the brains of the 1203 system. They contain the computation power for the AR 1204 system and processing for input and output. This layer is the main layer for processing and provides step-1207 by-step information of the actual flow of application. This 1208 layer uses the guideline on how all AR applications targeting 1209 neurological need to define their flow.

1210
The first step is to start the hardware. This includes the 1211 guidelines that have been described in the previous section. 1212 The hardware is easy to setup and it should be lightweight. 1213 Once the hardware setup is done and the application is 1214 launched, the application needs to interact with the ses-1215 sion dataset to restore the previous successful state. Then, 1216 the application must transfer into a data-gathering state, 1217 which will use interaction interfaces from the physical layer 1218 to gather data from the user. All the localization, scene 1219 analysis and mapping is handled by vision engine. Vision 1220 engine is the computer vision code that detects and interprets 1221 frames capture by AR devices. It is part of all AR devices 1222 as main component that helps detect objects and interpret 1223 environment.

1224
Enormous input and output data need to be stored and 1225 maintained, as the system needs to store processing data 1226 as well as the state of the system. Hence, the section of 1227 data collection has been introduced under the abstraction 1228 set. Although this has not been described, which specific 1229 hardware can be used as AR systems vary from head-mounted 1230 displays to mobile devices? What this section describes is 1231 a different type of data sets. The data/frames captured by 1232 the AR system for environment mapping are stored in the 1233 world knowledge dataset including data on the plane, lights, 1234 and occlusion. For an AR system to render different output 1235 there also needs to contain data storage that houses interactive 1236 content data. This includes variables that would be required 1237 to output processing data including content, superimposition 1238 data, and different layer fusion data.

1239
While another dataset may be part of any generic AR sys-1240 tem, what we have introduced in the data collection layer is 1241 the session management dataset. This dataset is critical to any 1242 application that treats neurological disorders as it maintains 1243 the application's state for the user so when the user relaunches 1244 the application the previous successful state is automatically 1245 restored keeping the application complexity and specifi-1246 cally start-up complexity very low. This also reiterates the 1247 design guidelines that application support auto-start and 1248 restore the last checkpoint to reduce huddle for technology 1249 illiterates.

1251
These include sub-systems that is interacting with users 1252 directly. This may include systems that take data from the 1253 VOLUME 10, 2022 are data collection applications that gather data from the 1279 patient-side application and provide it to the therapist to ana-1280 lyze the patient progress. They may be tracking application 1281 that gathers movement data or monitoring application that 1282 helps track patients' vitals.

1284
This paper presents a systematic literature review that com-1285 prehensively investigates the applications of AR techniques 1286 in neurology. A total of 47 papers were finalized from a 1287 subset of 135 papers that met a criterion to shortlist the 1288 relevant articles. This SLR identifies AR applications, the 1289 types of disorders being treated in the field of neurology 1290 through AR applications, and types of improvements along 1291 with limitations these AR applications pose. A point-based 1292 evaluation was done for all the collected papers.

1293
A taxonomy depicting different aspects of AR applications 1294 in neurology has also been presented. The identified limita-1295 tions in studies are highlighted, which are found while using 1296 AR for the treatment of neurological disorders. The technol-1297 ogy literacy, cognitive overloading, and movement errors are 1298 identified as some of the challenges of using the existing AR 1299 systems. To address these challenges, a model exhibiting a 1300 potentially robust AR system has been proposed. It is based 1301 on the type of disease, preferred interaction mechanism, and 1302 different computation techniques to improve results. The 1303 model emphasizes the specific input and output types as per 1304 the understandability of the patients and the different types of 1305 disorders. The general architecture has also been suggested 1306 clinical training using augmented reality,'' Virtual Reality, vol. 25