Exploring user perspectives on a robotic arm with brain–machine interface: A qualitative focus group study

Abstract

Brain–machine Interface (BMI) is a system that translates neuronal data into an output variable to control external devices such as a robotic arm. A robotic arm can be used as an assistive living device for individuals with tetraplegia. To reflect users’ needs in the development process of the BMI robotic arm, our team followed an interactive approach to system development, human-centered design, and Human Activity Assistive Technology model. This study aims to explore the perspectives of people with tetraplegia about activities they want to participate in, their opinions, and the usability of the BMI robotic arm. Eight people with tetraplegia participated in a focus group interview in a semistructured interview format. A general inductive analysis method was used to analyze the qualitative data. The 3 overarching themes that emerged from this analysis were: 1) activities, 2) acceptance, and 3) usability. Activities that the users wanted to do using the robotic arm were categorized into the following 5 activity domains: activities of daily living (ADL), instrumental ADL, health management, education, and leisure. Participants provided their opinions on the needs and acceptance of the BMI technology. Participants answered usability and expected standards of the BMI robotic arm within 7 categories such as accuracy, setup, cost, etc. Participants with tetraplegia have a strong interest in the robotic arm and BMI technology to restore their mobility and independence. Creating BMI features appropriate to users’ needs, such as safety and high accuracy, will be the key to acceptance. These findings from the perspectives of potential users should be taken into account when developing the BMI robotic arm.

1. Introduction

Severe damage to motor functions due to spinal cord injury, muscular dystrophy, stroke, amyotrophic lateral sclerosis, and other disorders can lead to paralysis and eliminate the ability to perform activities of daily living. Assistive technologies, such as a robotic arm, enable people with tetraplegia to restore their mobility and independence using various interface systems.^[1] The brain–machine interface (BMI) can be one of the ways to control assistive devices by connecting their intact neuromuscular pathways of the brain. BMI refers to a system that deduces the user’s intention from nerve data acquired from the brain and converts it into an output variable to control a computer, peripherals, or body aids according to the user’s requirement in real-time.^[2,3] Our research team aims to develop a BMI robotic arm for people with tetraplegia with a human-centered design (HCD).

The International Organization for Standardization (ISO) standardized HCD through ISO 9421-210:2019, as an approach to interactive systems development.^[4] To make it usable and useful, the key is to focus on the users, their needs, and requirements, and to apply human factors/ergonomics, and usability knowledge and techniques. The 6 principles of proper HCD are as follows: (a) the design is based upon an explicit understanding of users, tasks, and environments; (b) users are involved throughout design and development; (c) the design is driven and refined by user-centered evaluation; (d) the process is iterative; (e) the design addresses the whole user experience; and (f) the design team includes multidisciplinary skills and perspectives. The iterative process of HCD assistive technology development of the following 4 stages: (1) understanding and specifying the context of use; (2) specifying the user requirements; (3) producing design solutions; and (4) evaluating the design. In the present paper, our multidisciplinary design team focuses on stages 1 and 2, which are, identifying the context of use and requirements from potential users in the process of developing the robotic arm using a qualitative approach, the focus group interview (FGI).

For a better theoretical understanding of how assistive technology works in a context, this study followed a Human Activity Assistive Technology (HAAT) model that illustrates assistive technology usability with the interaction of (1) a human with (2) assistive technology to engage in (3) an activity in (4) a given context.^[5] The activity is the essential element of the HAAT model because it can define the overall purpose of the assistive technology system, that is, the process of doing something and the functional result of human performance. We hypothesized that the focus group interview with potential users would generate ideas of what activities users will be able to or expect to participate in using the BMI robotic arm, which can imply what elements and standards the design and development team should reflect in the device.

This study aims to explore, the activities in which people want to participate by using the BMI, their opinions, requirements, and usability of the BMI robotic arm from the perspectives of people with tetraplegia. The findings from this study will be used in the future BMI robotic arm development process to make it acceptable and applicable for potential users, such as people with spinal cord injury and muscular dystrophy.

2. Methods

2.1. Participants

Eight people with tetraplegia [four with spinal cord injury (SCI) and 4 with muscular dystrophy (MD)] participated in the focus group interview. All of them are electronically powered wheelchair users. The interview sample was purposive to focus on the target group of potential users. Potential users of the BMI robotic arm with severe impairment in motor and sensory systems were considered because SCI is an acquired disorder and MD is a congenital disorder. All the participants with SCI had severe motor dysfunction with a diagnosis above the C6 level. The average onset duration of the participants with MD is 21.25 years. Characteristics of interviewees are summarized in Table 1. They were individually informed about the study and gave their consent to participate. This study was approved by the Seoul National University Hospital Institutional Review Board (IRB No. 1403-056-565).

Table 1.

Focus group participants.

User ID number	Age (years)	Gender	Disability type	Period (years)	Housing type
U1	45	Male	Myopathy	25	Apartment
U2	42	Female	Myopathy	21	Detached house
U3	51	Male	Myopathy	15	Apartment
U4	56	Male	Myopathy	24	Apartment
U5	34	Male	SCI C5/C5 ASIA A	12	Apartment
U6	41	Male	SCI C6/C6 ASIA A	19	Apartment
U7	46	Male	SCI C5/C5 ASIA A	23	Apartment
U8	21	Male	SCI C6/C6 ASIA A	7	Studio

ASIA = American Spinal Injury Association, ID = Identification, SCI = Spinal cord injury, U = User.

2.2. Procedure

In this study, FGI was used to identify the BMI robot arm’s applications and the potential users’ perception of it. FGI is a qualitative research method that enables us to directly review individuals’ perceptions and experiences of specific topics and problems, and it is designed to obtain the participants’ different opinions and feedback on a particular topic.^[6] Deeper and richer information can be obtained on the selected topic in a relatively short period of time rather than through one-on-one interviews.^[7] The semistructured interview format we used can encourage interactive communication between the interviewer(s) and interviewee(s) by allowing both groups to ask questions to each other.

To stick with a principle of proper HCD, our design and development team consists of 7 experts from multidisciplinary fields with experience times ranging from 6 to 26 years, including rehabilitation medicine, neurosurgery, occupational therapy, and biomedical engineering. Our team, with diverse perspectives, generated 6 questions applied to FGI. Table 2 shows a list of the questions.

Table 2.

List of questions for focus group interview.

1.How do you think about brain-machine interface technology?
2.What would you like to use this technology for?
3.What are your thoughts on the invasive methods that require surgery?
4.What do you think of the noninvasive way of not requiring surgery?
5.What areas of rehabilitation assistive technology are the most useful for you?
6.What levels of the brain-machine interface are acceptable regarding the following characteristics? [Motion speed, accuracy, set up time, training period for operation, safety, size and portability, cost.]

Each participant has received a 5-day advance notice to guide the users through the process, including contents and questions. In order to make the interview effective, FGI was undertaken by a well-trained and experienced moderator with expertise in occupational therapy and assistive technology usability for more than 7 years. Eight people in the user group and 7 people in the expert group participated in the interview. In the beginning, participants had a short introductory session with videos regarding a case of using the robotic arm with BMI technology published online for approximately 15 minutes. They had a question-and-answer session with the expert group for about 30 minutes. Then, the moderator led a discussion session based on prepared questions for approximately an hour. All sessions were recorded by 2 video recorders and an audio recorder, and a stenographer transposed the conversation in real-time into a word document. Each of the FGI members had a microphone in front of them to help them sound comfortable and effortless as they spoke so that they could respond independently to the discussions without any communication assisting devices or help from their caregiver.

2.3. Data analysis

In this study, a general inductive analysis method was used. This method has been used to analyze qualitative data and entails condensing raw textual data obtained through interviews into summaries. A coding scheme was generated based on the research purposes. The final version of the transposed verbatim was explored by 2 researchers after a thorough examination using basic content analysis and theme identification to find clear associations with results and research objectives.^[8,9] Our development team discussed which narratives should be presented in the paper to reflect user perspectives. Three themes were extracted, including activity, needs, acceptance of the BMI, and usability. Below is the detailed description of each theme.

2.3.1. Activity.

Activities from the coded data were categorized according to 2 criteria. Firstly, we used the fourth edition of the Occupational Therapy Practice Framework: Domain and Process(OTPF-4), which is developed for health care professionals, educators, policymakers, and consumers to present a summary of interrelated constructs.^[10] A perspective from occupational therapy implies that the meaning of an activity may vary depending on the individual’s role at the time of activity being performed throughout their life span.^[11] For instance, the activity of writing an essay can be work or a productive activity for authors, education for students, and a leisure activity for some people. Each of the 2 occupational therapists categorizes the activities from the coded data into each occupation area to reflect on an individual’s context of usage following the human-centered approach. After the initial categorization, a discrepancy was obtained and resolved based on the discussion by 2 occupational therapists. Secondly, Generic Indicator Criteria (GIC) as a criteria tool was used as a link between activities and generic indicators.^[12,13] In the previous study for conceptualization and measurement of assistive technology usability, GIC was used to map generic indicators within the activities and participation section. Two occupational therapists did the same categorization process.

2.3.2. Needs and acceptance of the BMI.

This theme of the needs and acceptance is from reflection on attitudes toward the BMI robotic arm from the perspectives of the focus group. It describes how they respond to the BMI technology and the robotic arm, where and how the BMI technology can be utilized, and what should be included to deal with their concerns.

2.3.3. Usability.

Information on the usability from the coded data was categorized into operational speed, accuracy, setup, training period, size/portability, safety, and the cost of the device. These usability categories were derived from discussions of the multidisciplinary expert group because they will be the characteristics required to consider in the design and development process. Usability could reflect on expected standards for the actualized and developed machine in the future.

3. Results

3.1. Activities that users want/need to do using the robotic arm

Participants answered and mentioned activities that they want and need to do most, from 2 questions “What would you like to use this technology for?” and “What areas of rehabilitation assistive technology do you think are the most useful for you?”. Activities obtained by mapping the coded data with OTPF activity domains and generic indicators are shown in Table 3.

Table 3.

Coding and indicator classification.

Activities (or Occupations)[10]		Coded data	Generic indicator[13]
Activities of Daily Living	Feeding	• Drinking water or juice	Self-care
		• Having a meal
	Functional mobility	• Powered wheelchair use	Mobility
Instrumental	Driving and community mobility	• Using public facilities	Mobility
		• Pressing elevator button and getting into/out of an elevator	Outdoor activities
Activities of Daily Living	Communication management	• Using phone/smartphone	Communication
	Home establishment and management	• Operating electronic home appliances, for example, turning on and off light bulbs, unlocking front door for receiving packages or delivery, using remote controller for television	Home management activities
	Safety and emergency maintenance	• Shutting off the gas valve	Home Management activities
		• Emergency call in sudden, unexpected hazardous situations	Communication
Health Management	Physical activity	• Working out	Leisure activities
	Personal care device management	• Ventilator management	N/A
	Symptom and condition management	• Ulcer prevention	N/A
Education	Formal education participation	• Using computer to study	Learning
		• Writing essay assignment
Leisure	Leisure participation	• Taking photos using smartphone	Leisure activities
		• Writing novel/essay/diary using computer/smartphone
		• Sports participation

3.2. Needs and acceptance of the BMI technology

Participants’ responses were controversial but very positive only if the BMI technology could be actualized with safety and high accuracy because it could help them independently perform activities that are meaningful and important to their life. Participants described their needs and how the BMI technology could benefit their lives.

“Even if someone is around me 24/7, there are things that I want to do without other’s help in my own space. I think BMI is a very high-value technology in this respect. It is our strong desire for independent life. So, I think it can ultimately improve my quality of life” (U2)

“Due to degenerative progress following muscular dystrophy, eventually, I would not be able to use assistive computer mouse for my education and drive my powered wheelchair. In my case, BMI is a very necessary technology in the near future, also for people with severe disabilities”. (U8)

“I can use my computer only when I’m sitting down, but when lying down I use my smartphone a lot because it’s hard to use my computer. The problem is I use a mouth stick as a touch pen for smartphone when lying down and writing. […] If I drop it while using the stick, I can do nothing. Someone has to pass it on to my mouth”. (U7)

“We can’t ignore a safety aspect. There is a lot of traumas about it. [...] If there’s no one in the fire situation, or if the body tilts slightly, we [people with muscular dystrophy] can feel our life threatened because our voices can be lower, and our chests can be compressed. I think BMI has a huge advantage in that respect. Even if the invasive method is the only choice I have […] as long as my safety is ensured, I’m willing to use it”. (U4)

However, they also described why they hesitate to use the BMI technology, which is related to BMI acceptance. Their main concern was its effectiveness, with fatigue being a secondary concern. Furthermore, participants represent negative aspects of an invasive method because surgery and skin management would be risky.

“My primary concern is safety in hazardous situations due to accuracy or error rates of the BMI, like before the stairs. If the robotic arm moved in an unexpected way, it could lead to death.” (U1)

“I have experience of using the computer mouse with brain signals a couple of years ago. I felt tired and exhausted a lot and was sweating so much. It was hard. I don’t know how much the technology can be actualized in a level of being able to use it without too much effort.” (U4)

(Invasive method) “If the invasive chip in my brain became old and the malfunction rate got higher, I should receive the surgery again and again. I believe that the more I have more surgery, the more negative impact on my body. So, I would rather choose the noninvasive one”. (U3)

(Invasive method) “If I had an invasive chip in my head, should I always perform wound disinfection? Does that mean that my head skin can be infected? […] I am so worried about that. It’s my brain.” (U4)

3.3. Usability and expected standards

In the domain of BMI’s usability assessment of robotic arms, we divided the key components and characteristics into the following 7 categories: motion speed, accuracy, setup, the training period for operation, safety, size/portability, and cost. Table 4 shows a summary of the response in each usability category.

Table 4.

Potential user preferences as to usability in the brain-machine interface robotic arm.

Usability Features	Summarized Description
Motion speed	Approximately 1 minute per motion (range from 12 seconds to 10 minutes)
Accuracy	At least 80% accuracy for tasks such as food intake (certain tasks required a higher level of accuracy)
Setup	<15 to 30 minutes
Training period for operation	Approximately less than 1 or 2 months with intensive training sessions (range from 1 to 6 months)
Safety	If the robotic arm causes problems such as discomfort or any side effects, these problems can be resolved by outpatient visits
Size/portability	Users prefer a size that allows for easy mounting on the powered wheelchairs they use
Cost	Less than about USD 1800

For motion speeds, the early part of FGIs showed varying degrees of endurance in performing a particular action, from a minimum of 12 seconds to a maximum of 10 minutes, depending on the individual’s propensity or experience in daily life. For example, in the case of food intake, overall, the user group’s desired motion speed was approximately a minute.

“If it takes too long to operate the machine, I don’t think I should use it. I think I’ll ask a guardian or someone next to me for help. I mean, I think it’ll be okay if I have enough time for a person to slowly eat [...] I think about a minute for each action is enough for me”. (U5)

“I think I can wait a minute. [...] As you can see in the video clip, if I’m picking up a cup, or if I’m eating chocolate”. (U3)

For accuracy, the FGI’s entire process showed no significant differences among users, and for typical meals, it was shown that they were willing to accept at least 80% of the accuracy level. However, they described that higher accuracy is needed for certain tasks, such as drinking hot tea or coffee, which are related to their safety.

“For example, when you eat with a robotic arm, you can only get about 70 percent success, but you can’t get more than 10 percent of what you put in a glass of water and spill it on your pants. [...] If you pour rice into your mouth, only 7 out of 10 spoons need to be put into your mouth. If you spill it on your legs while trying to drink hot coffee, it’s okay for people with muscular dystrophy like me because our sensory system is fine, but if you lost sensory system, like people with spinal cord injury, you’ll be in big trouble”. (U4)

In the case of setup, they responded that setup time should be <15 minutes, which is the time required to use a computer at home. However, the preparation for the noninvasive BMI operation requires additional steps, such as applying the gel, wearing caps, and connecting electrodes. The noninvasive one takes approximately 20 to 30 minutes at the current technology status, whereas the invasive one takes a shorter time. Therefore, the preferred setup time was <30 minutes for the entire preparation process.

“I’m using an on-screen keyboard, which takes about five to seven minutes for the person next to me to turn on the computer and to help me hold the computer mouse so that I can use the on-screen keyboard. Also, it takes about 10 minutes to transfer from my bed to my powered wheelchair and drive to my computer. So, I think I can accept this level”. (U1)

For the training period, the range was from a month to 6 months, considering past rehabilitation treatment experiences of the participants, or the period of adaptation to the currently powered wheelchairs or assistive devices. More than half of the participants prefer intensive, short training periods, such as 5 sessions each week for a couple of months, while a couple of participants mentioned that they could tolerate long training periods from 1 to 2 sessions a week for about 6 months. Taken together, less than 2 months with relatively high intensity or frequency of training sessions were preferred.

“I told you about three to six months of training. I think it’s okay if I practice once or twice a week until I achieve 75 percent accuracy”. (U2)

“If training is for an hour every weekday, I’d like to have a two-month period until I will be able to use it as skillfully as you used before”. (U8)

“For me, about a month is appropriate. I think getting intensive training like five times a week would be better rather than intermittent training a couple of times per week, because by doing so I will get used to it sooner, like within a month. So, if possible, I want to do it intensively”. (U8)

“It took me about a month to get used to my powered wheelchair I’m using. If BMI machines were commercialized, I should do it for about a month considering my experience. At first, I couldn’t go out because I was afraid. After a while, I could go out because I kept practicing inside at home. If it is assured that I can skillfully use my BMI robotic arm after sufficient practice, I will go out with it”. (U8)

Regarding safety, the discussion was divided into small side effects, inconveniences, and serious side effects to the extent that outpatient visits and further operations were required. Small side effects that require outpatient visits have shown that users are not burdened by the constant use of hospitals for rehabilitation treatment. However, it was suggested that with serious side effects requiring further operations, this would be appropriate once during use or once every 10 years due to fears of general anesthesia.

Feedback relating to small side effects or inconveniences was:

“I think it’s okay to skip it once or twice a year, or four or five times a year. I’m going to the hospital anyway, so I think that’s okay”. (U1)

Regarding serious side effects, the feedbacks were:

“My experience is that surgery with anesthesia is hard for people with cervical spinal cord injury. I heard that breathing can be difficult, so I can’t breathe easily. I mean, I should risk my life for the surgery, but if I have to do it again, I think I can do it once in my life”. (U7)

“I think I’ll be fine about once every 10 years. As I said earlier, if using BMI can be very helpful to us, I think it’s OK to have it re-opened. Instead, once every 10 years, I think it’ll be a little”. (U4)

For size and portability, since all the participants were powered wheelchair users, they preferred the dimensions of the BMI robotic arm to the extent to that it could be mounted on their wheelchairs when it comes to size and weight levels. The feedbacks were:

“It would be perfect if the device size and weight is portable on the powered wheelchair while driving”. (U1)

“I always have a bag on the back of my wheelchair when I go out”. (U7)

For the cost, only portable computer equipment (for example, smartphones, iPad, and batteries) were identified to activate the robotic arm. The cost of surgery and the robotic arm fluctuates depending on the situation of the market. For most users, the level of government subsidies for the purchase of powered wheelchairs, which is USD 1800, is considered appropriate.

“For the disabled, there are many people who are financially challenged, so it would be a burden if it is too expensive. So, like buying a powered wheelchair, if the government supports a certain amount of the expense, it could reduce the burden. But without the support, I think maximum would be about 2,000,000 won(KRW) [approximately 1800 USD]? I think it’ll be a burden if it goes beyond this amount”. (U1)

4. Discussion

Recent advances in BMI technologies, such as intracortical microstimulation, have contributed to enhancing the motor-sensory control and performance of the robotic arms for people with tetraplegia.^[14–17] In this study, our design team constituted of experts from multidisciplinary fields and focused on the BMI system with potential users with tetraplegia from the beginning of the design and development process for HCD to make the machine more usable and convenient. Since a lack of consideration of user opinion and changes in user needs or priorities are the main factors for technology abandonment, consumer involvement and long-term needs of the consumers with iterative interaction systems were emphasized in the technology design and development process in this project.^[4,18] The purpose of this study was to learn from the perspectives of people with tetraplegia through the FGI to determine specific activities the BMI provides, as well as opinions and usability of the robotic arm with the BMI technology. Overall, participants expressed that the robotic arm using the BMI technology could enable them to gain independence across everyday life, while they also described concerns and factors about the successful use of the machine. Therefore, the usefulness of the BMI functions is the key to its acceptance by prospective users.

Through FGI, participants showed expectations on activities that the BMI could provide, and they wanted to perform using the BMI robotic arm. In order to bridge the language and terminology gap for better communication between medical and engineering fields, activity domain identification and mapping were provided with both OTPF and generic indicators.^[10,12] Identified activities are across various everyday life domains, including communication, home management, activities of daily living, participation in leisure, education, and society, and their personal care/prevention. Since the quality of life, especially in people living with a severe disability, depends on independence in their life, we found that obtaining environmental control using the BMI technology is essential to improving quality of life.^[11,19,20] Information on activities that potential users would need and want to use the BMI robotic arm serves as a useful guideline in the design and development plan, such as necessary elements of the machine required for better performance of the specific activities.^[20]

We found that participants with tetraplegia were dependent on their caregivers in many daily tasks and that those with MD will experience the additional loss of functions due to disease progression. For example, U7 implied that the BMI would be a necessary technology in the future to pursue his education because of the degenerative progression. As an individual’s function, interests, desired activities, and role are ever-changing throughout their life span,^[11] interactive systems and services with follow-up evaluations should be involved to ensure their useful operation.

In accepting the BMI robotic arm, the users’ prioritized elements are safety and high accuracy, just as their main concern was effectiveness. ISO defined effectiveness in assistive technology as accuracy and completeness with which users achieve specified goals.^[4] This reinforces the information gathered from our FGI on the importance of the prioritized elements of effectiveness.

Participants identified a secondary concern of fatigue. A focus group study suggested that if the experience of physiological and psychological fatigue associated with the brain-computer interface is too extensive, then the technology is unlikely to be successfully integrated into the usage for various activities.^[21] This indicates that fatigue in controlling the BMI has a strong impact on BMI acceptance.

In terms of invasive methods of the BMI, issues that emerged were a surgery that might be required multiple times and skin management against wound infection. Throughout the service delivery process, a follow-up management system after the surgery, such as monitoring the transfer rate of the invasive chip and skin status on infection, should be planned during the development process.^[5] Despite the risk of the surgery, more than half of the participants showed a willingness to choose the invasive method if the effectiveness of the invasive method is higher than that of the noninvasive method using dry- and gel-type electrodes. However, a client should be given various choices on the BMI methods with detailed descriptions of positive and negative aspects of each interface’s usability.

To successfully integrate the BMI robotic into their daily use, potential user preferences concerning usability should be reflected. Usability features utilized to assess the service items to be equipped with the robotic arms include motion speed, accuracy, setup, training period, size/portability, safety, and cost. Although the motion speed was somewhat different depending on the user and performance, it was found that users were willing to use the robotic arm if it was able to perform the desired motion within approximately 1 minute or so. Users were also willing to use a robotic arm if it showed about 80 percent accuracy for daily tasks such as food intake. However, certain tasks require a higher level of accuracy for the task performance because drinking hot coffee or tea, for example, could cause scalding or other risks to users who have no sensations in their body, including people with cervical spinal cord injuries. It was also shown that each user would accept the robotic arm if setup took <15 to 30 minutes. Meanwhile, for noninvasive procedures, users expect significant inconvenience if they needed to apply gel, wear caps, and connect electrodes for each use. In terms of the training period for proper operation, potential users wanted a period of less than 1 to 2 months with intensive sessions to expedite acquisition and adaptation in controlling the BMI robotic arm. In the case of safety, it was suggested that as long as the robotic arm functions well with high satisfaction, they showed a willingness to tolerate side effects or discomfort that can be treated through outpatient visits. However, in cases where surgery or further invasive operations are required, users expressed concerns due to pressure from anesthesia risky for people with cervical SCI. For size and portability, users preferred a size that can be mounted on their powered wheelchairs. The cost should not be greater than USD 1800, considering the economic difficulties experienced by the disabled and the government’s funding for various assistive devices they are using.

This study has several limitations. The sample was relatively small. We purposively recruited participants with SCI and MD as potential users of the BMI robotic arms, but the 2 population groups could not be representative of people with tetraplegia. For example, prospective users can be people with amyotrophic lateral sclerosis, stroke, traumatic brain injury, and other disorders with severe physical dysfunction. We also potentially have a bias toward Korean respondents because all the participants were Korean. Another limitation is user experience. Participants in this focus group took an introductory session regarding the current BMI technology, yet 7 out of 8 participants had never used the BMI. Thus, the perspective of the participants would be from the session and some experiences on other used technologies or assistive devices, such as controlling computer mouse using brain signals, assistive robots for self-feeding with button- or joystick-type interface, Smart Internet of Things (IoT) home, etc. Our findings were based on qualitative data, which do not have significant statistical power and cannot be generalized. Further work is required to examine more comprehensive standards and requirements, quantitatively and qualitatively, on usability features.

In summary, our focus group study presents the perspectives of potential users of the BMI robotic arm concerning activities that participants want to perform using the robotic arm, acceptance, and usability for the human-centered design and development process. By using the BMI robotic arm, potential users want to independently participate in activities of daily living, mobility, communication, health management, education, leisure activities, and community. Furthermore, effectiveness with safety and accuracy was the most essential factor for acceptance. Our finding provides desired usability features of the BMI robotic arm as to motion speed, accuracy, setup time, training period, safety, size/portability, and cost. This study represents an initial step in exploring the perspectives and acceptance of the BMI robotic arm for the human-centered design and development process. Future studies on larger, more diverse potential users should be conducted to develop a more usable and convenient robotic arm using BMI technology through the iterative, interactive system. This will lead to the development of a successful BMI robotic arm as an assistive device for people with tetraplegia to regain autonomy and independence, ultimately improve their quality of life, and plan a comprehensive service delivery system for commercializing it in the future.

Author contributions

JHL (corresponding author) was responsible for the entire process of this study. JYP, YJK, HSN, HGS, BMO, SK, and MSB advised on the overall technology status of the brain–machine interface. MYK (first author) was responsible for the analysis and description of transcription data.

Acknowledgments

The authors would like to thank the participants for participating in FGI.