Assessment of brain-machine interfaces from the perspective of people with paralysis

Abstract

Objective: One of the main goals of brain-machine interface (BMI) research is to restore function to people with paralysis. Currently, multiple BMI design features are being investigated, based on various input modalities (externally applied and surgically implantable sensors) and output modalities (e.g. control of computer systems, prosthetic arms, and functional electrical stimulation systems). While these technologies may eventually provide some level of benefit, they each carry associated burdens for end-users. We sought to assess the attitudes of people with paralysis toward using various technologies to achieve particular benefits, given the burdens currently associated with the use of each system.

Approach: We designed and distributed a technology survey to determine the level of benefit necessary for people with tetraplegia due to spinal cord injury to consider using different technologies, given the burdens currently associated with them. The survey queried user preferences for 8 BMI technologies including electroencephalography, electrocorticography, and intracortical microelectrode arrays, as well as a commercially available eye tracking system for comparison. Participants used a 5-point scale to rate their likelihood to adopt these technologies for 13 potential control capabilities.

Main results: Survey respondents were most likely to adopt BMI technology to restore some of their natural upper extremity function, including restoration of hand grasp and/or some degree of natural arm movement. High speed typing and control of a fast robot arm were also of interest to this population. Surgically implanted wireless technologies were twice as 'likely' to be adopted as their wired equivalents.

Significance: Assessing end-user preferences is an essential prerequisite to the design and implementation of any assistive technology. The results of this survey suggest that people with tetraplegia would adopt an unobtrusive, autonomous BMI system for both restoration of upper extremity function and control of external devices such as communication interfaces.

Figure 1

The eight illustrations provided to the survey participants, depicting (in order from top left) “30+ electrodes glued to the head,” “cap with wires,” “eye tracking glasses,” “wired device in the brain,” “wired device on the brain,” “wireless device in the brain,” “wireless cap,” and “wireless device on the brain.” Note that the same form factor was used for both the wireless transmitter and the wired preamplifier in the depictions of the implantable technologies to isolate the perceived contribution the sensor to the desirability of adoption.

Figure 1

The eight illustrations provided to the survey participants, depicting (in order from top left) “30+ electrodes glued to the head,” “cap with wires,” “eye tracking glasses,” “wired device in the brain,” “wired device on the brain,” “wireless device in the brain,” “wireless cap,” and “wireless device on the brain.” Note that the same form factor was used for both the wireless transmitter and the wired preamplifier in the depictions of the implantable technologies to isolate the perceived contribution the sensor to the desirability of adoption.

Figure 2

Examples of the text associated with each illustration, which described the design of the device: its usage (including information on donning and doffing the device, assistance needed, cleaning and maintenance); a description of any surgical procedures required; physical restrictions while using the device; and any known side effects.

Figure 3

a. Legend used throughout for creating the composite graphs of technology, control type, and likelihood of adoption. Red colored bars denote upper cervical injury, while blue colored bars denote lower surgical injury. Each of these groups is subdivided further into participants <10 years post-injury (lighter colors) and >10 years post-injury (darker colors). Categories of “likely” and “very likely” were depicted in increasingly saturated colors, with “very unlikely” and “unlikely” depicted in shades of grey, and “undecided” depicted as white. b. Composite graph showing likelihood of adopting any particular control type, independent of the BCI technology used as a sensor. c. Composite graph showing likelihood of adopting any BCI technology, independent of control type. Interest in each Control Type for any Technology Interest in each Technology for any Control Type

Figure 3

a. Legend used throughout for creating the composite graphs of technology, control type, and likelihood of adoption. Red colored bars denote upper cervical injury, while blue colored bars denote lower surgical injury. Each of these groups is subdivided further into participants <10 years post-injury (lighter colors) and >10 years post-injury (darker colors). Categories of “likely” and “very likely” were depicted in increasingly saturated colors, with “very unlikely” and “unlikely” depicted in shades of grey, and “undecided” depicted as white. b. Composite graph showing likelihood of adopting any particular control type, independent of the BCI technology used as a sensor. c. Composite graph showing likelihood of adopting any BCI technology, independent of control type. Interest in each Control Type for any Technology Interest in each Technology for any Control Type

Figure 3

a. Legend used throughout for creating the composite graphs of technology, control type, and likelihood of adoption. Red colored bars denote upper cervical injury, while blue colored bars denote lower surgical injury. Each of these groups is subdivided further into participants <10 years post-injury (lighter colors) and >10 years post-injury (darker colors). Categories of “likely” and “very likely” were depicted in increasingly saturated colors, with “very unlikely” and “unlikely” depicted in shades of grey, and “undecided” depicted as white. b. Composite graph showing likelihood of adopting any particular control type, independent of the BCI technology used as a sensor. c. Composite graph showing likelihood of adopting any BCI technology, independent of control type. Interest in each Control Type for any Technology Interest in each Technology for any Control Type

Figure 4

a. Legend (see Figure 3a). b-i. Composite graphs showing likelihood of adopting each of the 8 different BCI technologies for a given control capability.