473 - The Digital Motor Output: A conceptual framework for a meaningful clinical performance metric for a Motor Neuroprosthesis

Objectives: The able-bodied population have become increasingly reliant on continuous and independent control of smart devices for instrumental activities of daily living. Upper limb movement is typically required to generate the motor outputs that control these interfaces. For people with tetraplegia, these abilities are lost, significantly compromising their ability to interact with their environment.

Implantable brain computer interfaces (BCIs), including motor neuroprostheses (MNPs), hold promise for restoring lost neurological function. An implantable MNP can directly infer motor intent by detecting motor signals and transmitting the motor signal out of the brain to generate a motor output and subsequently control computer actions. To evaluate these implanted technologies, there is a need for an objective measurement of the effectiveness of MNPs in restoring motor outputs. Here, we propose the concept of digital motor outputs (DMOs) to address this gap.

Design: DMOs are decoded directly from a neural recording during an attempted limb or orofacial movement, then transformed into a command that controls a digital device. A DMO is diverse and can be categorized as discrete or continuous representations of motor control, with varying degrees of freedom.

Results: The clinical utility of the control of a single, discrete DMO has been reported in multiple studies, including the Stentrode thought-controlled digital switch (SWITCH Study; JAMA Neurol. 2023 Jan 9;80(3):270-8.) Collectively, these studies set the stage for the DMO to emerge as a quantitative measure of MNP performance.

Conclusions: Patients with tetraplegia face unique challenges due to the loss autonomy occurring from motor impairment. Implantable MNP BCIs hold promise to restore continuous digital autonomy. The DMO is an objective, clinically meaningful performance metric to assess the integrity of the restored motor outputs delivered by BCIs. Future clinical trials of BCIs must demonstrate that DMOs can be delivered in an accurate and reliable manner.

Scientific Paper Best Paper Award Abstract: Background and Objective

The able-bodied population have become increasingly reliant on continuous and independent control of smart devices for activities of daily living. Upper limb movement is typically required to generate the motor outputs that control these interfaces. For people with tetraplegia, these abilities are lost, significantly compromising their ability to interact with their environment.

Implantable brain computer interfaces (BCIs), including motor neuroprostheses (MNPs) hold promise for restoring lost neurologic function. An implantable MNP can directly infer motor intent by detecting motor signals, transmitting the motor signal out of the brain to generate a motor output and subsequently control computer actions. Despite recent progress in the BCI field, no implantable BCI technology that restores communication or mobility has yet received Food and Drug Administration (FDA) premarket approval or Medicare payment coverage to make it widely available and immediately accessible to users with physical disability. One of the challenges facing clinical translation of BCI technology has been the lack of clinically meaningful BCI performance metrics that can be used to assess efficacy in clinical trials. To evaluate these implanted technologies, there is a need for an objective measurement of the effectiveness of MNPs in restoring motor outputs.

We propose a metric for the minimal clinically meaningful performance of a motor MNP BCI used by people with tetraplegia, and a quantitative measurement of motor output achieved by an MNP. Our goal is to generate a framework that results in the emergence of a new generation of MNP clinical science that underpins the assessment of generalized system performance and guides further development of the field.

Design

The goal of implantable MNP BCI technology is to enable the user to independently produce a repertoire of functionally meaningful outputs at will. The mechanism of action driving this goal is through the detection of a motor intention, and translation of that motor intention into an alternate control signal. Thus, any core performance metric for an MNP should capture its ability to reliably translate a motor intention to a digital output - the “Digital Motor Output” (DMO). DMOs, carrying motor intent information, can then be mapped onto specific or generalized computer actions that can be used to control a personal computer or device. The ability of an implantable MNP BCI to produce a reliable DMO lends itself as a BCI-agnostic approach to measurement of clinical efficacy in this emerging field. To date, BCIs have typically had performance metrics that are linked to the control of a specific effector (e.g., bit rate, correct characters per minute). In a rapidly evolving technological landscape, dependence on specific effector hardware platforms (such as robotic arms, specific tablets or computers, etc.) can create uncertainty around the objective performance of a BCI. Through the use of DMOs, however, one could fully capture the objective performance of an MNP itself, independent of the effector being controlled with the MNP.

DMOs are decoded directly from a neural recording during an attempted limb or orofacial movement, then transformed into a command that controls a digital device. A DMO is diverse and can be categorized as discrete or continuous representations of motor control, with varying degrees of freedom. The relative performance of a DMO could be measured using technical methods previously outlined in the BCI literature, such as confusion matrix derived and information theoretic metrics. While these metrics could be useful for querying specific aspects of a DMO, a clinical outcome assessment of DMOs should capture the reliability of the DMO in a manner reflecting its intended application by the patient. Within the confines of a clinical trial, this may be practicably captured by assessing the accuracy of the DMO across a sufficient number of task repetitions, using a task that captures fundamental DMO performance, executed at an appropriately chosen time point after implant or activation. Using the example of DMO for digital device control, meaningful control can be achieved by successfully selecting a target in space, provided the selection can be made within a tolerable duration. Measuring the accuracy of a DMO for performing this task could be formalized as: Accuracy = Correct Trials / All Trials, for example.

Results

The clinical utility of the control of a single, discrete DMO has been reported in multiple studies, including the SWITCH Trial by Mitchell et al in 2023. Further, in a landmark study published by Vansteensel et al in 2016, a user with locked-in syndrome secondary to amyotrophic lateral sclerosis, with preserved eye-tracking capability, had an implanted BCI that could be used independently at home without expert intervention to control a single click action. This sets the stage for the DMO to emerge as a quantitative measure of MNP performance. Regardless of the resulting action mapped to the DMO – whether it entails computer interactions, robotic limb control, wheelchair navigation, or other applications – these demonstrations of the accurate completion of tasks using transmitted motor intent is what drives BCI usability in the real world.

Conclusions

Patients with tetraplegia face unique and important health challenges due to the loss autonomy that can occur with motor impairment. Implantable MNP BCIs hold promise to restore continuous autonomy through the restoration of motor intent transmission to control the digital environment. An objective performance metric that is clinically meaningful is needed for the assessment of the integrity of the restored motor outputs delivered by BCIs.

The assessment of DMOs in categories of discrete and continuous, and with varying degrees of freedom, is proposed as a functionally relevant and objective performance measure. A DMO demonstrated to be robust in this manner can be functionally applied in generic or universal computer actions, akin to the use of the hand to control generalized functions such as a computer mouse. Future clinical trials of BCIs must demonstrate that DMOs can be delivered in an accurate and reliable manner.