Design-development of an at-home modular brain-computer interface (BCI) platform in a case study of cervical spinal cord injury

Abstract

Objective: The objective of this study was to develop a portable and modular brain-computer interface (BCI) software platform independent of input and output devices. We implemented this platform in a case study of a subject with cervical spinal cord injury (C5 ASIA A).

Background: BCIs can restore independence for individuals with paralysis by using brain signals to control prosthetics or trigger functional electrical stimulation. Though several studies have successfully implemented this technology in the laboratory and the home, portability, device configuration, and caregiver setup remain challenges that limit deployment to the home environment. Portability is essential for transitioning BCI from the laboratory to the home.

Methods: The BCI platform implementation consisted of an Activa PC + S generator with two subdural four-contact electrodes implanted over the dominant left hand-arm region of the sensorimotor cortex, a minicomputer fixed to the back of the subject's wheelchair, a custom mobile phone application, and a mechanical glove as the end effector. To quantify the performance for this at-home implementation of the BCI, we quantified system setup time at home, chronic (14-month) decoding accuracy, hardware and software profiling, and Bluetooth communication latency between the App and the minicomputer. We created a dataset of motor-imagery labeled signals to train a binary motor imagery classifier on a remote computer for online, at-home use.

Results: Average bluetooth data transmission delay between the minicomputer and mobile App was 23 ± 0.014 ms. The average setup time for the subject's caregiver was 5.6 ± 0.83 min. The average times to acquire and decode neural signals and to send those decoded signals to the end-effector were respectively 404.1 ms and 1.02 ms. The 14-month median accuracy of the trained motor imagery classifier was 87.5 ± 4.71% without retraining.

Conclusions: The study presents the feasibility of an at-home BCI system that subjects can seamlessly operate using a friendly mobile user interface, which does not require daily calibration nor the presence of a technical person for at-home setup. The study also describes the portability of the BCI system and the ability to plug-and-play multiple end effectors, providing the end-user the flexibility to choose the end effector to accomplish specific motor tasks for daily needs. Trial registration ClinicalTrials.gov: NCT02564419. First posted on 9/30/2015.

Keywords: Neuroscience; Rehabilitation; Signal processing systems.

Conflict of interest statement

JJ reports support from Medtronic Inc, during the conduct of the study; grants and personal fees from Medtronic Inc, grants from Boston Scientific, outside the submitted work. Outside the submitted work, EB is a cofounder of PASCALL, a company developing closed-loop physiological control systems and a cofounder of Neuradia, a company developing agents to promote recovery of consciousness following general anesthesia or sedation. In addition, EB has a patent SYSTEMS AND METHODS FOR ANALYZING ELECTROPHYSIOLOGICAL DATA FROM PATIENTS UNDERGOING MEDICAL TREATMENTS. MI reports personal fees from Medtronic—Visualase outside the submitted work.

© 2022. The Author(s).

Figures

Fig. 1

System overview. Electrocotricography (ECoG) signals are recorded using two four-contact subdural strips placed on the surface of the sensorimotor cortex. ECoG signals were transmitted by a subcutaneous implant to an external receiver which delivers it to the minicomputer for processing. The decoder classified the signal as a motor imagery command that is then sent over Bluetooth to actuate the mechanical glove

Fig. 2

Remote ECoG Data collection. Results of data collected from the BCI system using the Activa PC + S Nexus device. A, B ECoG data from channels 1 and 3 and filtered through a 1 Hz high-pass filter. C, F Power spectra for channels 1 and 3 (shown as μ±σ). D, E Time–frequency spectrograms of averaged windows (6.4 s, N = 2051) of data that show the changes in power of frequency bands between 1–100 Hz from filtered, averaged, and normalized data during multiple transitions from the REST state (indicated by the first half of the time series) to the MOVE state

Fig. 3

Mobile application overview. The App functioned as the GUI for the subject to interact with the BCI software running on the computer. a The home screen displayed the currently selected input and output devices in use. The blue dot in the upper-right marked the system’s status. b An input device selection screen, allowing the subject to select from more devices. c A settings page that allowed the subject to adjust parameters (such as decoder threshold, end effector motor speed, etc.) for a given device. These settings were defined by the software Class’s Application Programming Interface (API) on the computer’s end and were delivered over Bluetooth for dynamic display. d A data collection session that presented prompts to the subject either for assessing accuracy or applying calibration to the device’s decoder

Fig. 4

Application control flow. a The main application is initialized by a daemon script—or background running process—that ensured the program was always running while the computer was on. The computer application ran multiple coroutines asynchronously to allow for nearly uninterrupted data streaming between input and output devices as well as for Bluetooth communication. b The main application process iteratively made calls to classes that manage input and output devices. These device manager classes contained public methods for obtaining device input and sending commands to output devices. These device classes communicated with their hardware counterpart over serial port communication. Importantly, an array of devices may exist for the subject to use. These could be individually selected via the App over BLE

Fig. 5

Bluetooth low energy communication Benchmark. The time delay observed during a data collection session (n = 750). Bluetooth transmission time delay was measured as the difference between the time at which the display prompted changed on the App and the time the prompt was changed on the computer system to initiate BLE (prompt to display notifying characteristic)

Fig. 6

System profiling. Sunburst diagrams representing the proportions of time spent to process incoming data (left) and send the decoded output to the glove (right). The center of each sunburst diagram represents the process to obtain the decoded neural signal (left), or to send the command to the glove (right). Each arc surrounding the center point represents a subprocess needed to be carried out to process incoming data (left) or send data (right). The length of the arc represents the proportion of time taken for a subprocess to complete relative to subprocesses that depend on it for completion. While there are many subprocesses, those relevant to the BCI software are highlighted. The remaining subprocesses are system sepectific processes such as input–output operations

Fig. 7

Decoder classification. Classification performance of the decoder associated with the Nexus telemeter input device. Month 0 indicates assessment of training data and the subsequent months indicate the number of months since training. The black dotted line indicates the global median of 87.53%

Fig. 8

System setup time. Elapsed time taken by the subject’s primary caregiver to set up the system. Repeated measurements once a day for several days. A Presents the total time to setup the system, while B presents the times for the different set up steps that sum to the total elapsed time. Calibration time is not included because it was not necessary during datily setups