Intracortical brain-computer interfaces (iBCIs) enable people with tetraplegia to gain intuitive cursor control from movement intentions. To translate to practical use, iBCIs should provide reliable performance for extended periods of time. However, performance begins to degrade as the relationship between kinematic intention and recorded neural activity shifts compared to when the decoder was initially trained. In addition to developing decoders to better handle long-term instability, identifying when to recalibrate will also optimize performance. We propose a method, “MINDFUL”, to measure instabilities in neural data for useful long-term iBCI, without needing labels of user intentions. Longitudinal data were analyzed from two BrainGate2 participants with tetraplegia as they used fixed decoders to control a computer cursor spanning 142 days and 28 days, respectively. We demonstrate a measure of instability that correlates with changes in closed-loop cursor performance solely based on the recorded neural activity (Pearson r = 0.93 and 0.72, respectively). This result suggests a strategy to infer online iBCI performance from neural data alone and to determine when recalibration should take place for practical long-term use.

Longitudinal iBCI Cursor Control using a Fixed Decoder

These data are released with the manuscript - by Pun, T. K. et al. Measuring instability in chronic human intracortical neural recordings towards stable, long-term brain-computer interfaces. Nature Communications Biology 7, 1363 (2024).

Written by Tsam Kiu Pun, 2024. Brown University.

Dataset Description

This dataset comprises intracortical neural signals recorded from two participants enrolled in the BrainGate2 pilot clinical trial (NCT00912041):

• T5: A 65-year-old right-handed male with a C4 AIS-C SCI incurred approximately 9 years prior to enrollment.
• T11: A 37-year-old right-handed male with a C4 AIS-B spinal cord injury (SCI) incurred approximately 11 years prior to enrollment.

This research was conducted under an Investigational Device Exemption (IDE) granted by the US Food and Drug Administration (IDE #G090003; CAUTION: Investigational device. Limited by Federal law to investigational use). 

All sessions took place at the participants' residences.

Intracortical neural recordings and neural features

Each participant had two 96-channel microelectrode arrays (Blackrock Neurotech, Salt Lake City, UT) placed in the dominant (left) hand/arm knob area of the precentral gyrus 2. The following neural features are included in this dataset:

• T5: Multi-unit threshold-crossing spike rates (RMS < -3.5) per electrode.
• T11: Multi-unit threshold-crossing spike rates (RMS < -3.5) and power in the spike band (250-5000 Hz) per electrode.

BCI behavioral tasks

• T5: Performed a closed-loop 2D random target selection task across 6 sessions spanning 28 days (trial days 2121-2149), using a fixed linear decoder. Each session included two to four 4-minute closed-loop blocks. Total task time: 84 minutes, 1200 trials.
• T11: Performed a closed-loop 2D point-and-click center-out-and-back task across 15 sessions spanning 4 months (trial days 658-800), using a fixed RNN-based decoder. Each session included two 5-minute task blocks, except for trial day 751 with one block. Total task time: 145 minutes, 1840 trials.
• In addition for T11, approximately 16 minutes of personal use (i.e. active web browsing) on trial day 658 and 10 minutes of random target tasks on trial day 665 were included to assess the robustness of MINDFUL using different task as the reference dataset. This data is stored in separate subfolder within the folder named "T11(additional)/". The same RNN decoder was used online.