High-density single-unit human cortical recordings using the Neuropixels probe
Chung, Jason et al. (2022), High-density single-unit human cortical recordings using the Neuropixels probe, Dryad, Dataset, https://doi.org/10.7272/Q6ST7N3B
This dataset contains the raw broadband and LFP data collected using the Neuropixels probe in human participants in the intraoperative setting. The Neuropixels probe allows for single-unit recordings simultaneously over 384 contacts. The experimental design and procedures are explained in Chung, Sellers et al (2022), Neuron. Each participant received an identifier (e.g. NP##). Multiple blocks of data were collected for some participants, denoted by B##. Recordings were started prior to probe insertion, to allow for visualization during insertion - thus spike sorting should not be conducted on the full duration of these raw recordings but rather following termination of insertion artifact.
Full methodological details are available in the published paper. Reproduced here is the most relevant information:
Probe preoperative preparation and sterilization
Neuropixels 1.0 NHP-short probes, with 10mm long shanks and metal dovetail caps (IMEC, Leuven, Belgium), were used for all recordings. These probes are adapted and modified versions of the Neuropixels 1.0 rodent probes (Jun et al., 2017) that are electronically identical, with the notable difference of increased shank thickness from 24μm to 97μm (Paulk et al., 2022). This increased thickness allowed for tolerance of greater mechanical forces for larger brain recordings. Two 27G subdermal needle electrodes (Ambu, Columbia, MD) were soldered separately to the probe flex-interconnect to serve as ground and reference using lead-free solder and two strands of twisted 36 AWG copper wire. The connected needle electrodes were approximately 20cm long, to allow for insertion into the skin flap following probe insertion. No additional modifications were made to the Neuropixels probe. Probes with soldered needle electrodes were installed into custom CleanCut mounting cards (Oliver Healthcare Packaging, Grand Rapids, MI) for storage and transport.
The Neuropixels probe was secured to the metal cap dovetail probe mount (IMEC, Leuven, Belgium). The probe mount was in turn attached to either a Elekta microdrive (Elekta, Stockholm, Sweden) or Narishige (Tokyo) micromanipulator (MM-3 or M-3333). The manipulator/microdrive was secured to the Mayfield skull clamp using an 3-joint mounting arm (Noga NF9038CA) and Nano clamp (Manfrotto 386BC-1, Cassola, Italy) assembly attached to the primary articulating arm and C-clamp of the Integra Brain Retractor System A2012 (Integra, Princeton, NJ). Probes, headstages, interface cables, Narishige micromanipulators, screwdrivers, and probe mount with metal cap dovetail were all separately sterilized according to standard protocols of ethylene oxide sterilization, while the Elekta micromanipulators were sterilized using Sterrad prior to use. Probes were not reused across participants.
Recording site selection
In all cases, the insertion of Neuropixels probes was targeted to cortical tissue that was destined for resection in the same procedure based on clinical criteria. In cases where a tumor or cavernoma was resected, the recording was targeted to radiographically normal tissue (no T2 hyperintensity) which would require resection as part of the transcortical approach. In all cases the recorded tissue appeared normal intraoperatively. The specific sites were selected to be the crown of surface gyri which allowed for direct visualization and monitoring of the insertion and penetration through cortical layers. See below for post-hoc insertion localization.
Probe positioning and insertion
Each participant was secured with a Mayfield skull clamp. A compatible C-clamp was then attached to the skull clamp, and in turn, a primary articulating bar was attached to the C-clamp (all items contained in Integra Brain Retractor System A2012). A clamp (Manfrotto 386BC-1) was used to attach a 3-joint mounting arm (Noga NF9038CA) to the articulating bar. The 3-joint mounting arm was connected to a micromanipulator (Elekta or Narishige MM-3 or M-3333) which held the Neuropixels assembly. All equipment was sterilized according to standard protocols using either Sterrad or ethylene oxide sterilization. The articulating arms were positioned to place the micromanipulator above the target insertion site, and the Neuropixels probe was lowered using the micromanipulator to a target depth of 6 to 8 mm from the brain surface, at a rate of 50–75 μm/sec. Insertion trajectory was approximately perpendicular to the surface. Insertion locations were estimated through a combination of intraoperative navigation, intraoperative photos taken during the surgery, and histology when available. In some cases, a piotomy was performed at the site of insertion, which also reduced the risk for probe fracture. We hypothesized this would also increase unit yield, although this did not appear to have an effect. Recordings were typically started to visualize probe insertion (Video S1).
Data were collected using a custom-constructed rig including a Windows machine, PXI chassis (NI PXIe-1071), PXI Multifunction I/O Module (NI PXIe-6341), NI SHC68-68-EPM shielded cable, Neuropixels PXIe Acquisition Module, and NI BNC-2110 Connector Block with SpikeGLX 3.0 (http://billkarsh.github.io/SpikeGLX/) acquisition software. In some experiments, speakers were used to present auditory stimuli and a microphone was used for recording; these signals were also acquired as analog inputs synchronized with the neural data. The Neuropixels probe was configured to acquire from 384 channels in a ‘long column’ layout, providing the greatest possible depth span of recording while acquiring from a contact at each depth. Total recording span was 7.67mm. Ground and reference needle electrodes were inserted into the scalp adjacent to the craniotomy. AP gain was 500 and LF gain was 250. During data acquisition, all non-essential equipment in the operating room was unplugged or run using battery power in order to reduce electrical noise. By iteratively turning off pieces of equipment, when possible, and listening to the high-pass filtered signal from a recording channel, we found the operative table, electrocautery, and electronics associated with intravenous lines should be on direct current, including syringe pumps and blood/fluid warmers. Other key environmental sources of noise included neuromonitoring and the Brainlab neuronavigation system.
Data were acquired using SpikeGLX 3.0. The raw data are stored in binary files which can be read into Matlab or Python. Example functions provided by SpikeGLX developers are available at https://billkarsh.github.io/SpikeGLX/#post-processing-tools for reading raw data into Matlab and Python.
Howard Hughes Medical Institute