3. Experimental Methods and Setup

Figure 3.1A shows the standard dual-arm stereotaxic apparatus for rats that we used, manufactured by Stoelting Company. The apparatus includes manipulator arms, a rat adaptor, U-shape frame, and baseplate. When setting up, ensure sufficient space around the experiment table for freely operating around the stereotaxic apparatus. Before securing the rat head, remove the manipulator arms from the apparatus. Then use the left and right ear bars on the U-shape frame, along with the incisor bar and nose clamp on the rat adapter, to immobilize the rat head. Beginners typically need repeated practice to master this process. Below is the operating procedure used in our laboratory.

Full size image

After securing a rat on the stereotaxic apparatus, perform the craniotomy as below. First, shave the surgical areas on its head and nose, then clean the areas with alcohol swabs. To fix reference and ground electrodes on the nose for signal recording, make a midline incision along the nasal tip using a scalpel, gently scrape away the connective tissue on the nasal bone surface, and clean it with saline solution (Fig. 3.2). Using a microdrill, make two holes on either side of the nasal bone suture, about 5 mm apart. Fix a stainless steel bone screw into each hole to serve as the reference and ground electrodes, respectively, and wire the screws to the amplifier. We used rat bone screws from BASi (Bioanalytical Systems, Inc.), measuring ~ 1 mm in diameter and 2 mm in length (No. MF-5182, https://www.bioanalytical.com/products/MF-5182). Alternatively, the reference and ground electrodes can be placed on the posterior skull, provided they are positioned away from recording sites (Schjetnan and Luczak 2011).

Full size image

Next, make a midline sagittal incision along the scalp, approximately 1.5–2 cm (Fig. 3.2). Remove the subcutaneous connective tissue to expose the skull, and gently scrape skull surface clean. If bleeding occurs, use a thermocautery to staunch bleeding sites. Clean the skull surface with saline solution and then let it dry. Locate the bregma and mark it with a black pen. According to electrode implantation targets, draw the outline for the cranial window. For our experiments in the unilateral hippocampal CA1 region, we typically used a window of about 8 × 5 mm on the left skull. Use an electric microdrill to grind evenly along the window outline, flushing away the drilling powder with saline solution until the thinned outline faintly reveals the red dura underneath. Switch to a small carving knife for light etching. Once the bone along the outline is almost cut through, use a bone rongeur to pry open a corner of the bone flap. Gently remove the bone flap to expose the underlying dura. Then, carefully inspect and clean bone fragments along the window edge using a bone rongeur, smoothing the edge to prevent fragments from piercing the blood vessels under the dura, which may cause bleeding during the subsequent experiment period. This smoothing process can also prevent electrode damage when inserting along the skull window edge.

The rat stereotaxic apparatus features two manipulator arms, one on each side (Fig. 3.1A). Additional separated micro-manipulators, such as those from World Precision Instruments (WPI), can be placed around the stereotaxic apparatus to hold more electrodes or devices. For experiments requiring multiple electrodes, first position each manipulator arm, then attach the electrodes sequentially. To successfully insert electrodes through the small cranial window into the target sites in brain, some electrodes (holder arms) need to be adjusted to appropriate tilt angles. In our experiments, we typically inserted the recording array electrode vertically while setting other electrodes at specific angles. The entry position coordinates for electrodes on the exposed dura must be calculated accurately based on their target sites and tilt angles (as detailed in Sect. 3.2.1).

We used FHC concentric bipolar electrodes for electrical stimulations (see Sect. 3.4.2). This type of electrodes is relatively rigid and durable. Before mount it on the manipulator arm, clean the electrode by rinsing it with deionized water or gently wiping it with an alcohol disinfectant wipe. Then fix it to the manipulator using the corner clamp attached to the small cube at the holder head (Fig. 3.1A). This cube has 1 mm-spaced grooves along its three sides. Embed the electrode needle into one groove and secure it with the clamp. Finally, connect the two electrode wires to the output ports of electrical stimulator (see Sect. 3.5.4).

We used NeuroNexus microelectrode to record neural signals (see Sect. 3.4.1). Due to its extreme thinness and fragility, this electrode should be placed last. Begin by rinsing the electrode surface with deionized water. When placing the electrode, handle it carefully and prevent its tip from touching any surrounding objects to avoid breakage. NeuroNexus electrodes for acute recordings have a rod (about 2 cm long and 3 mm wide) between the electrode shank (containing contacts) and the electrode connector. Press this rod against the cube on the holder head (Fig. 3.1A), then secure it with the tightening clamp. Once secured, wire the preamplifier to the electrode connector.

Notably, when mounting an electrode, ensure that its rod is parallel with the holder rod. Otherwise, the electrode will not advance straight along the direction of holder axis (the moving direction) but instead enter the brain tissue at an angle, cutting and damaging the tissue like a knife.

Some electrodes, like the fragile NeuroNexus arrays (only 15 μm thick and made of film), are not strong sufficiently to penetrate the dura on their own. In such cases, the dura at the electrode entry site must be punctured before electrode insertion. The sharp beveled tip of a syringe needle can be used for this purpose. During puncturing, lift the electrode as high as possible to avoid accidental contact and damage. When puncturing the dura near blood vessels, be prepared to immediately control any bleeding if vessels beneath are pierced. The puncture hole is typically too small to see with the naked eyes. Slowly advance the electrode from the dura surface while carefully observing changes at the entry point. If the dura hasn't been punctured, the electrode tip will create a visible indentation when it pressing against the dura surface, indicating a need for re-puncturing. Although these operations can be performed with the naked eye, using a magnifying glass makes them easier.

After penetrating the dura and positioning all electrode tips on its surface, cover the stereotaxic apparatus with a shielding cage (or Faraday cage) and start signal recordings. While advancing the electrodes, monitor real-time changes in recorded neural signals to determine electrode positions within the brain (see Sect. 3.2.3).

The FHC electrodes we used are quite rigid and can pierce the dura on their own. However, when an electrode directly pierces the dura, its tip can visibly indent and compress the underlying brain tissue. Therefore, it is better to puncture the dura first. The perforation hole should be minimal—just large enough to allow the thin electrode entry. If the hole is too large, brain tissue may gradually protrude through the opening over extended experiment period.

After surgery, add saline or artificial cerebrospinal fluid over the cranial window to keep the exposed dura and surrounding skull surface moisture. The incised scalp forms a natural circular “dam” on the exposed skull (Fig. 3.2) that can hold fluid. Surface tension from scalp oils can prevent fluid spillage even when its level is slightly above the “dam”. During experiment, replenish the fluid as needed to prevent the exposed surface from drying out.

The acute in-vivo rat experiments described in this book were conducted under urethane anesthesia. Urethane (ethyl carbamate) is an acetylcholinesterase inhibitor. It is commonly used as an anesthetic in neurophysiological experiments on rats and other small animals, providing mild and long-lasting anesthesia. At normal doses, urethane has minimal impact on neuronal activity, possibly causing slight increases in neuronal firing thresholds and reduced firing rates. It has negligible effects on both glutamate-mediated excitatory synapses and GABA-mediated inhibitory synapses. It only minimally suppresses sensory responses to stimuli such as body touch (Sceniak and Maciver 2006; Wu et al. 2007; Shirasaka and Wasterlain 1995). Additionally, this anesthetic barely affects respiratory ventilation, and artificial ventilation is not needed during experiments.

If euthanasia is required after experiments, a quick and simple method involves injecting a saturated potassium chloride solution into the rat heart (Cartner et al. 2007). The sudden increase in the extracellular concentration of potassium ions can cause immediate cardiac arrest, leading to death.

The bregma on rat skull typically serves as the origin of the three-dimensional coordinate system (X, Y, Z) to determine the entry point on the brain surface for each electrode. The process involves three main steps: (1) Look up the rat brain stereotaxic atlas (Paxinos and Watson 2007) for the coordinates of the target sites in brain—the desired points for the electrode tips. (2) Measure the electrode inclination angles from the experiment setup. (3) Calculate the three distances: the anterior–posterior and lateral distances between the entry point and the bregma, and the electrode advancing distance after entering the brain. These distances, typically measured in millimeters, guide the movements of the X, Y, and Z axes on manipulator arms or micromanipulators.

As shown in Fig. 3.3A, when an electrode is advanced vertically into the brain, (i.e., at a zero inclination angle), determining the coordinate data of its entry point on the brain (or skull) surface is straightforward. You simply need to look up the anterior–posterior and lateral distances between the target site and the bregma in the brain atlas. These distances are just the displacements along the X- and Y-coordinates, respectively, with a negative value indicating a position posterior to the bregma. However, as illustrated in Fig. 3.3B, when the holder (with electrode) of WPI micromanipulator forms an angle with the vertical, the displacement along its X-axis is different from the displacement of electrode tip in the anterior–posterior direction on the skull surface. The two displacements form the right and oblique sides of a right triangle, labeled in Fig. 3.3B as “X-axis displacement” and “Anterior–posterior displacement”. In this case, calculating the X-axis displacement involves two steps: First, determine the “anterior–posterior displacement” from the bregma to the entry point, based on both the target site and the electrode inclination angle. Second, calculate the required “X-axis displacement” using trigonometric functions to the right triangle shown in Fig. 3.3B. It's important to note that the X-axis displacement is not equivalent to the anterior–posterior distance between the bregma and the entry point—confusing them will lead to inaccurate electrode positioning. Figure 3.3B illustrates this for an inclination angle in the X-axis direction, the same calculation method applies to an inclination angle in the Y-axis direction.

Full size image

Additionally, it is important to distinguish between an intended entry point on the skull surface and one on the dura surface above the brain cortex, since skull thickness affects the calculations. Since the bregma is marked on the skull surface, for entry points on the exposed dura surface, you must subtract the skull-to-dura distance when calculating the post-entry advancing depth.

When placing multiple electrodes or other implanted devices, position them one by one at their entry points in a posterior-to-anterior order. For each electrode, first, align the electrode tip with the bregma, and read the initial coordinates in three directions on the electrode manipulator. Then, raise the electrode high enough to prevent contact with the skull and other electrodes during its movement. Finally, move the electrode tip to the entry point by adjusting the coordinates based on the calculated displacements.

Precise positioning requires meticulous attention to detail. The bregma and lambda on the rat skull should be adjusted to the same horizontal plane. Although size variations among adult rat skulls are typically negligible, body weight-related differences do exist. For accurate positioning, the distance between the bregma and interaural line needs to be corrected based on the actual skull size. The stereotaxic atlas provides a standard distance of 9.0 mm (Fig. 3.3A). When the actual distance of experimental rat differs significantly from this value, apply a correction factor—the ratio of the actual distance to 9.0 mm—to the anterior–posterior coordinate. For more positioning guidelines, refer to stereotaxic atlas literature. A classic reference is “The Rat Brain in Stereotaxic Coordinates” by Paxinos and Watson (2007), which is available in multiple editions.

The rat stereotaxic apparatus shown in Fig. 3.1A is a basic setup that uses knobs to manually move the manipulator arms in three directions with a resolution of about 0.1 mm. Although this manual adjustment can meet basic requirements for recording neuronal signals such as field potentials and unit spikes, turning the knobs may cause slight vibrations in the manipulator arm and attached electrode. These vibrations can damage neurons near the advancing electrode, degrading the quality of signal recordings, particularly single-neuron firing (spikes). Consequently, recordings become limited to firing from distant neurons, resulting in smaller amplitudes and lower signal-to-noise ratios. The situation can worsen when the manipulator arm lacks sensitivity and requires greater force to move. To minimize these vibrations, manual adjustments should be as steady as possible.

RWD Life Science Co. Ltd., a manufacturer of rat stereotaxic instruments in Shenzhen, China, once developed a fine-tuning manual device modified from a spiral micrometer. This device can be easily installed on the manipulator arm. In such setup, the original Z-coordinate knob serves as a coarse adjustment to lower the electrode on the brain surface. Then the micrometer knob can provide fine adjustment for advancing the electrode in the brain. Although still manually operated, the device requires minimal force to turn the knob. The micrometer itself has both coarse and fine adjustments, greatly improving movement resolution and enhancing recording quality. Skilled operators can apply gentle force to turn the knobs, minimizing vibrations in the manipulator and electrode. One drawback is the difficulty in monitoring electrode displacement by reading the micrometer scale during implantation. Electric fine-tuning devices and those with digital displays have since provided more convenient for this manipulation.

Electric micromanipulators can minimize damage to brain tissue during electrode implantation. In addition to stereotaxic instruments with motorized manipulator arms, various standalone electric micromanipulators are available. We used an electric single-axis micromanipulator from the Finnish company Sensapex (Fig. 3.4). The device consists of two parts: a micromanipulator and a controller (Fig. 3.4A, B), achieving 10 nm step precision with a 20 mm total travel distance. The user-friendly controller can display digital readouts for displacement, speed, and other parameters. When connected to a computer via USB port, its accompanying software can control the micromanipulator movement. The micromanipulator can be mounted onto the holder of stereotaxic instrument (Fig. 3.4C). During work, the Z-coordinate knob on the stereotaxic instrument serves as coarse adjustment to position the electrode tip on the brain surface. The Sensapex micromanipulator can then drive the electrode into the brain, advancing it smoothly with consistent force and minimal vibration. Using this motorized micromanipulator, when we moved a NeuroNexus array electrode up and down in the rat brain repeatedly, we successfully recorded unit spikes from the same adjacent neurons each time the electrode returned to a same position—indicating minimal damage to neurons from the electrode movements.

Full size image

An electric micromanipulator may produce high-frequency noise from their micromotor coils, which may interfere with recordings. To address this issue, these devices typically come with a grounding post. The noise usually occurs only during electrode movement when the micromotor is active, and ceases once movement stops.

The unique laminar structure of hippocampal region can facilitate accurate induction and recording of various neural signals. In in-vitro experiments using hippocampal slices, this laminar structure can be seen with the naked eyes or under a microscope, making it easy to verify correct electrode positioning. However, in in-vivo experiments with intact brains, it is impossible to directly observe recording and stimulation sites in real-time. Relying solely on post-experimental histological analysis of stained brain slice to verify electrode positions can significantly reduce experimental efficiency. Instead, the efficiency can be greatly improved by determining electrode positions through real-time signal recordings during experiments. The neuronal firing patterns and the unique waveforms of evoked potentials captured by an array electrode can enable the real-time position determination, as described below.

As shown in Fig. 3.5, the recording electrode was a NeuroNexus 16-channel linear array (model A1 × 16-5mm-25-177). The electrode shank was 5 mm long, with 16 contacts spaced 25 μm apart, spanning 375 μm at the tip (detailed in Sect. 3.4.1). The electrode was vertically inserted into the rat brain at an entry point 3.6 mm posterior and 2.6 mm lateral to the bregma. As it was advanced, real-time recordings showed unit spikes in signals within the 500 Hz to 5 kHz frequency range. These spikes were extracellularly recorded single-neuron action potentials (see Sect. 4.1). This ongoing monitoring helped us determine the precise electrode position.

Full size image

When the electrode passed vertically through the cerebral cortex into the hippocampal CA1 region (about 2 mm deep from the brain surface), it encountered the thin pyramidal cell layer (pcl) with dense somata. Figure 3.5A shows that the recording contacts in this layer (channels 6–11) captured more spikes (visible as short vertical lines with varying lengths) than the contacts in the layers containing basal dendrites (above) or apical dendrites (below). It should be noticed that spikes can also be recorded beyond the pcl. Interneurons are distributed across almost all layers of the CA1 region (Klausberger and Somogyi 2008; Pelkey et al. 2017), though at a much lower density than pyramidal neurons in the pcl. Because of their low density, interneurons are not always recorded, which explains why clear spikes appeared only across the pcl region in Fig. 3.5A. However, interneurons fire at a much higher rate than pyramidal neurons, and once recorded, their spikes can appear quite dense (see Fig. 5.27A and 5.28A). In the layers of basal and apical dendrites, where pyramidal neurons’ cell bodies are absent, recorded spikes typically come from the rapidly firing interneurons. Since these spikes usually originate from a single interneuron, they can show consistent amplitudes.

As the electrode advanced further, it passed sequentially through the hippocampal CA1, DG, CA3 and again DG regions (Fig. 3.5B). During this process, spikes appeared and disappeared across recording channels. When the electrode reached the thalamic nuclei (approximately 5 mm from the brain surface), spikes appeared on all 16 channels (Fig. 3.5C). This occurs because of the widespread distribution of neuronal cell bodies within the nuclei. A similar all-channel spike recording occurred when the electrode first entered the cerebral cortex beneath the dura. However, if cortical neurons had been damaged during dura piercing before electrode insertion, the recorded spikes in the cerebral cortex would be reduced or absent.

NeuroNexus array microelectrodes are extremely thin, typically 15 μm in thickness, causing minimal damage to brain tissue. These electrodes can be inserted and removed from brain tissue repeatedly while still recording spikes from the same group of neurons (Csicsvari et al. 2003). When a recording contact is close to neuronal soma, the amplitude of recorded spikes can reach hundreds of microvolts to even over one millivolt. Given that high-frequency noise in recording signals is typically less than 20 μV in amplitude, this recording quality can reach a high signal-to-noise ratio (SNR). Additionally, as the electrode advances gradually, mechanical forces can stimulate adjacent neurons, increasing their firing rates and producing dense spikes. These large, dense spikes migrate from one channel to the next as the electrode moves forward slowly. The spike migrations can indicate the survival of nearby neurons. However, damage can occur if the electrode surface is rough and exerts excessive pulling force on nearby neurons. Similarly, electrode vibration can also kill nearby neurons. In these cases, recorded spikes may gradually disappear as they migrate along channels. Importantly, spike disappearances do not always indicate neuron damage. They may result from increased distance between the recording contact and the neuron soma, or from other factors such as the neuron entering a temporary quiescent state.

NeuroNexus array electrodes are reusable, but their surfaces can become rough after multiple uses due to accumulated deposits. This roughness prevents the recording contacts from adhering closely to neurons, resulting in lower-quality signal recodings and poor signal-to-noise ratios.

In addition to determining electrode position through spontaneous firing recordings, we can verify the positions of both stimulation and recording electrodes in real-time using evoked-potentials triggered by electrical pulses (Kloosterman et al. 2001; Csicsvari et al. 2003). As shown in Fig. 3.6A, in the hippocampal CA1 region, a pulse applied to the Schaffer collaterals can activate the downstream post-synaptic pyramidal neurons, while a pulse applied to the alveus can antidromically activate the pyramidal neurons somata. This setup allows recording both orthodromic and antidromic evoked-potentials from the same neuronal population using a single recording electrode.

Full size image

As shown in Fig. 3.6B, the 16-channel recording electrode detected spontaneous spikes concentrated in the channels 6–11, which indicated their locations in the CA1 soma layer. A pulse (0.3 mA) to the Schaffer collaterals produced distinctive fEPSP waveforms in the apical dendritic layer (channels 14–16) and OPS waveforms in the soma layer (channels 6–8), respectively (Fig. 3.6C). The OPS latency was approximately 5.5 ms. Also, a pulse to the alveus induced APS waveforms in the soma layer (Fig. 3.6D). The APS latency was only ~ 1.5 ms, much shorter than the OPS latency. Thus, the locations of recording channels were further confirmed by these evoked-potentials.

Synchronized activation of a neuronal population can create unique evoked waveforms in the extracellular space of various CA1 layers through ion fluxes (refer to Fig. 2.5). These waveforms can provide real-time guidelines for accurately placing recording and stimulation electrodes. While the amplitude of evoked potentials (reflecting the number of activated neurons) can vary with pulse intensity (see Figs. 2.7 and 2.8), their basic waveforms remain consistent. Improper electrode positioning can distort these waveforms, indicating the need to adjust electrode positions.

After craniotomy surgery, we usually began the experiment by inserting the recording electrode first. As the electrode advanced, we determined its entry into the hippocampal CA1 region by monitoring both the insertion depth and the sequential appearance of unit spikes across channels. Using an array electrode with appropriately spaced channels, the determination was highly reliable. We then inserted the stimulation electrodes. The bipolar stimulation electrodes we used can produce a small activation area precisely confined to the target (detailed in Sect. 3.4.2). Large-amplitude evoked waveforms can appear only when the stimulation effectively activates neurons in the recording area. Therefore, the stimulation sites needed precise alignment with the recording area that had been fixed by the position of recording electrode. During the insertion, we applied test pulses to track the position of stimulation electrode and verify its proper placement.

The alignment requirements for orthodromic and antidromic stimulations differ somewhat (López-Aguado et al. 2002). In the Schaffer collaterals projecting to the recording area, each axon terminal has multiple branches with synapses covering a broader area. Therefore, orthodromic stimulation of the Schaffer collaterals can activate neurons across a larger area in the downstream CA1 region. This means that the alignment between the orthodromic stimulation and recording area needs less precision.

When inserting the orthodromic stimulation electrode from the dorsal hippocampus, it passes through the alveus, stratum oriens (so), and pyramidal cell layer (pcl) before reaching the stratum radiatum (sr) where the Schaffer collaterals are located (Fig. 2.5A). When the stimulation electrode is in the so layer, an applied pulse can activate the axonal fibers there, specifically the commissural fibers (CF) from the contralateral hippocampal CA3 pyramidal neurons (Fig. 2.13A, B). This activation can also evoke OPS in the CA1 pcl layer. Do not mistake this for activation from Schaffer collaterals. The electrode position can be distinguished by observing the polarity of fEPSP waveforms in the so and sr layers. When stimulating the CF fibers in the so layer, the fEPSP is negative in that layer, while a positive fEPSP appears in the sr layer due to the current return loop. As the stimulation electrode advances to the sr layer to activate the Schaffer collaterals, these polarities reverse—the fEPSP becomes negative in the sr layer and positive in the so layer (Richardson et al. 1987). The polarity flip can serve as a marker for the electrode movement among layers. Additionally, the apical dendrites of pyramidal neurons in the sr layer have an extensive longitudinal distribution, enabling simultaneous fEPSP recordings across multiple channels (Fig. 3.6C).

In antidromic stimulation, when the stimulation electrode advances downward from the cerebral cortex to the hippocampus, it reaches the alveus on hippocampal dorsal surface without deep entry. Unlike orthodromic activation which spreads widely at axon terminals, antidromic stimulation can activate just one pyramidal neuron soma per axon, resulting in a smaller activation area (López-Aguado et al. 2002). Therefore, the alignment between the antidromic stimulation site and the recording site needs more precise.

In addition, stimulation electrodes can also record neural signals. Monitoring unit spikes collected at their exposed tips while advancing in the brain can help determine their positioning. Once the electrode is near its target site, switch it to stimulation mode.

After the experiment ends, the rat brain can be extracted for slice staining to verify electrode positioning. Alternatively, we developed a simple assessment method: record the coordinate data on the manipulators for each electrode, raise all electrodes from the brain, and carefully remove the rat from the stereotaxic apparatus. Then return the electrodes to their original coordinates in the brain. This can allow examination of the relative positions among electrode tips to verify placement. This method is particularly helpful for beginners analyzing the recorded evoked-potentials and spontaneous spikes.

Figure 3.7A shows a side view photograph taken during an experiment. Three electrodes protruded above the left skull, including the recording electrode (RE) and two stimulation electrodes—the orthodromic stimulation electrode (OSE) and the antidromic stimulation electrode (ASE). The RE was inserted vertically into the hippocampal CA1 region. The OSE, targeting the Schaffer collaterals, was inserted at a 30° angle to the vertical (from front to back). The ASE, targeting the alveus, was inserted at a 20° angle to the vertical (from left to right). Figure 3.7B shows a posterior view photograph taken after the experiment, with electrode positions restored after rat removal. Figure 3.7C is a side view photograph with a superimposed hippocampus schematic. These photos have some distortion due to the shooting angles. On-site examination of the relative positions between electrodes, combined with analysis of evoked-potential waveforms from orthodromic and antidromic stimulations (Fig. 3.7D), can help assess electrode positioning.

Full size image

Figure 3.7D shows the recording signals from eight channels on the RE array with a 50-μm spacing. The middle channels were in the CA1 soma layer, recording OPS and APS. Upper and lower channels were in the basal and apical dendrites, respectively, with apical dendrite channels recording fEPSPs. The ASE, possibly positioned slightly deeper (see Fig. 3.7C) or aligned with the RE less precisely, produced relatively small APS waveforms. The OSE, located in the apical dendritic layer, produced large fEPSP and OPS waveforms with accurate position and alignment. Both OSE and ASE were FHC concentric bipolar electrodes with a tip length of 300 μm, an outer diameter of 250 μm, and an inner diameter of 75 μm (see Sect. 3.4.2). The tips of these electrodes in Fig. 3.7C are faintly visible. The RE tip, containing the contacts, was only about 30 μm in width—almost invisible in this photograph. At just 15 μm thick, the entire RE shank appears as a thin line in the well-focused photograph in Fig. 3.7B.

We performed neural signal recordings in rat brains using Michigan style microarray electrodes produced by NeuroNexus company (website: http://neuronexus.com/). These electrodes feature one or more thin-film silicon shanks with contacts coated in iridium or gold to enhance conductivity and biocompatibility. Their conductive traces are shielded by insulating layers. Standard Michigan electrodes are only 15 μm thick, minimizing tissue damage during insertion. However, these thin-film electrodes are extremely fragile and prone to breakage. Therefore, the shanks that penetrate the brain tissue typically measure only a few millimeters in length (3 or 5 mm, Fig. 3.11A). For deeper brain regions, electrodes with longer insertion shanks (e.g., 10 mm) are available, but these have an increased thickness of 50 μm.

Full size image

NeuroNexus electrodes have two types of connectors (Fig. 3.11A), designed for acute and chronic experiments respectively. Acute experiments typically involve short-term recordings in anesthetized animals. With proper maintenance, acute electrodes can be reused while still yielding high-quality signals. Chronic experiments usually involve long-term recording over days or months, requiring the electrodes to be fixed to the animal skull with dental cement (Fig. 3.11B). These chronically implanted electrodes typically cannot be reused.

NeuroNexus electrodes offer a variety of contact configurations. For examples, the basic linear design has contacts evenly spaced along a shank (Fig. 3.12A). The tetrode-style configuration arranges contacts in diamond-shape groups of four (Fig. 3.12B), imitating traditional wire-made tetrodes. The “Poly2” design features two columns of contacts in a staggered pattern, creating a two-dimensional array within a single shank (Fig. 3.12C). NeuroNexus also provides customized configurations for specific needs (Fig. 3.12D). Notably, placing contacts along the shank edges can bring them closer to neurons, resulting in better signal-to-noise ratios in recordings.

Full size image

These array electrodes are available in diverse specifications, varying in their shank and contact numbers, areas, spacing, and distribution patterns. Electrodes with a single contact column are one-dimensional (Fig. 3.12A), while those with multiple shanks or multiple columns on a single shank are two-dimensional (Fig. 3.12B, C). Several two-dimensional electrodes combined together can form three-dimensional recordings. The contact number (also known as channel number) includes 4, 16, 32, 64, 128, and beyond. Basic electrodes have 4 to 16 contacts on one shank (single column) and can reach up to 64 contacts (multiple columns). Contact spacing ranges from 25 to 200 μm, and shank spacing varies from 125 to 500 μm. Contact area ranges from hundreds to thousands of square micrometers. Customized specifications can be made to meet specific research requirements.

Neurons close to the electrode surface may be damaged by friction during electrode advancement. While neurons near the electrode tip or edges can remain largely unaffected, those along the electrode shank may be damaged by the electrode movement. Early electrodes had thicker and rougher substrates and coatings, which reduced the survival probability of nearby neurons. This made it difficult to record high-quality unit spikes from individual neurons, limiting these electrodes mainly to recording LFPs. Thanks to advances in manufacturing technology, modern electrodes can effectively record unit spikes. In our experiments, unit spikes recorded from rat hippocampal neurons occasionally reached amplitudes of up to 1 mV, indicating a close attachment of the neuronal soma to the contact.

A contact with a small area can be in close proximity to only one or two neuronal somata, resulting in just one or two types of large-amplitude spikes in a multiple unit activity (MUA) recording. This can facilitate spike detection and sorting (Buzsáki 2004; Blanche et al. 2005). In contrast, a large contact can collect unit spikes from many surrounding neurons, resulting in overlapped spikes that are difficult to identify. Therefore, electrodes with a larger contact area are suitable for recording LFPs, while those with a smaller area are better for recording unit spikes. Electrodes with a moderate contact area can accommodate recordings of both types of signals. For example, our commonly used electrode type (A1 × 16-Poly2-5mm-50s-177), abbreviated as Poly2, has a contact area of 177μm² (Fig. 3.12C) and can record both signals effectively. In comparison, another type of four-shank electrode (A4 × 4-4mm-200-200-1250) has a large contact area of 1250 μm² and can simultaneously collect overlapped spikes from many neurons, making spike identification and classification difficult. However, reducing the contact area may increase electrode impedance, leading to more noises (Bretschneider and de Weille 2006). This issue can be resolved through electrode surface modification. NeuroNexus uses iridium oxide thin film deposition for surface modification which can activate the contact surface to reduce electrode impedance.

NeuroNexus array electrodes with sufficiently large contact areas can also function as stimulation electrodes. Moreover, on a same electrode, each contact can be designated for stimulation, recording, or both—switching between the two functions as needed.

Besides the NeuroNexus electrodes originally from the University of Michigan, the Utah electrodes, developed by the University of Utah, are another prominent type of implantable array microelectrodes (Campbell et al. 1991). These Utah electrodes, also known as needle arrays, consist of multiple fine needles protruding on a silicon substrate. A typical electrode contains 100 needles arranged in a 4 × 4 mm area with 400 μm spacing. Each needle is insulated except for its platinum-coated tip, which can collect electrical signals. The needles are sturdy and resistant to breakage. In 2006, the Utah electrode made history as the first array electrode implanted in the human brain, enabling a brain-computer interface for “thought controlling machines” (Hochberg et al. 2006). However, due to the limited length of needles, this type of electrodes is primarily used for recordings from superficial brain regions and for peripheral nerve stimulations (Branner et al. 2001).

In summary, these array electrodes are implantable devices that can be used to detect extracellular signals or deliver electrical stimulations. They can simultaneously collect electrical signals from populations of neurons, including unit spikes, LFPs, and stimulus-evoked potentials. This makes them ideal tools for studying brain mechanisms, developing neural regeneration technologies, and advancing brain-computer interfaces (Buzsáki 2004; Blanche et al. 2005; Erofeev et al. 2022).

We used concentric bipolar electrodes from the American company FHC (website: http://www.fh-co.com/) for precise stimulations in the rat hippocampal region, typically using the model “CBCSG75” (Fig. 3.13A). In the model name “CBCSG75”, “CB” stands for concentric bipolar design, “C” represents the 250 μm diameter of outer pole—stainless steel tube, “S” represents a three-segment nested structure (inner pole, insulation, and outer pole)—each segment 100 μm long (Fig. 3.13B), and “G” represents the 75 μm diameter of inner pole—made of platinum-iridium alloy (Pt/Ir). The final “75” represents the total electrode length of 75 mm.

Full size image

When applying electrical stimulation through such an electrode using a stimulator with isolator (as those introduced in Sect. 3.5.4), the stimulation current flows from the inner pole, travels about 100 μm, and then enters the outer pole. This creates an electric field confined to a small region near the electrode tip. Such localized field can enable precise stimulation of brain microstructures while minimizing interference with nearby signal recordings. It can also prevent amplifier saturation that may be caused by strong or sustained stimulation currents. Additionally, the surface area of outer pole is about three times larger than the inner pole, resulting in a much stronger electric field around the inner pole. Therefore, the inner pole acts as the working electrode, while the outer pole serves as the return electrode for the current loop.

Stimulation safety should be assessed when applying electrical stimulations through an electrode (refer to Sect. 1.4.4). It relates to multiple factors, with Faradaic reaction posing a primary risk. This reaction can damage surrounding tissue and corrode the electrode. To ensure accurate experimental results, even in animal studies, it is better to keep the stimulation current below the Faradaic reaction threshold. For current pulse stimulation, safety can be assessed using an index k, which is calculated from the charges delivered per pulse and the electrode surface area. A stimulation is considered safe when k falls below 1.5–2.0 (Shannon 1992; Cogan et al. 2016; Gilbert et al. 2023). The k is calculated byHere the charge Q (μC/pulse-phase) is the product of pulse current and pulse width. The charge density D (μC/cm²/pulse-phase) is the Q divided by the electrode surface area. The inner pole of CBCSG75 electrode has a surface area of 0.00028cm² (Fig. 3.13B). Appling a current pulse with a width of 100 μs and intensity of 0.3 mA results in k = 0.47, well below the lower safe limit of 1.5. Keeping this pulse width and increasing the pulse intensity to 1.6 mA yields k = 2.0 (the upper safe limit). In our rat experiments, we typically used a 100 μs pulse with an intensity less than 0.6 mA. These currents applied by such an electrode are safe even with monophasic pulse stimulations. Let alone that we typically employed charge-balanced biphasic pulses, which further enhanced stimulation safety. Unless otherwise specified, the stimulation pulses used in this book are all biphasic current-pulses with a negative phase followed by a charge-balanced positive phase, each 100 μs wide.

Reducing electrode surface area can significantly affect safety. Figure 3.14 illustrates another type of FHC concentric bipolar electrodes (CBBRC75) with a smaller inner pole of only 0.0000049cm² in surface area. For a current pulse with a width of 100 μs and an intensity of 0.3 mA, k = 2.26, exceeding the upper safe limit of 2.0. We previously used this type of electrodes to apply continuous high-frequency pulse stimulation at 200 Hz for 1 min in the rat hippocampus to investigate whether the stimulation would cause neuronal damage (Yuan et al. 2021).

Full size image

Figure 3.15A shows the experiment setup, with the stimulation electrodes (OSE and ASE) and the recording electrode (RE) positioned in the rat hippocampal CA1 region. Both stimulation electrodes were CBBRC75. The OSE delivered orthodromic high-frequency stimulation (O-HFS) and single-pulse orthodromic test stimulation (OTS) to the Schaffer collaterals. To assess the state of downstream CA1 pyramidal neurons during O-HFS, the ASE was placed at their axons (alveus) to apply antidromic test stimulation (ATS)—a single biphasic pulse every 5 s. All the current pulses had a width of 100 μs per phase and an intensity of 0.3 mA.

Full size image

Figure 3.15B shows a recording from the CA1 pyramidal cell layer during 1-min 200 Hz O-HFS of biphasic pulses. The first pulse of O-HFS triggered an OPS similar to the baseline recording. Subsequently, due to local inhibitory circuits (see Sect. 2.3), the OPS disappeared. However, when an ATS was applied to the alveus at 20 ms from O-HFS onset, it still produced an APS (see the enlarged inset at bottom left of Fig. 3.15B). As the effect of inhibitory circuits diminished, PS reappeared. The ATS applied at about 5 s from O-HFS onset produced an APS along with other PSs. (Note: here “PS” refers to a population spike which lacked regular latency with a preceding pulse.) Later, HFS-induced axonal block (refer to Chap. 5) prevented O-HFS pulses from producing more PSs, but the ATS applied every 5 s was consistently able to produce a large APS. After the end of O-HFS, OTS-induced OPS gradually recovered. The results indicate that even when the k value exceeded the safety limit, sustained HFS with charge-balanced biphasic-pulses would not cause neuron damage.

However, results differed when O-HFS was switched to monophasic negative pulses while maintaining other stimulation parameters (Fig. 3.15C). The neuronal response during the initial O-HFS period was similar to that of biphasic-pulse O-HFS. However, after about 10 s O-HFS, a spreading depression (SD) event occurred, characterized by dense PSs followed by a slow wave in the LFP, consistent with previous reports (Herreras and Makarova 2020; Bragin et al. 1997a, b; Herreras et al. 1994). Following the SD, during ongoing O-HFS, ATS was no longer able to produce APS. Even after O-HFS ended, it took several minutes for APS to recover gradually, while OTS-induced OPS failed to recover. This indicated that the CA1 neurons kept their ability to respond to external activation, while the directedly stimulated afferent fiber (Schaffer collaterals) was damaged and failed to activate. Additionally, during the period following SD, unit spikes disappeared (forming a “silent period”) and only partially recovered after O-HFS ended (see MUA signal at the bottom of Fig. 3.15C). These results indicate that when the stimulation electrode surface area is too small, causing excessively high charge density to raise the k value beyond the safety limit, continuous HFS of monophasic-pulses can damage neural tissues such as the Schaffer collateral axons.

The hippocampal region is one of brain areas where SD readily occurs. A key feature of SD is its slow propagation. Figure 3.16A shows SD waves recorded from the CA1 soma and apical dendritic layers (pcl and sr) by two channels 150 μm apart. The SD waves propagated from the sr layer to the pcl layer at a very slow speed. Besides electrical stimulations, other inducements can also trigger SD events, such as high extracellular K⁺ concentration and intense epileptiform discharges. Figure 3.16B shows an SD event induced by locally injecting a high-concentration potassium chloride solution into the hippocampus. The two recording channels were still 150 μm apart. Dense firing preceded the SD waves. The slow SD propagation can be determined through the time delay between the SD troughs recorded in the pcl and sr layers.

Full size image

The SD wave in field potential is caused by massive neuronal depolarization. Aristides Leão first discovered SD in electroencephalogram (EEG) recording in the 1940s (Leao 1944; Bures 1999), when he observed it as a period of EEG suppression following epileptic seizures. SD is also associated with conditions such as migraine, ischemic brain injury, and cerebrovascular disorders (Gorji 2001; Lauritzen 1994; Parsons 2004; Kunkler and Kraig 2003). A key phenomenon accompanying SD events is the dramatic changes in extracellular ion concentrations (including K⁺, Ca²⁺, Na⁺, and Cl⁻). These changes disrupt the normal ion concentration gradients between intracellular and extracellular environments, causing morphological changes in neurons, such as toxic swelling (Oka et al. 2022; Martins-Ferreira et al. 2000; Kraig and Nicholson 1978; Krnjević et al. 1980; Somjen and Giacchino 1985).

Although neural damage can trigger SD, its occurrence doesn't necessarily mean damage. Neuronal activity may return to normal after SD, especially when induced by high extracellular potassium. However, a failure of neuronal activity recovery long after SD can signify neural damage, as shown in Fig. 3.15C. Therefore, the appearance of SD can serve as a warning sign for tissue damage. SD in the hippocampal region is often accompanied by dense epileptiform discharges before its slow wave (Figs. 3.15 and 3.16). Even after reversible SD, normal neuronal activity may take tens of minutes to recover. Additionally, a “silent period” without neuronal firing can typically occur immediately following SD waves. During this period, the field potential becomes almost flat (Fig. 3.15C). If a similar silent period suddenly appears during an experiment and the rat is still alive, you can replay the recorded signal to check for possible SD occurrence before inspecting any setup failures. If SD has occurred, the field potential and spikes may reappear over time. This situation may occur when an experiment procedure, such as electrode advancement during implantation, excessively damages neural tissue.

Note that the actual SD wave contains a negative DC shift (Herreras et al. 1994), which was filtered out by the AC coupling inputs used in our amplifier setup. As a result, the SD wave typically appeared as an alternating waveform—first a negative phase followed by a positive phase (Fig. 3.16), which was consistent with other reports (Bragin et al. 1997b).

Neural electrical signals are very weak and need amplification before being converted from analog to digital. Our lab used general electrophysiological amplifier and data acquisition devices rather than specialized neural equipment. As shown in Fig. 3.17, we used a Model 3600 16-channel extracellular amplifier with headstage (preamplifier) from A-M Systems Inc. For data acquisition, we used a PL3516 PowerLab data recorder from ADInstruments Inc. (Australia). The multi-channel neural signals collected by an electrode first passed through the headstage into the amplifier, where they underwent a 100-fold amplification. They then entered the PowerLab recorder for additional amplification and analog-to-digital conversion (ADC), usually at a 20 kHz sampling rate. Because the PowerLab recorder had its own amplification, setting the preceding amplifier to a 100-fold gain was sufficient to record sub-millivolt and smaller neural signals.

The Model 3600 amplifier features low-noise and high-gain properties, designed for amplifying microvolt electrical signals (Fig. 3.18A). It can be equipped with two types of headstages (Fig. 3.18B). One is the miniature headstage used solely for recording (the small square device in the figure). The other is the larger, dual-function headstage capable of both recording and stimulation (the triangular device in the figure). The dual-function headstage offers versatility through proper wiring and settings, allowing designated channels to switch between recording and stimulation modes, enabling a single electrode contact to both collect signals and deliver stimuli. When a channel switches to stimulation mode, its corresponding contact disconnects from the recording circuit and connects to the pre-wired stimulator. This disconnection prevents stimulation signals from entering the recording circuit, protecting the amplifier from saturation and damage since stimulation signals are much larger than neural signals.

Full size image

The front panel of the 3600 amplifier features a touchscreen (Fig. 3.18A), enabling settings for each channel, including high-pass and low-pass cutoff frequencies, gain, notch filter and recording mode. A-M Systems also offers the Model 1700—a four-channel amplifier with similar performance and functionality to the Model 3600, but with a traditional panel of knobs and switches instead of a touchscreen. For applications requiring fewer signal channels, the Model 1700 is a suitable alternative.

Electrical devices used in electrophysiological experiments typically require a separate ground (or “clean ground”). Do not share their ground with lighting, refrigerators, or other electrical equipment. Appropriate shielding measures are also necessary. A copper plate on the experiment table can improve shielding. Our experiment table, measuring 60 cm wide and 150 cm long, was covered with a 2 mm thick copper plate connected to the experimental ground. Ample space surrounded the table, allowing operators to work from various directions. Some equipment, like amplifiers and stimulators, occupied half the table, while the other half accommodated rat stereotaxic instruments. During experiments, we covered the stereotaxic system with a shielding cage (or Faraday cage) with an open-bottom design, made of an aluminum frame and copper mesh. The aluminum bars along the cage bottom edges contacted the copper plate on the table, providing shield grounding. Notably, only when a cage is grounded can it provide effective shielding. The copper plate can offer an additional function: operators can discharge electrostatic charges from their bodies by touching it, preventing transfers to animal and devices. This function is particularly important in cold and dry winters when static electricity easily accumulates on bodies and clothing.

Our shielding cage was custom-built using L-shape aluminum bars to form a trapezoidal cuboid frame—slightly narrower at the top than at the bottom. Copper mesh covered all sides except the open bottom. One side was designed as a liftable curtain, facilitating observations and operations during experiments. Even with this curtain raised, we still achieved excellent signal-to-noise ratios in our recordings, without significant external interference. The interior space in the cage allows for simple experimental operations. For complex procedures, such as craniotomy and electrode replacements, the cage can be completely removed for easier access.

It is important to be aware of the technical specifications of an amplifier before using it. For example, the Model 3600 amplifier has a maximum input voltage of only 1–2 V (determined by the headstages shown in Fig. 3.18B). Typical electrophysiological signals fall within this range. However, when input signals exceed its linear amplification range, the amplifier can saturate. Similarly, if the amplifier gain is set inappropriately, signals can be clipped due to saturation.

External noise and electrical stimulations can cause amplifier saturation. When applying electrical stimulations to a brain region being recorded, large stimulation signals may leak into the amplification circuit, causing saturation which can persist long after the stimulation ends. This amplifier malfunction can disrupt normal signal recording for minutes or even hours. More seriously, high-voltage inputs, such as a electrostatic voltage, may completely destroy an amplifier. Therefore, special care should be taken to prevent electrostatic buildup during experiment, especially in cold, dry winters when static electricity can easily accumulate and reach thousands of volts momentarily. Wearing a grounded wrist strap and keeping appropriate humidity can help reduce electrostatic risk. Before handling devices like electrode holders and manipulators, touch your free hand to a grounded copper plate to discharge any potential static electricity. Another protective approach is to turn off amplifier inputs during operations. The Model 3600 amplifier offers “ON” and “OFF” options for each channel—“OFF” disconnects the channel input. The Model 1700 amplifier has “Record” and “Stimulate” options—switching to “Stimulate” disconnects the amplification circuit. These measures can prevent amplifier damage when handling animals or adjusting electrodes during experiments. If you need to monitor recording signals during operations (when the amplification circuit must remain connected), take extra precautions against electrostatic interference. Additionally, improper connections of ground and reference wires, or their unreliable connections—such as broken wires or poor contact—can also cause amplifier saturation.

When amplifier saturation occurs, the recording signal can become a flat line or its amplitude may drop significantly. In this case, immediately disconnect the input signals or turn off the amplifier power, then check all instrument connections for troubleshooting. Be aware that severe saturation may cause unstable, low-frequency, high-amplitude drifts in recordings during the recovery period after turning the amplifier back on. Do not mistake these artifacts for actual field potentials from the brain.

A known input signal is essential for checking a newly established or updated amplification setup and its settings, or for verifying proper function after troubleshooting. Since electrophysiological amplifiers have a limited input range, using arbitrary signals could cause amplifier saturation or damage. Model 3600 amplifier features a “Calibration Signal” output on its front panel (in the right of Fig. 3.18A), which can generate a 1.0 kHz sinusoidal signal with four amplitude options: 1, 10, 100, and 1000 mV. This calibration signal provides a safe source for testing amplifier wiring. For example, after setting up the signal amplification and acquisition system as shown in Fig. 3.17, connect this calibration signal to each socket of the electrode connector sequentially. By examining the frequency and amplitude of the recorded signals in the LabChart display window—the software coming with the PowerLab recorder (refer to Sect. 3.5.3), you can verify the recording of each channel. This process validates the entire signal path from the electrode connector through amplification to sampling and recording.

Electrophysiological signals, including neural signals, are classified as weak electrical signals. Their measurements are susceptible to various types of noise and interference, including 50 Hz mains interference and pulse interference from switching electrical equipment. Using differential amplification can help eliminate interferences by recording the potential differences between the measurement and reference points, effectively removing common-mode noises.

Additionally, proper grounding and shielding are crucial for eliminating interference. The experiment setup should use an independent ground, separate from lighting and non-experiment electrical equipment. For effective shielding, place the experimental animal, along with its connecting devices and micromanipulators, inside a shielding cage. All connecting wires also need appropriate shielding. For example, use shielded cables to connect the stimulator outputs and the stimulation electrodes implanted in the rat brain. It is important to ground these shields to prevent external noise from entering the rat brain and interfering with the recorded signals. The following examples show various interferences we once encountered in our recordings in rat brains.

Mains interference in a signal is easy to identify. As shown in Fig. 3.19, a thickened signal (top left) indicates potential interference. Zooming in on the time axis can reveal the superimposed sine waves clearly (bottom). You can determine mains interference by measuring the sine frequency—typically 50 Hz, or 60 Hz in North America and some other countries.

Full size image

Many amplifiers feature a notch filter that can remove mains interference, as in the Model 3600 and 1700 multi-channel amplifiers. However, when the interference frequency overlaps with the signal frequency, filtering out the interference frequency can also cause signal loss. Furthermore, real filters are not perfect—a 50 Hz notch filter can weaken signals across a frequency range around 50 Hz, leading to more loss. For some neural signals, such as EEG and LFP, frequencies around 50 Hz are often not negligible. Therefore, careful consideration is necessary when using a notch filter to remove mains interference. Action potentials from individual neurons (unit spikes) typically have much higher frequencies—above 300 Hz—and high-pass filtering can fully eliminate mains interference in spike signals. Generally, proper grounding and shielding can suppress mains interference to an acceptable level without filtering.

Switching electrical equipment on and off or rotating electrical knobs can cause high-frequency noises. In recorded signals ranging from 500 Hz to 5 kHz, these noises may resemble unit spikes (Fig. 3.20). However, they can be readily identified in multi-channel recordings. In Fig. 3.20, the first two channels (Ch1 and Ch2) were from the right hippocampal CA3 region, with the two recording contacts only 25 μm apart—allowing simultaneous recordings of spikes from same neurons. The third channel (Ch3) recorded from the left hippocampal CA3 region, making it impossible to detect spikes from the neurons measured in the other two channels. Yet, high-frequency noises appeared simultaneously on all the three channels. Thus, spike-like signals occurring simultaneously at distant contacts can be noises, rather than real spikes.

Full size image

Today, cell phones have become inseparable from our daily lives. Cell phone interference can be picked up by neural signal recording setups. The best way to prevent this is to keep cell phones away from the experimental setup. Interference can become especially severe when a phone is dialing or about to ring. When placed nearby, phones can introduce significant noise into the recorded signals.

Figure 3.21 shows the noises introduced by cell phone ringing in a recording from the rat hippocampal CA1 region. In the original signal recorded with a frequency band of 0.3 Hz to 5 kHz, the noises appeared as pulse trains in the enlarged inset (upper), with a pulse width of about 0.7 ms and a frequency of ~ 200 Hz. In the high-pass filtered signal (MUA signal) with a 500 Hz cutoff frequency, these noises roughly resembled neuronal spikes. However, the enlarged view showed clear differences between the noises and real spikes (bottom).

Full size image

A powered-on cell phone placed near the experiment setup can generate noises even without incoming calls or conversations. To minimize their interference, keep cell phones away from the experiment region. The signal shown in Fig. 3.21 was recorded around 2010. As cell phone technology has advanced, the characteristics of the interference may have also changed.

In an acute in-vivo rat experiment, the rat head was fixed in a stereotaxic apparatus (Fig. 3.7A). When recording neural signals from its brain, subtle vibrations from the head, such as mouth or face trembling, can cause interference that presents as high-frequency, cluster-like noises in recordings (Fig. 3.22A). Such noises can disrupt the recording of unit spikes in the 500 Hz to 5 kHz frequency band, since both frequency ranges overlapped. An enlarged view shows that these noises appeared as continuous oscillatory waves (Fig. 3.22B) with similar waveforms and amplitudes across adjacent recording channels. In contrast, spikes were separate waveforms with significant amplitude variations across different recording channels (Fig. 3.22C). If this type of noises is suspected during experiment, checking the rat facial movements can help confirm its presence.

Full size image

Careful attention to experimental procedures can help minimize the impact of rat facial movements on neuronal spike recordings. Forceful operations with a microdrill or scalpel during craniotomy may loosen the rat fixations. Therefore, after completing the craniotomy and before placing electrodes, carefully re-examine and reinforce all fixations, including ear bars, nose clamp, and incisor bar. This step can effectively reduce or eliminate the “cluster” noise introduced by rat facial movements during signal recordings. As the anesthetic wears off over long experiments, additional anesthetic may be needed to stop rat movements. Once the rat quiets down, the noises will disappear. Thus, the appearance of these noises can also serve as an indicator for anesthesia depth. In chronic experiments without anesthesia, electrodes are fixed to the rat skull with dental cement. When recording signals from a freely moving rat (Fig. 3.11B), its actions such as chewing food or scratching face and head may introduce noises in the recordings similar to those shown in Fig. 3.22. Do not mistake the noises as neural signals.

When recording multi-channel electrical signals, malfunctions in signal wires can lead to “crosstalk”. For instance, we once encountered this issue while using an array electrode to record unit spikes in the rat hippocampal CA1 region.

As shown in Fig. 3.23, we used a NeuroNexus acute electrode (model A4 × 4-4mm-200-200-1250) with four recording shanks, each featuring four linearly arranged contacts. Figure 3.23A shows the contacts on the first shank, numbered from top to bottom as channels 1–4 (Ch1–Ch4). The contacts were spaced 200 μm apart, both between contacts on the same shank and between shanks. This substantial spacing should have prevented different contacts from recording firing (unit spikes) from the same neurons simultaneously (Buzsáki 2004). However, the recorded signals showed that the spikes in Ch1 consistently resembled those in Ch3, although with reduced amplitudes. This was puzzling given the 400 μm distance between the two channels. According to the orthodromic and antidromic evoked potentials in the hippocampal CA1 region (see Sect. 3.2.3), we determined that Ch3 was positioned in the soma layer, recording spikes from nearby neurons. Then, Ch1, located 400 μm above, was likely beyond the CA1 region. Where were the Ch1 spikes from? Why didn’t Ch2—located between Ch1 and Ch3—record the same spikes (see the enlarged insets in Fig. 3.23A bottom)? How could the spikes bypass the closer contact (Ch2) to reach the more distant one (Ch1)?

Full size image

We numbered the recording channels based on the contact layouts on the four shanks, which differed from the pin numbering on the connector at the electrode tail (Fig. 3.23B, bottom). After the experiment, we checked the electrode carefully. According to the site map provided by manufacturer, Ch1 and Ch3 were connected to adjacent pins 1 and 2 on the connector, respectively (the red numbers in Fig. 3.23B). We discovered a broken PCB trace connecting the Ch1 contact to its pin—the pin 1 solder pad. As a result, the pin 1 was left floating. When connected to the amplifier, this floating pin pad acted like an antenna, picking up signals from the adjacent pad of pin 2 through electromagnetic coupling, thereby producing a crosstalk signal. This explained the similarity in spike recordings between Ch1 and Ch3. The broken trace was not a manufacturing defect but resulted from repeatedly applying tape for fixation. After this incident, we no longer applied tape directly to the electrode PCB.

Crosstalk occurs through electromagnetic coupling and can only appear with high-frequency alternating signals. The faster a signal changes, the more crosstalk it can generate. Direct currents do not generate crosstalk. Since neuronal spikes contain high-frequency components and change rapidly, they can produce significant crosstalk. This is evident in the MUA signals shown in Fig. 3.23A, which fall within the 500–5000 Hz frequency band. Similarly, pulse-evoked population spikes (OPS and APS) contain high-frequency components that can generate substantial crosstalk when experimental setups malfunction.

Crosstalk signals closely resemble normal signals, so be careful not to confuse these false signals with real neural signals. A crosstalk signal typically maintains a stable amplitude ratio to the original signal. To check for its presence, monitor the sequential shifting of spikes in recording channels while advancing a recording array into the brain. For example, as shown in the upper part of Fig. 3.23A, when the four-channel shank gradually enters the hippocampal CA1 region, Ch4 passes through the soma layer first and records spikes, followed by the Ch3, 2, and 1. Spike signals should appear on these channels sequentially in this order. However, when Ch1 was faulty as in this case, spikes appeared simultaneously on Ch3 and 1 when Ch3 entered the soma layer. Later, when the electrode was lowered to position Ch1 in the soma layer, no spikes were recorded because Ch3—the source of crosstalk—was already too far from the soma layer. Notably, this example shows that signal crosstalk can even occur between seemingly “unrelated” channels that are far apart, rather than between adjacent recording channels. This occurred only because the pin arrangement at the electrode connector differed from the channel (contact) arrangement at the electrode tip (Fig. 3.23B).

These above examples of interference represent just a few cases we encountered in our experiments. Noises are diverse and can originate from multiple sources, such as improper experimental procedures, amplifier equipment, and inadequate shielding measures. After establishing an experiment setup, simple tests can help identify and eliminate different types of noises, which is important for accurate neural signal recordings.

After amplification, neural signals undergo ADC and subsequent analysis and processing. As shown in Fig. 3.17, we used a PowerLab recorder for signal sampling. PowerLab recorders (http://www.adinstruments.com/) are widely used in research and teaching across fields like physiology, pharmacology, pathophysiology, and neuroscience. The recorder hardware mainly comprises a CPU microprocessor, memory, analog signal amplifiers, ADC, digital-to-analog converters (DAC) and other components.

As shown in Fig. 3.24A, we used the Model PL3516 recorder, which has 16 signal channels, each equipped with its own programmable amplifier. Through the LabChart software, these amplifiers can be configured independently for settings such as gain, filter bandwidth, input coupling mode (AC or DC), sampling frequency, and processing modes. The recorder automatically grounds the input amplifiers of idle channels. Once a gain setting is changed, the amplifier is grounded to measure DC offset, enabling the software to automatically correct any DC drift and offset. The recorder contains two 16-bit DACs that can output analog voltage signals via the Output1 and Output2 ports on the panel, which can be used to trigger stimulators and other instruments. Additionally, the recorder features 8-bit digital input and output ports.

Full size image

As shown in Fig. 3.24B, the recorder has LabChart software for powerful data display and analysis. We used LabChart Pro version 8, which consisted of core, module and extension components. The core components provide multi-channel recording, display, and analysis for up to 32 channels. The modules provide specialized tools such as peak analysis, spike histogram, ECG analysis, blood pressure analysis, and video capture. The extension component enables data export in various formats, including MATLAB and binary coding.

A notable advantage of LabChart is that it can be independently installed and run on computers not connected to PowerLab hardware. This is beneficial for offline analysis and processing of experimental data. Additionally, its Help menu is exceptionally comprehensive, offering both operational guidance and clear explanations of signal processing methods and principles. These features make LabChart an effective teaching tool. When using PowerLab and LabChart, consider the following points.

The amplification gain of the PowerLab recorder is software-controlled for each channel, with a maximum gain of 2000. The recorder can directly amplify and sample many types of signals without additional preamplifiers. Its gain is not presented as an amplification factor, but as an input range (or measurement range)—spanning from ± 2 mV to ± 10 V in 12 options. Users can set a desired gain by choosing an appropriate range.

The recorder features a 16-bit ADC with sampling resolution determined by the measurement range setting. For example, when the range is set to ± 10 V, the sampling resolution is 2 × 10 V/2¹⁶ = 305 μV. Table 3.1 lists the resolution and error for each range, based on the maximum value of a 16-bit binary number (approximately 64,000). Representing gains by ranges can help maximize ADC resolution. In the LabChart display window (Fig. 3.25), the areas outside the range are shaded, which allows users to easily assess amplification settings. The figure shows a signal recorded in rat hippocampal region during the application of a biphasic pulse train (0.4 mA, 200 Hz) with a measurement range of ± 2 V. The signal was displayed in two split windows with different time scales: a ~ 1.2 s segment in the right window and a ~ 6.5 ms enlarged segment in the left window which revealed clipped stimulus pulses within the ± 2 V range. Before entering the PowerLab recorder, this signal underwent 100-fold amplification, making an actual input range of ± 20 mV. This range setting was appropriate since we aimed to measure neural responses to the stimulation, treating the clipped pulse artifacts as negligible. While expanding the range would have captured the full pulse artifacts, it would have reduced the resolution of recorded neural signals. However, if valuable signals are lost due to insufficient range, immediate adjustment is necessary.

Full size image

Should we always choose the maximum possible range with enough margin to prevent signal clipping? The answer is no. A larger range can result in a lower ADC resolution and poorer accuracy (Table 3.1). When PowerLab converts analog signals to digital, the amplitude can only take on a finite number of discrete steps. A 16-bit ADC divides the measurement range into 2¹⁶ (65,536) steps. LabChart uses the middle 64,000 steps as the nominal range, treating the extremes as out-of-range values. For a ± 10 V range, the minimum resolvable voltage is 0.3125 mV. However, if the range is set to ± 10 mV, the minimum resolvable voltage becomes 0.3125 μV (Table 3.1)—a 1000-fold improvement in ADC resolution.

For example, the amplitude of normal LFP signals ranges approximately ± 1 mV. Under anesthesia, slow waves in LFPs can be slightly larger but typically not beyond ± 2 mV. With a pre-amplifier gain of 100-fold, the input signal to PowerLab recorder ranges within ± 0.2 V. A setting of this range results in an ADC resolution of 6.25 μV (Table 3.1). When simultaneously recording LFPs and unit spikes, small spikes—such as those with an amplitude of only ~ 50 μV (5 mV after 100-fold pre-amplification)—can occupy 800 ADC discrete steps (5 mV/6.25 μV = 800). This produces a sufficient sampling precision. However, when simultaneously recording large evoked population spikes (PS) with amplitudes > 10 mV, the PowerLab range must be set to ± 2 V, resulting in ~ 50 μV spikes occupying only 80 ADC steps—a lower precision.

As shown in Fig. 3.26A, a signal can be displayed as a step curve to reveal its discrete sampling data by selecting “Use Step Interpolation” from the dropdown menu in the LabChart window. In contrast, the conventional display uses “Linear Interpolation” mode, showing a smooth curve after processing (Fig. 3.26B). Note that the dropdown menu shows the alternative mode available for selection, opposite to the current display mode.

Full size image

In summary, the measurement range (i.e., amplification gain) must be set carefully. If the recorded signal approaches the shaded area and shows clipping in the LabChart display, the range must be increased. Conversely, if the signal consistently occupies only a small portion of the range, far from the shaded area, the range should be decreased to improve ADC resolution. For optimal ADC resolution, use the minimum possible range as long as signals of interest are not clipped.

The input signals to the PowerLab recorder cannot exceed ± 15 V in amplitude. Its input impedance of about 1MΩ (10⁶ Ω) and 100pF poses another limitation. When using microelectrodes to record electrophysiological signals—even metal wires with relatively low impedance for extracellular recordings—the electrode impedance is typically too high to connect directly to such a recorder. Therefore, signals from microelectrodes must first pass through a pre-amplifier with sufficiently high input impedance before connecting to the recorder. For example, the above-mentioned Model 3600 amplifier has an input impedance of about 10¹² Ω and 50pF, making it suitable for amplifying microelectrode signals.

The amplifier parameters of the PowerLab recorder can be set in LabChart. The cutoff frequency of its low-pass filter can be set between 1 Hz and 25 kHz. Its inputs have coupling options of direct current (DC) and alternating current (AC), with AC coupling having a low-frequency cutoff of 0.15 Hz. Its frequency response (− 3 dB frequency) reaches 25 kHz for the maximum input range of ± 10 V. Essentially, each channel's input connects a 25 kHz low-pass filter. In differential input mode (Fig. 3.24A), the common-mode rejection ratio (CMRR) at 100 Hz is 100 dB. The input crosstalk is 75 dB. The root mean square (RMS) of input noise is less than 350 µV at ± 10 V input range. The maximum sampling rate varies with the number of recording channels—400 kHz for 1–2 channels, decreasing to 20 kHz per channel for 9–16 channels.

The front panel of the PowerLab recorder features two analog output channels—Output1 and Output2 (Fig. 3.24A). They can function either as two independent single-ended outputs or together as a differential output. The recorder uses a 16-bit DAC with a maximum output current of ± 50 mA and an output impedance of about 0.5Ω. Through the LabChart software, users can set the output voltage range to one of six options: ± 10, ± 5, ± 2, ± 1, ± 0.5, and ± 0.2 V. Like the ADC, the DAC resolution varies with the output range setting. Users can configure the voltage waveforms of these analog outputs in the LabChart “Stimulator Dialog” window, which offers positive, negative, differential, and independent outputs. By default, Output1 is set as positive and Output2 as negative. When operating in the differential mode, the output between the two terminals doubles the amplitude of each individual output.

The recorder offers two recording modes: continuous and trigger. The continuous mode records signals throughout the entire recording period for later analysis. The trigger mode only captures desired data, initiated by a specific signal from the “Trigger” port located on the front panel (Fig. 3.24A). This “Trigger” accepts TTL-compatible signal, with trigger thresholds of + 1.3 V for rising edge and + 1.1 V for falling edge, featuring a 0.3 V hysteresis and a 50kΩ input impedance. The trigger signal should not exceed ± 12 V, and must last at least 5 μs. When the signal crosses the trigger threshold, a yellow light illuminates on the front panel. The trigger mode is especially useful for recording evoked potentials, as it can conserve storage space by avoiding the large data produced by continuous long recordings. A simple method to use the trigger mode is to connect the recorder's analog outputs (Output1 or Output2) to the “Trigger” inputs and configure the settings in the “Stimulator Dialog” window to produce desired trigger signal. Alternatively, a pulse signal from an external stimulator can serve as the trigger signal.

The LabChart software offers powerful features for display, calculation, and data analysis for both online and offline situations. As shown in Fig. 3.25, its display can be split into two windows, allowing real-time recording in one window while enabling replay and analysis in the other to quickly measure and assess recorded signals. During offline analysis, these split windows can also facilitate signal comparisons across different time periods. Figure 3.27 shows another example of recording signals from the rat hippocampal CA1 region. The top 16 channels recorded wide-band (0.3–5000 Hz) signals in real-time, while the bottom 16 channels showed their high-pass filtering (> 500 Hz) results from real-time calculations. This configuration allows simultaneous online observations of both low-frequency LFPs (top 16 channels) and high-frequency unit spikes (bottom 16 channels).

Full size image

The LabChart display has powerful features. Users can easily adjust zoom levels on both vertical and horizontal axes, and customize channel color, grids, and other display attributes. During recording, users can add comments to mark experimental operations, events and other relevant information in real-time. Additionally, the settings can be saved as templates for future recordings.

LabChart saves recorded documents in a unique format with the file extension “.adicht”. These files can be converted to other formats, such as text (.txt) or MATLAB (.mat), using the “File|Export” menu. Although decimal places in the data can be customized, adding more decimals doesn't necessarily mean increasing data accuracy. For voltage values, the appropriate number of decimal places should be determined by the ADC resolution. Similarly, for time values, the decimal places should be determined by the sampling rate.

Additionally, LabChart offers a down-sampling function to reduce the sampling rate of recorded data. When exporting a file as text using the “File|Export” menu, users can set the down-sampling factor (denoted as m) in the “Down sample by” field of the “Export as Text” dialog box. Rather than simply selecting one data for every m samples, LabChart averages m data and outputs their average as a single data. This averaging process acts as low-pass filtering (or smoothing) to reshape the data frequency-band. Following the sampling theorem—which states that the sampling rate must be at least twice the highest frequency of the original signal (known as the Nyquist frequency)—LabChart first applies a low-pass filter to limit the signal's highest frequency. This ensures accurate down-sampling that complies with the Nyquist frequency requirement.

LabChart offers two viewing modes: “Chart” and “Scope,” accessible through its “Window” menu. “Chart” mimics a multi-channel paper tape recorder, while “Scope” mimics an oscilloscope. These modes reflect early experimental setups, where oscilloscopes provided real-time signal observation and paper recorders and magnetic tape machines stored data. Inspired by these historical setups, LabChart developers incorporated both modes into the software and have maintained them throughout its development. It should be noticed, however, that while the older devices recorded continuous analog signals, PowerLab recorders capture digital signals. These digital signals are quantized in both amplitude and time by ADC, differing from analog signals.

Although most of our commonly used operations are in Chart mode, Scope mode offers specialized features for segmented signal display, allowing for superposition and averaging of signal segments. For example, to calculate an average waveform from 100 evoked potentials produced by pulse stimulations, simply use “Peak Analysis” to identify pulse artifacts, then let Scope to automatically align the 100 evoked potentials using these artifacts and calculate the average. Scope can display the results as superposed evoked potentials along with their average waveform, providing an intuitive comparison. Additionally, all calculation results can be exported in a text file.

“Peak Analysis” and “Spike Histogram” are two powerful LabChart modules. The Peak Analysis can automatically detect and analyze signal peaks in recordings—including unit spikes, population spikes (APS and OPS) and stimulus artifacts—by setting appropriate trigger thresholds in amplitude or slope. It can calculate multiple spike parameters, such as amplitude, width, slope and area. These measurements are useful for analyzing evoked potentials, including APS, OPS and fEPSP. The Spike Histogram can automatically identify unit spikes and create various histograms (refer to Chap. 4).

“Spectrum” menu in LabChart is a powerful tool that uses the periodogram method to estimate frequency spectra of selected signals, as detailed in Sect. 4.3. “Data Pad” is another useful tool for calculating various time and amplitude parameters. For users with programming expertise, LabChart offers macro commands to design and implement more complex processes for automated data detections and analyses. All these features are thoroughly documented in the LabChart Help guide.

The electrical stimulators used in our studies were from the A-M Systems company (http://www.a-msystems.com), including models 2100, 2200, 2300, and 3800. The Model 2100 stimulator is a single-channel pulse stimulator with electrical isolation. It can generate monophasic and biphasic pulses of voltage or current waves. The stimulator can be triggered manually or by an external electrical signal, and features a “Gate” input connector that allows an external signal to restrict its final output. Its front panel consists of three sections plus a power switch on the far right (Fig. 3.28). The left “START” section controls triggering mode; the middle “TIMING” section sets pulse parameters (delay, burst width, duration and period) using a 10 MHz internal clock; the right “OUTPUT” section provides an isolated output with controls for pulse polarity, baseline level and range. Here, users can configure output signals by selecting monophasic/biphasic and voltage/current pulses and setting pulse intensity.

Full size image

As shown in Fig. 3.29, the Model 2300 stimulator is a single-channel pulse stimulus isolator, equivalent to the OUTPUT section of the Model 2100 stimulator (Fig. 3.28). The device uses built-in rechargeable batteries to produce electrically isolated output. Its front panel features a power and battery charging switch in the upper left corner. Note that stimulus signals cannot be output during charging. The panel provides options of voltage and current outputs, each with three different intensity ranges that work with the fine-adjustment knob below to set the amplitude of output pulses. The device itself cannot control pulse timing. Timing and polarity of the output pulses are controlled by external input signals through the 9-pin D-type connector in the lower left corner. The simplest control signal is a TTL-compatible pulse to the connector's “On/Off” pin, which gates the output. In short, the 2300 stimulator provides voltage or current pulses at a set intensity but relies entirely on external signals for timing control.

Full size image

As shown in Fig. 3.30, the Model 2200 stimulator resembles the Model 2300 in appearance and size. However, it is an analog stimulus isolator that can output various waveforms within its bandwidth range by following the control signal input through the “Signal in” BNC connector. In contrast, both the above-mentioned Model 2100 and 2300 are digital stimulators that can only output pulse waves. The front panel of Model 2200 allows users to select between voltage or current outputs with adjustable amplitude ranges. It also features a 9-pin D-type “Control” connector with a “Gate” pin to control the signal through the “Signal in”. Its power switch and output connectors are similar to those of the Model 2300.

Full size image

Figure 3.31A shows the Model 3800 eight-channel programmable stimulator. Its companion Model 3820 stimulus isolator (Fig. 3.31B) runs on rechargeable batteries. This versatile digital stimulator can produce both voltage and current pulse outputs. By connecting it to a computer via USB, users can easily configure all eight channels through software that enables flexible stimulation designs (Fig. 3.31C). It can produce monophasic and biphasic pulses with varying amplitudes and intervals. The stimulator can also create complex pulse sequences by combining up to eight programmed channels to a finally output through a single terminal.

Full size image

Neural signals often need to be recorded simultaneously during electrical stimulations. To prevent interference, the stimulator output circuit must be electrically isolated from the recording circuit, avoiding a shared ground connection. All the above-mentioned stimulators feature such isolation. The Model 3800 stimulator is specially equipped with a 3820 isolator for each channel. These isolators run on rechargeable batteries.

Neural signals are extremely weak, typically requiring thousand-fold amplification before sampling and recording. The amplifiers have a small input range—exceeding this range can cause saturation or even damage. During extracellular electrical stimulation, the applied stimulus intensity far surpasses both neural signals and the amplifier input range. Therefore, stimulation isolation is critical to minimize interference with signal recordings. However, even with isolations, stimulus signals can still be collected by a recording electrode when the recording site is close to the stimulation site (only 1–2 mm in our experiments). This recorded stimulation waveforms, called stimulation artifacts, can have an amplitude much greater than neural signals. Before extracting and analyzing neural signals (such as unit spikes), the stimulation artifacts must be removed, as detailed in Sect. 4.5.

Off-the-shelf stimulators typically can only produce stimulus waveforms with constant or regularly varying parameters. They cannot generate arbitrary stimulation signals, such as pulse sequences with randomly varying intervals (varying frequencies), pulses with continuously varying amplitudes, and so on. These stimulators also cannot perform closed-loop stimulations. To resolve these limitations, we developed LabVIEW software for a data acquisition card to control a stimulator and create a stimulation system. It was able to output complex pulse sequences and arbitrary analog waveforms, and to perform closed-loop stimulations based on real-time recorded neural signals (Cai et al. 2018; Yang et al. 2021; Feng et al. 2012; Hu et al. 2015), as introduced briefly in the next section.

LabVIEW, developed by National Instruments (NI), is a graphical programming tool. With an NI data acquisition card (DAQ), a customized LabVIEW program can create a virtual instrument for data acquisition and instrument control. As shown in Fig. 3.32, we designed a stimulation system comprised three devices: a personal computer (PC), a data acquisition card (USB-6251, National Instruments Inc., USA), and a Model 2200 stimulator.

Full size image

The system operated as follows: a LabVIEW program controlled the DAC in the NI card to convert a digital signal into an analog one, creating the desired stimulation signal. This signal then fed into the “Signal Input” port of the 2200 stimulator, which output the final stimulation signal. Additionally, the LabVIEW program used a digital output port in the NI card to generate a TTL-compatible square wave, which connected to the stimulator “Gate” pin to control its output. When the Gate opened, the stimulator produced a stimulation signal precisely following the analog signal in the “Signal Input” port, but with an adjustable gain. To test the final stimulation current waveforms, a resistor could be connected across the stimulator output terminals to convert current into voltage (Fig. 3.32, right).

The LabVIEW program generated digital signals for desired stimulation waveforms (Yang et al. 2021). Figure 3.33 shows the method to produce a biphasic-pulse sequence with interphase gap (Fig. 3.33A). At a sampling rate of 20 kHz, the minimum pulse width of 0.05 ms met the common requirement for neuromodulation stimulation—typically above 0.06 ms (Merrill et al. 2005). The LabVIEW program created a data array for a biphasic-pulse sequence, with each element occupying a 0.05 ms duration and having a value proportional to the pulse voltage amplitude after DAC. “0”s (zero values) were inserted between adjacent phases to create the interphase gaps and pulse intervals (Fig. 3.33B). This array configuration can generate pulse sequences with varying pulse intervals (IPI), as shown below.

Full size image

Figure 3.34 shows four types of IPI distributions in pulse sequences studied in brain stimulations: uniform, normal, gamma, and Poisson (Wyckhuys et al. 2010; McConnell et al. 2016; Birdno et al. 2008; Feng et al. 2019). The first three are continuous functions while the Poisson distribution is discrete, all generated using LabVIEW VIs for random numbers. These distributions are determined by specific parameters: uniform distribution using minimum and maximum values, normal distribution using mean and standard deviation (SD), gamma distribution using mean and coefficient of variation (CV), and Poisson distribution using mean. Note that except for the inherent discreteness of the Poisson distribution (Fig. 3.34D), the discrete points shown in the other distributions resulted from the limitation in array resolutions (Fig. 3.34B, C). Additionally, for our experimental requirements, we limited the IPI range to 1–500 ms.

Full size image

The data array for a stimulation pulse sequence (Fig. 3.33B) can be generated internally by the LabVIEW program, or externally by other methods before being imported into LabVIEW. This flexibility allows users to design pulse sequences by creating text documents using common software such as Notepad, Excel or MATLAB. Beyond pulse sequences, this data array design enables the stimulation system shown in Fig. 3.32 to output any arbitrary stimulus waveforms.

Alternatively, for pure sine wave stimulation, a simple method is to use a general signal generator as the signal source for the 2200 analog stimulator, while using the “Output” ports of the PowerLab recorder (see Sect. 3.5.3) to provide a gate signal to control the stimulator output. Additionally, this analog stimulator can be replaced by a digital one (such as the 3800 stimulator shown in Fig. 3.17), if the user only needs pulse sequences with a constant intensity or a few discrete intensity steps.

The hardware configuration of the stimulation system shown in Fig. 3.32 is readily available, requiring only a general data acquisition card and a stimulator—standard equipment in most laboratories. This setup accommodates various stimulation needs, including constant and time-varying, digital pulse and analog signal stimulations. It also enables the exploration and development of other new stimulation patterns, as shown in Part II of this book.

Closed-loop stimulation represents a crucial advancement in neuromodulation therapy. This approach delivers electrical stimulation based on real-time brain activity, enhancing both treatment efficiency and patient safety. In epilepsy control, for instance, timely stimulation can suppress epileptiform activity more effectively when an impending seizure has been predicted through monitoring real-time brain neural signals (Sunderam et al. 2010; Smith et al. 2010). However, current off-the-shelf stimulators lack the ability to respond to neural signals for closed-loop stimulations.

We used the stimulation system shown in Fig. 3.32 to implement closed-loop stimulations (Feng et al. 2012; Hu et al. 2015). As shown in Fig. 3.35, LFP signals detected by the recording electrode (RE) in the rat hippocampal region were amplified and sampled via the ADC port in the NI card. The LabVIEW program performed real-time analysis of these signals. When an LFP signal met the preset criteria, one of the digital ports in the NI card sent a “Gate” signal to trigger the stimulator to deliver desired stimulation, which was then applied to a specific brain target via the stimulation electrode (SE). The dashed lines in Fig. 3.35 indicate the wiring for setup testing, with a PowerLab recorder simultaneously collecting neural signals (such as LFPs) and stimulation signals (either the stimulator's trigger signal or its output). Checking these recordings can verify whether the stimulation system is operating as required.

Full size image

We previously used this system to apply pulse stimulations at the peak or trough of theta-rhythm waveforms in the rat hippocampus (Feng et al. 2012). The electrical activity of neuronal ensembles in brain can generate various rhythms in LFPs with amplitudes up to millivolt (Buzsáki 2006). Electrical stimulation applied at different rhythm phases may produce different effects. For example, pulses applied at the peaks versus troughs of hippocampal theta rhythms (2–7 Hz) can induce two distinct synaptic changes: long-term potentiation (LTP) and long-term depression (LTD), resulting in excitatory and inhibitory effects, respectively (Hyman et al. 2003; Huang and Kandel 2005; Huerta and Lisman 1996). To study the differences in evoked potentials at the peaks and troughs of theta rhythms, we designed a LabVIEW program to automatically identify theta rhythms and predict their upcoming peaks and troughs based on LFP amplitudes and rhythm cycles. Figure 3.36 shows typical recordings in the hippocampal soma layer (pcl) and apical dendritic layer (sr). We applied biphasic paired-pulse stimulation to the Schaffer collaterals in the rat hippocampal CA1 region, with a pulse intensity of 0.2 mA and a phase width of 100 μs. The clear theta rhythms in the sr layer before stimulation indicated that the pulses in Fig. 3.36A, B were respectively applied at a peak and a trough of theta waveforms. The enlarged evoked potentials from the first pulse show the OPS in the pcl layer and the fEPSP in the sr layer. The OPS amplitude induced at the theta peak was larger than that at the trough.

Full size image

In addition, we used a threshold method to detect large epileptiform discharges, enabling closed-loop stimulation to suppress epileptiform activity (Hu et al. 2015). Figure 3.37A shows an example from our experimental results recorded in the rat hippocampal CA1 region. Once turned on, the closed-loop stimulation system automatically identified the initial PS of discharges and triggered the stimulator to deliver a 130 Hz pulse train to the Schaffer collaterals. This stimulation was able to effectively reduce the subsequent PSs in discharges.

Full size image

The automatic detection in this closed-loop system can be switched on or off accordingly. When turned on, the LabVIEW program can detect the initial PS in discharges and deliver the preset pulse train with a desired delay. Once the initial PS has been detected, the detection function turns off until the pulse stimulation ends. The frequency and duration of the applied pulse train can be set as needed. Delivering stimulation before an epileptiform discharge fully develops can effectively suppress seizures. This approach can conserve stimulator power and reduce the risks of interfering with normal neurological functions and causing brain damage by continuous electrical stimulations. Using this closed-loop stimulation system, we studied the effect of high-frequency pulse stimulation on epileptiform discharges (Cao et al. 2016), as detailed in Sect. 8.4.