MRI‐Compatible and Conformal Electrocorticography Grids for Translational Research

Abstract Intraoperative electrocorticography (ECoG) captures neural information from the surface of the cerebral cortex during surgeries such as resections for intractable epilepsy and tumors. Current clinical ECoG grids come in evenly spaced, millimeter‐sized electrodes embedded in silicone rubber. Their mechanical rigidity and fixed electrode spatial resolution are common shortcomings reported by the surgical teams. Here, advances in soft neurotechnology are leveraged to manufacture conformable subdural, thin‐film ECoG grids, and evaluate their suitability for translational research. Soft grids with 0.2 to 10 mm electrode pitch and diameter are embedded in 150 µm silicone membranes. The soft grids are compatible with surgical handling and can be folded to safely interface hidden cerebral surface such as the Sylvian fold in human cadaveric models. It is found that the thin‐film conductor grids do not generate diagnostic‐impeding imaging artefacts (<1 mm) nor adverse local heating within a standard 3T clinical magnetic resonance imaging scanner. Next, the ability of the soft grids to record subdural neural activity in minipigs acutely and two weeks postimplantation is validated. Taken together, these results suggest a promising future alternative to current stiff electrodes and may enable the future adoption of soft ECoG grids in translational research and ultimately in clinical settings.

3 crosses. The electrode sites are covered by screen-printing with a dispersion made of platinum nanoparticles (ø=0.27-0.47µm, Strem Chemicals Inc.) and PDMS at a weight ratio of 70%, using cyclohexane (Sigma Aldrich) to help mixing. The mixture is degassed and then cured at 55°C for 4 hours. To connect to the device, an uncured silver epoxy paste (Epotek H27D) is used as a medium between the stretchable gold thin film and a surface mounted zero insertion force device (Hirose) used as connector, connected then to a custom-made polyimide-copper flexible printed circuit board, over molded with silicone adhesive (One component silicone sealant 734 clear, Dow Corning). The device outline is defined by fs laser machining with subsequent release in deionized water. The brain phantom is made by dissolving 3%wt agarose (Electran) in phosphate buffered solution (Gibco PBS, pH 7.4, 1X) in a mold.

Testing
The electrodes are characterized by electrochemical impedance spectroscopy in a 3-electrode configuration (Gamry Instruments Reference 600 Potentiostat) in saline solution (Gibco PBS, pH 7.4, 1X) at room temperature with a platinum wire counter and Ag/AgCl (Metrohm, El. Ag/AgCl DJ RN SC: KCl) reference electrode. The frequency is swept from 1MHz to 1Hz, using a 100mV excitation signal.
Conformability testing on mock-brain PDMS (Sylgard 184, Dow Corning, including a blue silicone dye (Smooth-On, sil pig blue)) thin films were fabricated as described in the microfrabrication section above, varying the spin-coating speed to achieve different thicknesses. Polyimide (Dupont, PI 2611) thin films were prepared by spin-coating the liquid monomer onto silicon wafers and baking the layer according to the manufacturer's instructions, the different thicknesses were achieved by varying the spin-coating speed. Mock-brains were prepared by dissolving agarose into PBS (1%wt) and filling a silicone mold created by negative molding of a 3D printed human brain hemisphere. Once the mock-brain was demolded, the surface was wetted, then the different layers (PDMS and PI) were laid sequentially onto the surface and wetted with saline again.
Pictures were acquired with a handheld camera.

Cadaver head and MRI
In a human cadaver specimen, a large craniotomy is made over the temporal lobe exposing the lateral sulcus, the frontal and the temporal lobes. The dura mater was carefully cut and a flap was opened to expose the cortex. The electrode grid is placed on the surface of the brain.
Saline is applied to the brain to keep it hydrated and ensure optimal capillary action between the electrode array and the brain surface. The grid is then covered again by the dura mater, bone flap and skin. The cadaver specimen is imaged using a clinical computed-tomography (CT) scanner (GE 750 HD CT, General Electric USA, 120-kVp tube voltage, 22.9 effective mAs tube current without current modulation, pitch of 1.375, slice thickness of 0.625, and table speed of 55 mm/rotation) and a 3T MRI scanner (Siemens Magnetom TrioTim 3T, sequence parameters shown in Supplementary Figures Supplementary Figure 14, 15). The images are exported and processed using a DICOM viewer (Miele-LXIV). For the lateral sulcus approach, the cortex is exposed of the sulcus and the blood vessels are carefully separated between the two lobes. The soft electrode array is inserted by first folding it in half.
A cellulose gel is inserted over the array to fill the void above the device. The dura mater and bone flap are closed again together with the skin before imaging.

MRI phantom measurements
The imaging phantom is prepared by dissolving 3%wt agarose (Electran) in phosphate The images are exported using a DICOM viewer (Miele-LXIV) without further processing except adjusting for the position, and adjusting brightness and contrast. For the heating measurements, the device under test is placed on top of a phantom a thermocouple is placed over the electrode sites and the system is covered by insulating foam to ensure close contact between the electrodes and the temperature probe. A T1-weighted turbo spin echo sequence (parameters in Supplementary Figure 18) is run for 15 min split in 3min sequences and the temperature is measured continuously using an ADC (0.1 Hz acquisition frequency). The device is either placed in the axis of the scanner or in orthogonal directions to it.

Acute minipig experiment
Göttingen minipig or young farm pigs (n = 2) were used in the acute in vivo recording experiment. First, each animal was sedated with an intramuscular ketamine injection, 5 followed by isoflurane delivered with a face mask. Once the animal was fully anesthetized, an intravenous catheter was placed in the ear vein. A mix of propofol and saline solution was administrated as anesthetic during the surgery. The animal was taken off the face mask and intubated with active respiration. Heartrate, body temperature and blood oxygen pressure were monitored continuously. The animal was placed on the surgery table (over a heating blanket) covered with sterile drapes with top of the head exposed. A large frontal to posterior incision over the skull was performed. The skin and underlying muscles were separated from the skull and pulled aside with forceps, to expose as much skull area as possible. Four burr holes were drilled with a 5 mm diameter drill at the corners of the exposed skull. The skull between the holes was drilled to open a square bone flap. The bone flap was removed to expose the dura. The dura mater was cut open with a scalpel blade and flapped over until 2mm lateral from the midline. The brain was continuously kept wet with saline solution. The arrays were placed over the defined cortex region by holding the fPCB cable with a custommade 3D printed frame. The snout was stimulated with bipolar needle electrodes (Ambu Neuroline Twisted Pair Electrodes 745 12-100, Ambu) (3, 5 and 8 mA amplitude, cathodicfirst biphasic pulses, 1 Hz repetition rate, 300 µs pulse width) using an external pulse generator (AM System). Evoked potentials were recorded differentially between the electrode site and a subdural ground wire on the frontal end of the craniotomy (triggered averaged recording (n~60), band-pass filtered 1-5000 Hz, notch filter at 50 Hz and harmonics) using a commercially available amplifier and data acquisition system (PZ5 preamplifier and RZ2 base station, Tucker-Davis Technologies).

Chronic minipig experiment
An Aachener minipig was used in the subchronic in vivo recording experiment. The animal was anesthetized with isoflurane with a face mask. Once the animal was fully anesthetized, an intravenous lead was placed on the ear. Heartrate, body temperature and blood oxygen pressure were monitored continuously. The animal was placed on the surgery table (over a heating blanket) covered with sterile drapes with top of the head exposed. A large frontal to posterior incision over the skull was performed. The skin and underlying tissue were separated from the skull on an area the size of the chronic titanium chamber housing the connector. A cranial window was opened with a bone drill. The bone flap was removed to expose the dura. The dura mater was cut open with a scalpel blade and flapped over. The brain was continuously kept wet with saline solution. The soft electrode array was placed on 6 the brain and gently pushed under the skull. After fixing a reference wire under the dura and a ground wire on the skull, the dura was closed again and covered with artificial dura mater.
The bone flap was placed back and secured with titanium pieces onto the skull. A titanium chamber was screwed into the skull using screws and the inside of the chamber was filled with dental cement securing the connector in place. An acrylic cap was screwed over the chamber to protect the connector during the movement of the animal. After two weeks of post-operative recovery, the animal was placed into a soundproof chamber equipped with a microphone. A wireless amplifier headstage (Multichannel systems GmbH) was plugged into the connector on the head of the animal to measure the brain activity. Both signals were acquired synchronously. The brain signals were acquired at 20 kHz, low-pass-filtered below 10 Hz and down-sampled at 200 Hz. The brain recording was averaged across 76 spontaneous vocalizations triggered by their onset time. The interval during which the averaged evoked response exceeded the baseline noise level was isolated according to a Welch test ensuring P<0.05 statistical significance. Spatial activity maps projected onto the cortical anatomy were thresholded according to this significance level.

Supplementary Figures
Supplementary

Electrode
Pitch 10 mm 4 mm 10 mm 10 mm 20 µm in [2] 2 mm [3] 0.   Supplementary Table 2. Materials properties used for the calculation of the bending stiffness. The value for PDMS is extracted from [4] and for polyimide from [5] .