Medical device, therapy system and method

Abstract

The present invention provides a system for treating a patient having a spinal cord injury, the system comprising: a first flexible implantable device having a first plurality of electrodes; a second flexible implantable device having a second plurality of electrodes; and a controller communicatively coupled to the electrodes of the first and second implantable devices, and configured to receive an input signal from the electrode of the first implantable device, process the input signal to generate a stimulation signal, and transmit the stimulation signal to the electrode of the second implantable device. The device may be configured for implantation at a plurality of sites in a patient, including the brain, spinal cord, nerve, or limb. The controller may be configured to: process the input signal by decoding the input signal to determine an expected stimulation of a patient's nerve at the head end of the location of the first device on the spinal cord; and generating a stimulation signal that produces the intended stimulation of the nerve of the patient when the stimulation signal is applied to the spinal cord of the patient by the second device. Methods of treating spinal cord injury are also provided.

Description

Technical Field

[0001] The present invention relates to medical devices, treatment systems, and methods. In particular, but not exclusively, the present invention relates to implantable devices, such as implantable devices for treating spinal cord injuries. Background Art

[0002] Spinal cord injury (SCI) is a disabling neurological condition that can lead to irreversible loss of sensory, motor, and autonomic functions. Spinal cord injury (SCI) affects more than 15 million people worldwide, including an annual increase of 250,000 to 500,000 people. SCI occurs when the neuronal bundles of the spinal cord are damaged.

[0003] SCI also imposes a socioeconomic burden on individuals, their families, and the healthcare systems that support them.

[0004] Recent advances in targeted spinal cord stimulation (SCS) have made it possible for patients with SCI paralysis to regain assisted walking [1-3]. These pilot studies relied on open-loop initiation of stimulation algorithms or external closed-loop triggering via inertial measurement units. Alternatively, closed-loop SCS has also been used by decoding internal neural signals from the cerebral cortex [4]. These methods aim to restore motor function after SCI and focus on dorsal stimulation of the spinal cord by using commercially available SCS devices that are primarily targeted at chronic pain. The mechanism is hypothesized to indirectly activate motor neurons through proprioceptive circuits located dorsally [5].

[0005] Previous spinal cord recording methods employed the use of penetrating electrodes [9, 10], and advancements in electrode manufacturing have allowed for the implantation of multiple electrode arrays to increase the area of simultaneous neural recording [25, 26]. However, penetrating electrodes can cause spinal cord injury along the electrode insertion tract and translate into problems faced in long-term applications, which are wire breakage

[25] and chronic nerve injury due to mismatched elastic moduli

[27] . Summary of the Invention

[0006] An object of the present invention is to solve one or more of the above problems.

[0007] Aspects of the present invention aim to provide systems and methods that can effectively treat or overcome the effects of spinal cord injury and restore the motor level of patients with SCI, preferably without invasive surgery or the use of minimally invasive surgery.

[0008] Aspects of the present invention also aim to provide systems and methods for treating spinal cord injury using inductive sensing and electrical stimulation.

[0009] Aspects of the present invention provide systems and methods for treating patients with spinal cord injuries, which use implanted electrodes to detect signals from the patient's brain or nervous system, process those signals and generate stimulation signals and transmit them back to the patient's nervous system.

[0010] A first aspect of the present invention provides a system for treating a patient with a spinal cord injury, the system comprising: a first flexible implantable device having a first plurality of electrodes; a second flexible implantable device having a second plurality of electrodes; and a controller communicatively coupled to the electrodes of the first and second implantable devices and configured to receive an input signal from the electrodes of the first implantable device, process the input signal to generate a stimulation signal, and transmit the stimulation signal to the electrodes of the second implantable device.

[0011] The devices may be configured for implantation at a series of sites within the patient. Non - restrictively, the devices may be configured for implantation on a nerve, a muscle (e.g., a muscle in the patient's limb), the spinal cord or the brain; around or near a nerve, a muscle (e.g., a muscle in the patient's limb), the spinal cord or the brain. In particular embodiments, the first device may be implanted in the patient's brain or adjacent to the patient's spinal cord or nerve, e.g., rostrally (i.e., closer to the brain) relative to the SCI site. In particular embodiments, the second device may be implanted near the spinal cord or nerve, or in the patient's limb, caudally (i.e., away from the brain) relative to the SCI site.

[0012] In some embodiments, either or both of the first and second implantable devices are configured to conform to the patient's spinal cord or nerve when implanted within the patient. Implantation at a site adjacent to the patient's spinal cord or nerve allows for a tight coupling between the electrodes and the patient's nerve, both for detecting signals from the nerve and for stimulating the nerve. Configuring the device to conform to the spinal cord or nerve can further improve this coupling.

[0013] To improve the conformability of the device, devices of some embodiments have a gel layer disposed on a surface of the device opposite the surface of the device having the plurality of electrodes (i.e., if the electrodes are on the "front surface" of the device, the gel layer is disposed on the "back surface" of the device).

[0014] Preferably, the gel layer is expandable or swellable, e.g., by hydrophilicity and absorption of water from the tissue surrounding the implantation site. Thus, the gel layer can expand after implantation and fill the space around the implantation site. This results in the device applying a light (i.e., non - compressive) pressure to the spinal cord when the device is placed subdurally (or against the nerve), which can hold the device in the desired position and provide tight contact between the electrodes and the spinal cord or nerve.

[0015] In some embodiments, either or both of the first and second implantable devices are configured such that they are biased into a cylindrical configuration. In this way, the devices can self-roll or self-hoop such that they naturally conform to the spinal cord, nerve, or other implantation site in order to hold the device in the desired position and / or provide close contact between the electrodes and the spinal cord or nerve. A variety of mechanical methods can be used to create such a bias. For example, in some embodiments, the device can be formed from multiple polymer layers and a stress mismatch created between the polymer layers, which causes a natural bias towards rolling up. The stress mismatch can be achieved by making different polymer layers (e.g., the top and bottom layers) have different thicknesses, by asymmetric thermal annealing, or by designing mechanical features into the device (e.g., by etching).

[0016] In some embodiments, either or both of the first and second implantable devices are configured such that when implanted in a patient, they substantially encircle the patient's spinal cord or nerve, and preferably completely encircle the spinal cord or nerve. By encircling the spinal cord or nerve, the device can closely conform to the spinal cord or nerve and good coupling between the electrodes and the nerve can be achieved. By having the first device encircle the spinal cord, the controller can use the spatial (i.e., circumferential) arrangement of the electrodes to interpret and identify signals from all the nerves in the spinal cord to determine the signals to act upon. Similarly, by having the second device encircle the spinal cord, the controller can stimulate the selected electrodes based on the position of the selected electrodes to cause stimulation of specific nerves within the spinal cord.

[0017] In some embodiments, the first and second implantable devices each have a distal portion, and each said distal portion includes a respective one of said plurality of electrodes, further wherein the distal portions of the first and second implantable devices are the same. Other parts of the devices can also be the same. Thus, a simplified system can be used that utilizes the same device for both sensing and stimulation.

[0018] Preferably, the controller is configured to: process the input signal by decoding the input signal to determine the expected stimulation of the patient's nerves at the cephalic end of the first device location on the spinal cord; and generate a stimulation signal that, when applied to the patient's spinal cord via the second device, produces the expected stimulation of the patient's nerves. Thus, by picking up nerve signals above the SCI site (i.e., on the brain side) and stimulating the nerves below that site, the system can effectively bypass the SCI site in order to effectively reapply the desired stimulation to the nerves for forward stimulation of the muscles.

[0019] Preferably, the controller is configured to perform the steps of receiving, processing, and transmitting substantially in real time. For these purposes, "real time" should be understood to mean having a delay similar in magnitude to the delay physiologically present in an uninjured spinal cord.

[0020] In certain embodiments, the controller is configured to process the input signal by integrating the input signal in the time domain and the spatial domain. The controller is further configured to identify unique topographical features from the integrated input signal to identify the major spinal tracts that are activated. Integration in the time domain and the spatial domain and / or identification of topographical features from the input signal can allow for processing of complex signals obtained from multiple nerves in, for example, the spinal cord to identify signals of interest and determine appropriate stimulation output signals corresponding to those signals.

[0021] The controller can be coupled to the electrodes of the device by a wired or wireless connection. The wireless connection can use any known protocol. The wired connection can improve (i.e., reduce) latency and can reduce the complexity of an individual device, but it also requires more complex surgical planning and implantation of leads.

[0022] The controller itself can be implantable, which is particularly beneficial when the coupling to the electrodes is wired.

[0023] Preferably, the controller can be configured to wirelessly receive power from a separate power source. In such a configuration, the controller and the device can be implanted in a patient, but a power source does not need to be implanted. Using an external power source means that no further surgery is required to replace the power source.

[0024] The first plurality of electrodes and / or the second plurality of electrodes can be arranged in a regularly spaced array. Different configurations of the array can be selected. For example, the electrodes can be arranged in multiple rows (e.g., longitudinally extending along the distal end of the device such that the rows of electrodes can encircle a nerve or the spinal cord). The electrodes in adjacent rows can be

[0025] Preferably, each implantable device has a total thickness of less than 100 μm. More preferably, each device has a total thickness of less than 50 μm, and in some embodiments, a total thickness of less than 10 μm.

[0026] Preferably, the device is configured to be implanted and / or removed through an incision no greater than 1 cm.

[0027] The systems of the above aspects can include some, all, or none of the above optional and preferred features in any combination.

[0028] A further aspect of the invention provides a flexible implantable device having a plurality of electrodes. The device of this aspect is preferably one of the devices used in the systems of the above aspects and can have some, all, or none of the above preferred and optional features of such a device.

[0029] According to a further aspect of the invention, there is provided a method of treating a human or animal body, the method comprising implanting a medical system or device according to the above aspects.

[0030] A further aspect of the invention provides a method of treating a patient with a spinal cord injury, the method comprising the steps of: percutaneously implanting a first plurality of electrodes at a position upstream of the spinal cord injury site; percutaneously implanting a second plurality of electrodes at a position downstream of the spinal cord injury site; processing signals from one or more of the first plurality of electrodes to determine an expected stimulation of a patient's nerve downstream of the spinal cord injury location; generating a stimulation signal which, when applied to the patient's nerve through one or more of the second plurality of electrodes, produces the expected stimulation of the patient's nerve.

[0031] The electrodes may be implanted at a series of sites within the patient. By way of non-limiting example, the implantation may be on a nerve, muscle (e.g. in a patient's limb), spinal cord or brain; around or near a nerve, muscle (e.g. in a patient's limb), spinal cord or brain. In particular embodiments, the first plurality of electrodes may be implanted in the patient's brain or adjacent to the patient's spinal cord or nerve, e.g. cephalad (i.e. closer to the brain) relative to the SCI site. In particular embodiments, the second plurality of electrodes may be implanted near the spinal cord or nerve, or in the patient's limb, caudal (i.e. away from the brain) relative to the SCI site.

[0032] In certain embodiments, one or both of the first plurality of electrodes and the second plurality of electrodes are implanted such that they substantially surround the patient's spinal cord or nerve, and preferably completely surround the spinal cord or nerve. By surrounding the spinal cord or nerve, the electrodes can closely conform to the spinal cord or nerve and good coupling between the electrodes and the nerve can be achieved. By implanting the first plurality of electrodes to surround the spinal cord, the spatial (i.e. circumferential) arrangement of the electrodes can be used in the processing step to interpret and identify signals from all the nerves in the spinal cord to determine the signals to act upon. Similarly, by surrounding the spinal cord with the second plurality of electrodes, the step of generating the stimulation signal can generate a stimulation signal that stimulates the selected electrodes based on the position of the selected electrodes to cause stimulation of a specific nerve within the spinal cord.

[0033] The method further comprises the step of percutaneously removing the electrodes after the treatment. The advantage of small devices with flexible electrode arrays is that they can be removed without significant difficulty and without causing harm to the patient at or around the implantation site. Thus, for example, if an alternative therapy is being used, or if the electrodes need to be replaced or repositioned, the electrodes can be removed.

[0034] Unless otherwise specified, any feature (including optional or preferred features) described for one of the above aspects equally applies to the devices, systems, and methods in combination with any of the other above aspects. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The present invention will be described by way of example with reference to the accompanying drawings, in which:

[0036] Figure 1 shows the design and characteristics of an implantable device used in an embodiment of the present invention, as well as the implantation surgical procedure forming part of an embodiment of the present invention;

[0037] Figure 2 shows a system according to an embodiment of the present invention, in which the device is implanted at a site on a patient's spinal cord;

[0038] Figure 3 shows the manufacturing Figure 1 steps of the device;

[0039] Figure 4 shows the Figure 1 use of the device as a nerve stimulator;

[0040] Figure 5 shows the use Figure 1 of the device for circumferential spinal cord recording and testing;

[0041] Figure 6 shows the various steps and results from the analysis of the recorded spinal signals;

[0042] Figure 7 shows the principle of a bioelectronic bypass at the site of an acute spinal cord injury;

[0043] Figure 8 shows the experimental results from a study aimed at demonstrating the preservation of spinal cord circularity in animals implanted with a device according to an embodiment of the present invention;

[0044] Figure 9 shows the design and characteristics of an alternative implantable device used in an embodiment of the present invention;

[0045] Figure 10 shows the results from the analysis of the foreign body reaction associated with the implantation of the Figure 9 device;

[0046] Figure 11 shows the use Figure 9 of the device to record electrode signals from a target nerve;

[0047] Figure 12 shows the Figure 9The intended implantation of the devices and the signals from nerves recorded by those devices;

[0048] Figure 13 shows the results of a stimulation test using Figure 9 the devices; and

[0049] Figure 14 shows the functional elements of a system according to one embodiment of the present invention. Detailed Description

[0050] It has been shown that direct stimulation of ventral horn motor neurons by ventral stimulation is more effective in activating muscles and allows a greater degree of selectivity [6]. This finding has led to recent efforts to develop ventral SCS devices [7, 8]. In addition, the spinal cord near the injury site, rather than external algorithms or the cerebral cortex, represents a more feasible and patient-friendly target for the neural recording and decoding of motor intent [9, 10]. However, both the anatomical approach to the ventral spinal cord and the materials used to limit iatrogenic nerve injury present challenges. In addition, to date, due to limitations in recording and stimulation techniques (which only allow coverage of a limited area of the spinal cord), spinal cord signals have not been used as control signals for closed-loop SCS.

[0051] By leveraging advancements in thin-film bioelectronic devices that more closely match the mechanical stiffness of neural tissue [11, 12], embodiments of the present invention include non-penetrating conformal bioelectronic arrays that are capable of circumferentially docking with the spinal cord. This has been demonstrated to be possible without causing iatrogenic spinal cord injury in rats. As further described below, this approach allows for comprehensive recording of nerve signals around the spinal cord in high spatial and temporal dimensions, which provides the possibility for whole spinal cord signal analysis as well as simultaneous measurement of descending and ascending tracts. Embodiments of the present invention combine recording and stimulation and use these circumferential spinal cord implants to achieve direct, low-latency spinal cord electronic bypass to restore hindlimb movement in acute SCI. This has been tested in a rat model and has also been tested in a human cadaveric spinal model.

[0052] Devices and Surgical Design

[0053] Figure 1 shows the details of the devices used in embodiments of the present invention, as well as further aspects associated with the devices and surgical design.

[0054] Figure 1 A is an exploded view of a microdevice 10 as used in embodiments of the present invention. The complete fabrication of the device is further described below in conjunction with Figure 3 The device 10 has a proximal end 12 and a distal end 14. Figure 1 B shows Figure 1The device shown in A has dimensional references and also shows the toroidal flexibility of the device.

[0055] To achieve circumferential implantation around the spinal cord, the distal end 14 of device 10 has a thin flexible electrode array 16, which consists of 32 electrodes 18 arranged in a staggered linear configuration. This configuration maximally captures neural signals distributed around the spinal cord while minimizing crosstalk contamination

[13] . The device is attached with a reinforced gold ring that allows a 7-0 polypropylene suture (Ethicon Inc., Raritan, NJ, USA) to pass the device through the ventral side of the spinal cord. Array 16 has a parylene-C substrate (with titanium and gold electronics) and is fabricated using standard photolithography techniques

[14] .

[0056] Preferably, the flexibility of the distal end 14 of device 10 is such that it can conform to and / or wrap around the part of the patient that is intended to interface with it. For a device intended to interface with the human spinal cord, this means that a bend radius of no more than 0.5 cm is preferred (the radius of the smaller human spinal cord), and a preferred bend radius is no more than 0.25 cm. Since device 10 fabricated and tested for this purpose is designed to interface with the rat spinal cord, they are adaptively fabricated to have a much smaller bend radius. For a device intended to interface with a single nerve, the bend radius is preferably in the micron range, e.g., no more than 500 μm, preferably no more than 100 μm, and ideally even smaller.

[0057] The bend radius is measured based on the inner curvature and is the minimum radius at which an element (in this case the device) can be bent in at least one direction without damaging it. The bend radius defined here refers to elastic deformation as opposed to plastic deformation, such that a device bent to a radius greater than the minimum bend radius under an applied force will at least partially return to its original shape upon removal of the applied force. In other words, in these embodiments, the device in this aspect can be bent to an inner curvature of 0.25 cm (spinal cord interface device) and still function exactly as it did before being bent.

[0058] The conductive polymer poly(3,4-ethylenedioxythiophene) (PEDOT), with poly(styrenesulfonate) (PSS) as its counterion, is used to reduce the impedance of the electrodes and thus improve the performance of recording and stimulation

[15] . Figure 1D shows an optical image of device electrode 18, showing the PEDOT:PSS coating. The footprint of electrode interconnect 15 is minimized to allow the device to be implanted through the gap between the outgoing nerve roots. The circumferential spinal interface device 10 is manufactured in a range of sizes (lengths of 8.90 mm, 9.90 mm, and 10.90 mm) to allow adaptation to variations in spinal anatomy in order to implant a device that maximizes electrode coverage around the spinal cord. It should be noted that these specific sizes were chosen for rat studies, as further described below. The sizes of the devices for human use are accordingly selected. In particular, for a device intended to circumferentially wrap around the human spinal cord, the distal end of device 14 (with electrode array 16) is preferably at least 3 cm in length, and the length can be between 3 cm and 5 cm (to circumferentially wrap around a human spinal cord that typically has a diameter of 1 cm to 1.5 cm). For a device intended to wrap around a nerve, the size of the device will be correspondingly smaller. The number and configuration of electrodes 18 in electrode array 16 are also selected accordingly. In a square configuration, the electrode size is 0.1 mm × 0.1 mm.

[0059] The device was characterized in vitro using electrical impedance spectroscopy (EIS). Figure 1 C shows the EIS measurements from device 10, where n = 13 electrodes, from 2 devices. Figure 1 The inset in C shows the range of impedance values seen at 1 kHz, where individual device electrodes show low impedance (1 kΩ - 6 kΩ) at 1 kHz.

[0060] The device was implanted proximal to the L1 anatomical level in the rat to directly interface with the spinal cord rather than the outgoing nerve roots that are predominantly distal at the L3 - L4 level. Figure 1 E shows the process flow of the surgical and device placement steps performed prior to the recording or stimulation experiments discussed below. Figure 1 F is an intraoperative image of device 10 wrapped around the longitudinal axis of spinal cord S. The bottom of the image shows the superior articular facet of L1.

[0061] To detect any potential iatrogenic spinal cord injury during device implantation, neuro-monitoring was performed regularly during implantation and did not show an increase in the threshold amplitude required to evoke motor evoked potentials (MEP)

[16] . After laminectomy, the device was inserted using a leading 7-0 polypropylene suture (Ethicon Inc., Raritan, NJ, USA). After successful device implantation, electrodes located on the spinal midline were used to calibrate the topographic distribution analysis of the acquired signals.

[0062] Figure 2A system is schematically shown having two devices 10 (such as those previously described, located at positions wrapped around the patient's spinal cord S). These sites can be selected on either side of a lesion in the patient that causes SCI. The devices are connected to a controller (not shown) via leads 22.

[0063] Manufacturing the device

[0064] Figure 3 Illustrated are the manufacturing steps in an example method of manufacturing device 10 (such as those described above with respect to Figure 1 The steps of the method are described in more detail below.

[0065] Using a PDS2010 Labcoter TM 2 (Specialty Coating Systems Ltd, Woking, UK) performs the first parylene deposition on a 4-inch silicon wafer (Microchemicals GmbH, Ulm, Germany), where 4 g of parylene-C dimer forms a 2-μm layer. Subsequently, the first photolithography for gold electrode patterning is performed using a MA / BA6 mask aligner ( MicroTec, Garching, Germany) to spin coat nLOF 2035 negative photoresist (MicroChemicals GmbH, Ulm, Germany) at 500 revolutions per minute (rpm), 1000 rpm for 5 seconds, and 6000 rpm, 8000 rpm for 30 seconds to achieve a thickness of 2.00 μm - 2.50 μm, followed by soft baking at 110 °C for 60 seconds, ultraviolet light exposure for 7 seconds, and post-baking at 108 °C for 150 seconds. 726 MIF (MicroChemicals GmbH, Ulm, Germany) is used as a developer for 20 seconds - 30 seconds, then washed with distilled water and dried with compressed nitrogen.

[0066] Using a Plasma Pro 80 RIE (Oxford Instruments Plasma Technology, Bristol, UK), the sample is activated using oxygen plasma under vacuum (60 seconds at 100 W and 0.6 mbar of O2) to improve the adhesion between the parylene and the electronic components, which are titanium (Ti) and gold (Au) deposited by electron beam evaporation (Kurt J. Lesker Company, Jefferson Hills, PA, USA). Under vacuum (<6×10 -6Evaporate 10 nm of Ti under vacuum and let it stand for 10 min, then deposit 100 nm of Au. Then place the sample in acetone for 30 min - 60 min, and then carefully remove it with a swab stick. Finally, rinse them with isopropyl alcohol (IPA) and DI.

[0067] Deposit the second layer of parylene using the same steps as above. Subsequently, use a MA / BA6 mask aligner ( MicroTec, Garching, Germany) to perform photolithography on the profile of the device, spin-coat 10XT (MicroChemicals GmbH, Ulm, Germany) negative photoresist at 500 rpm, 1000 rpm for 5 s and 3000 rpm for 30 s to obtain a thickness of approximately 6.00 μm, then soft bake at 110 °C for 120 s, and perform ultraviolet light exposure in 6 steps of 5 s with a 5 s interval between steps. Then place the sample in 726MIF (MicroChemicals GmbH, Ulm, Germany) developer for 6 min - 7 min and wash with DI. Then use Plasma Pro 80 RIE (Oxford Instruments Plasma Technology, Bristol, UK) to complete reactive ion etching (RIE) at an etching rate of 200 nm / min. Subsequently, deposit a third layer of 2 μm sacrificial parylene separated by soap (2% Micro-90 cleaning solution, Cole-Parmer Instrument Company Ltd, St. Neots, UK), and then use 10XT (MicroChemicals GmbH, Ulm, Germany) for further photolithography to expose the electrodes and contact pads, spin-coat at 3000 rpm for 30 s, and then bake at 110 °C for 120 s. Then expose and develop the sample, and then perform RIE as described above until the exposure of the metal is completed.

[0068] The conductive polymer poly(3,4-ethylenedioxythiophene) (PEDOT) is combined with poly(styrenesulfonate) (PSS) as the counterion of PEDOT to increase the conductivity of the electrode

[15] . PEDOT:PSS is prepared according to the method described previously

[34] , where 19 mL of CLEVIOS TMPH 1000 PEDOT:PSS (Heraeus Holding, GmbH, Hanau, Germany) was mixed with 1 mL of ethylene glycol (EG, to improve the conductivity of the film) and 2 drops of 4-dodecylbenzenesulfonic acid (DBSA) (to enhance the uniformity of the spin-coated film). The mixture was sonicated for 10 min - 15 min and filtered through a 1.2 μm polytetrafluoroethylene (PTFE) filter.

[0069] Subsequently, the PEDOT:PSS layer was spin-coated at 3000 rpm for 30 s, baked at 110 °C for 60 s, and two cycles of spin-coating at 1500 rpm for 30 s and baking at 110 °C for 60 s were performed. Finally, the sacrificial parylene layer was peeled off to expose only the PEDOT:PSS layer, and the sample was hard-baked at 130 °C for 1 h. The device was encapsulated using a 33-channel flat flexible cable (Mouser, UK) with a bonder (Fineplacer, Finetech GmbH & Co. KG, Berlin, Germany) using a 5 μm anisotropic conductive film (ACF, Jetro, Japan).

[0070] The device was used to stimulate the spinal cord circumferentially

[0071] A first set of tests was performed using a device for selective spinal cord stimulation such as described above. These tests investigated the application of spinal cord stimulation around the periphery of the spinal cord, but with particular focus on the effect of tonic stimulation on the ventral side of the spinal cord. First, the charge storage and charge injection capacity of the device were characterized. Using microelectrodes, it is important to ensure that the delivered charge does not damage the neural tissue. The application of PEDOT:PSS on conductive metal (Au) electrodes significantly increased the charge storage and charge injection capacity

[17] and allowed a wider stimulation limit. These data are summarized separately in Figure 4 A and Figure 4 B.

[0072] Figure 4 A shows representative cyclic voltammetry (CV) traces and derived charge storage capacity (CSC) data (n = 10). Figure 4 B is a representative transient voltage plot after an electrical pulse and derived charge injection capacity (CIC) data (n = 8).

[0073] Furthermore, the application of the device to evoke specific motor actions, ideally located at the ankle, hamstring, and calf of the animal under study, was investigated. Other studies covered whether stimulation at the ventrolateral threshold amplitude level would elicit a motor response when the dorsal region of the spinal cord was stimulated, as Figure 4 C shows that it does not elicit a motor response.

[0074] Figure 4C shows the results of stimulation studies in a rodent model, which indicate specific control of several lower limb motor groups and lack of effect on the dorsal part of the spinal cord. Figure 4 E is two static extractions from a live video recording of lower limb muscle flexion during spinal cord stimulation at 20 μA.

[0075] The stimulation experiments ( Figure 4 C and Figure 4 E) are supplemented with computer simulation data ( Figure 4 D), observing current spread to infer possible evocations at different points throughout the spinal cord. Figure 4 D shows the results of finite element simulations of the electric field strength in the rat spinal column. Simulations were performed at multiple peak current levels, with the minimum and maximum currents used during in vivo experiments being 10 μA and 200 μA respectively.

[0076] A tonic stimulation pattern of 50 Hz - 100 Hz and a 10 - 20 pulse train were used to evoke smooth movements in a controlled bilateral manner. As Figure 4 shown in C, ankle, hamstring, calf, gluteal, and flank muscles were evoked. EMG data were collected from the gastrocnemius (calf), tibialis anterior, and vastus lateralis (quadriceps) muscles using steel needle electrodes. For each electrode, three recordings were made and superimposed, showing differences in muscle evocation between two ventral electrodes spaced 0.5 mm apart. Figure 4F Shows EMC recordings (n = 3) of ventral electrode stimulation. Figure 4G Shows EMG recordings (n = 3) of ventral electrode stimulation using electrodes positioned 0.5 mm more laterally. EMG recordings were taken from the ventral midline, all on the left side. Needle electrodes were used in the gastrocnemius (calf), tibialis anterior, and vastus lateralis (quadriceps) muscles, and EMG recordings were made at 30 ks / s using an Intan RHS unit.

[0077] For simulation modeling, a three - dimensional finite element simulation was performed using COMSOL Multiphysics 6.0 (the current interface of the AC / DC module) to solve the point form of Ohm's law assuming current conservation, in order to analyze the electric field distribution in the rat spinal column. Fixed DC electrode injection currents of 10 μA or 200 μA at the electrode surface were simulated. The model geometry is a three - dimensional extension of a two - dimensional geometry based on an MRI scan of the rat T10 vertebra 41. The geometry consists of muscle, bone, epidural cavity, dura mater, cerebrospinal fluid, white matter, and gray matter regions. The device is represented by a ring of 4 - μm - thick low - conductivity (0.001 S / m) material, which represents a parylene - C electrode with a size of (100×100) μm 2 placed directly on the dura mater surface. All regions are approximated as fully isotropic conductors. The spinal column model is embedded in a (50×50) mm 2A cube with muscle tissue conductivity and a grounded outer boundary. These figures show the electric field strength in a plane perpendicular to the electrodes.

[0078] Use the device to record spinal signals circumferentially

[0079] Then use the same device as previously described for stimulation studies to record nerve signals around the spinal cord.

[0080] After safe implantation of the device, as demonstrated by the preserved hindlimb MEP responses, acute electrophysiological tests were performed by recording the sequential activation of bilateral MEP triggered by the hindlimb motor cortex and the somatosensory evoked potentials (SSEP) via the sciatic, tibial, and peroneal nerves through the device. The electrodes around the spinal cord were circumferentially placed such that a comprehensive topographical representation of the amplitude of the acquired signals could be obtained at each sampling interval, as Figure 5 shown in A.

[0081] Figure 5 A shows the process flow for creating the topographical map: First, record the raw signals from 32 channels around the spinal cord. Second, reference and filter the channels to remove artifacts. Third, orient the features on the map based on the electrode positions and coverage. Fourth, create the complete topographical map based on the peak amplitudes of the nerve signals.

[0082] This spinal cord topographical map takes advantage of the ordered arrangement of the ascending and descending tracts through the spinal cord. This is illustrated when stimulating the left sciatic nerve produces a hot spot on the spinal cord topographical map corresponding to the expected site of the dorsal column, as Figure 5 shown in B, Figure 5 where B is the topographical map created from the somatosensory evoked potential recordings.

[0083] The density of the recordings by the electrodes around the spinal cord also allows for better discrimination of the various hot spots around the spinal cord. The descending motor pathways are mediated by the rubrospinal and corticospinal tracts in rats [18 - 20]. This phenomenon is shown by a spinal cord topographical map that indicates the simultaneous appearance of two peak activation sites after MEP stimulation, corresponding to the approximate sites of these descending motor axons. Figure 5 C shows how the motor evoked potential (MEP) recordings indicate that the peak amplitude hot spots are mainly located in the ventrolateral and lateral regions of the spinal cord, corresponding to the approximate sites of the ventral corticospinal tract (CST) and the rubrospinal tract (RST), respectively.

[0084] In addition to using channel reference to isolate nerve signals, wavelet decomposition can also be used as a denoising technique to automatically isolate nerve signals from the instrumental and biological noise generated by electrocardiogram (ECG) signals and respiratory activity

[21] . This is shown in Figure 5Shown in D, which shows the extraction of the cardiac and neural components of the raw signal. Using any of the neural signal extraction methods allows the study of the signal waveform to obtain conventional measurements of latency, amplitude, and neural conduction velocity in a low-noise environment. Figure 5 E shows how to analyze the recorded compound action potential (CAP) in more detail, and the details of the waveform analysis are provided here. Figure 5 F shows the left and right SSEP and MEP waveforms from representative electrodes on the spinal cord.

[0085] A subset of the original recording was independently analyzed using wavelet decomposition to verify the ability of the method to automatically isolate neural signals from biological and instrumental noise, thereby reducing the manual effort associated with the previously described processing. By using the discrete wavelet transform (DWT, originally proposed by Diedrich et al.

[21] ) for denoising and multiresolution analysis of the signal, a direct method for identifying ECAP was employed, as this choice is superior to other alternatives when dealing with similar neural signals.

[0086] The 5th-order orthogonal Daubechies with 3-level decomposition were empirically selected based on their similarity to the ECAP of interest and to filter out low frequencies. The hard thresholding rule wavelet denoising technique proposed by Donoho

[45] was selected, which essentially applies a threshold to the coefficients at each decomposition level in the orthogonal time-frequency domain to further isolate the structures of interest and then transforms them back to the original domain for denoising. The threshold is calculated as follows:

[0087]

[0088] where σ is the standard deviation of the Gaussian noise and N is the number of samples in the coefficients. Since orthogonal wavelets are used, the coefficient lengths at all decomposition levels are the same as those in the original signal. At each level "l", σ is calculated as:

[0089]

[0090] where c l is the value of the wavelet detail coefficient at each level, and 0.6745 is the 75th percentile of the standard normal distribution

[45] . The detail coefficients are then hard-thresholded to reconstruct the denoised signal as follows:

[0091]

[0092] Spikes on the reconstructed signal are then identified by using a detection threshold defined as in equation (1). A spike window centered on the detected peak with a length of 4 ms is extracted to obtain the corresponding waveform. The cardiac component of the signal is also separated by applying continuous wavelet decomposition using the Ricker wavelet family at a 5-ms scale.

[0093] Circumferential spinal cord recording analysis

[0094] Integrating rich neural signal data in the time domain and spatial domain allows identification of unique topographical features that reflect activation of the major spinal tracts. A sequence of alternating left and right, MEP and SSEP stimulations is generated, producing a t-pattern corresponding to the expected sites of spinal tract activation. This allows clear visualization of where the nerve tracts are activated and produces unique topographical features, enabling accurate classification of the source signals.

[0095] For this test, the recording device was connected to a 32-channel recording probe (Intan Technologies, Los Angeles, CA, USA) via a custom omnetics / ZIF connector PCB and a flexible flat cable, and signals were acquired at 30 kHz through an RHS2000 stimulus / recording controller (Intan Technologies, Los Angeles, CA, USA) with a 50 Hz notch filter to reduce line noise from electrical current sources. The device was grounded through stainless steel wires implanted in the paravertebral muscles, and impedance measurements were obtained.

[0096] MEP stimulation was performed using 50-μm tungsten microelectrodes insulated with polyamide (California Fine Wire Company, CA, USA), which were fixed to a stereotaxic frame and grounded to a stainless steel cerebellar screw. The microelectrodes were guided to cortical areas representing hindlimb motor function as determined by previous mapping studies

[42] . Motor cortex stimulation was performed with a 200 Hz sequence of 20 alternating square wave pulses with a pulse width of 100 ms to maintain a short stimulation duration

[43] , starting at 50 μA and increasing in steps of 25 μA - 50 μA until the minimum stimulation energy required to elicit a barely perceptible muscle contraction, called the threshold level, was determined. If a satisfactory MEP was not obtained, the motor cortex stimulation area was offset 3 mm - 5 mm in the depth and / or rostral-caudal direction.

[0097] SSEP stimulation was performed using a custom nerve hook electrode that was applied to the sciatic, peroneal, and tibial nerves, anchored by a silicone sealant (Kwik-Sil TM , World Precision Instruments, Hitchin, UK), and grounded through a stainless steel wire placed at the base of the tail. The stimulation protocol consisted of a single alternating square wave pulse of 100 ms, starting at 50 μA and increasing in steps of 10 μA - 25 μA until the threshold level was determined.

[0098] EMG was recorded using electrodes made of stainless - steel 25 - gauge needles, which were inserted into the quadriceps muscle and grounded via a stainless - steel wire placed at the base of the tail. An alternating left / right, SSEP / MEP stimulation sequence was applied to provide data for analysis and machine learning. Each stimulation sequence lasted approximately 50 - 70 seconds, with a 1 - second interval between each stimulation pulse. All stimulation protocols were applied using an RHS2000 Stimulation / Recording Controller (Intan Technologies, CA, USA).

[0099] The raw signals were imported into MATLAB (R2021a) without further filtering to preserve key time - domain information. Active channels were referenced to resting channels to reduce electrocardiogram, respiration, stimulation, and movement artifacts. An automated peak detector was used to identify and demarcate the evoked compound action potential (ECAP) as a time - locked positive waveform after stimulation

[44] , and it was verified by visual inspection of the waveform to ensure that false signals were not included in the analysis. Since the signals in the 32 channels had different amplitudes at each sampling interval, a custom MATLAB script was used to plot the amplitudes in each channel around the spinal cord in a topographical heatmap, thereby visualizing the activation regions at each sampling interval. This topographical heatmap generated spatial features that roughly corresponded to the regions of potential spinal tract activation and could be used to correlate with the sources of the evoked potentials.

[0100] Figure 6 A shows representative recordings from four quadrants of the spinal cord interface covering the left dorsal, left ventral, right dorsal, and right ventral regions of the spinal cord during SSEP and MEP stimulation events. Also shown is the topographical heatmap at a single time point and the associated EMG recording.

[0101] The high sampling rate of the obtained recordings also enabled interrogation of the perievent potential amplitudes after the evoked potential, highlighting regions of peak amplitude shift around the spinal cord. After evaluating the topographical data following sciatic nerve SSEP stimulation, it was possible to identify multiple signal peaks consistent with previous studies [22, 23]. However, these peaks had different relative amplitudes in various channels, indicating that the topographical hotspots shifted over time after stimulation.

[0102] Figure 6 B is a plot from 4 representative channels showing how the signals recorded after sciatic nerve stimulation evolved over time. This phenomenon was reflected in most recordings, and the inventors believe that this shift reflects complex interneuronal interactions, where dorsal column neurons are activated and subsequently send impulses to ventral - based corticospinal and spinothalamic neurons.

[0103] At each sampling interval, circumferential spinal cord recordings produce topographic peak signal amplitudes distributed across 32-channel recordings, each sampling interval being unique for different evoked potential sources. A thresholding method based on the mean of the peak amplitudes allows for feature extraction of the signals, preparing them for data analysis. This allows the use of supervised machine learning methods to identify these features for classifying the evoked potential sources using the KNN algorithm. When classifying the source signals according to the following categories, this results in a classification accuracy of 93.8%: left MEP, right MEP, left SSEP, and right MEP. Figure 6 C is the classification matrix, showing the positive and negative detection rates of left / right, SSEP / MEP stimulation events.

[0104] The density of the signal information obtained from circumferentially placed electrodes also enables the further discrimination of the source signals down to the level of the peripheral nerve branches. The tibial nerve, which provides and receives sensorimotor input signals from the anterior compartment of the leg, and the peroneal nerve, which reflects the posterior lateral compartment of the leg, are stimulated in an alternating sequence. The stimulation of these two sciatic nerve branches produces unique spatial patterns represented by 32 channels on the electrodes, allowing for the classification of the source signals. Figure 6 D is the classification matrix showing the tibial and peroneal stimulation events.

[0105] Using linear and non-linear dimensionality reduction methods, automatic strategies for extracting signal features for classification were also investigated. After wavelet decomposition to identify cardiac and neural peaks, this allows unsupervised machine learning classification algorithms to identify different clusters, allowing for a comparable classification accuracy of the tibial / peroneal nerve stimulation source signals, although at a higher computational cost and processing time. This provides a second method for classifying the alternating stimulation of the peroneal and tibial nerves, as Figure 6 shown in E, Figure 6 E is a 3D graphical representation of the principal component analysis, showing how the tibial and peroneal stimulations are clustered based on the waveform shape of a single dorsally positioned electrode. Figure 6 A representative waveform comparison from an electrode located dorsally is shown in F, showing the temporal evolution of the average clustered waveforms of the tibial and peroneal stimulation events from a single dorsally located electrode.

[0106] The high-dimensional raw signal data obtained from the 32 channels need to be further processed before classification modeling. Two strategies were implemented. First, feature extraction was performed to characterize the peak amplitude of each ECAP

[46] . Then a threshold was determined based on the mean of the recorded peak amplitudes of each of the 32 recording channels, effectively converting the continuous raw signal data into classification variables.

[0107] In the second strategy, linear and nonlinear dimensionality reduction methods were used to reduce the dimensionality of the high-dimensional space corresponding to the waveforms extracted after wavelet denoising. In particular, principal component analysis [47, 48] (PCA) and uniform manifold approximation and projection

[49] (UMAP) were applied and compared. UMAP is a nonlinear method that, despite its higher computational cost, shows superior performance to linear PCA in datasets with significant nonlinear dependencies.

[0108] Dimensionality reduction is necessary because modeling with high-dimensional raw signal data leads to overfitting and reduced prediction performance

[50] , which can be addressed through feature extraction and dimensionality reduction.

[0109] The processed data was used to train a machine learning model, with 60% of the data divided into a training set and 40% into a classification set. For supervised machine learning, the k-nearest neighbor (KNN) algorithm

[51] with automatically optimized hyperparameters (k = 1, city block distance metric) was used in MATLAB (2021a). The classification data was aggregated to generate a confusion matrix to calculate the overall classification accuracy and classification loss rate for each evoked potential source. Tibial and peroneal SSEP classifications were aggregated separately as they are subsets of the left / right SSEP. Finally, an unsupervised KNN-based method with k = 2 (for the tibial and peroneal SSEP dataset) or k = 4 (for the alternating left / right, SSEP / MEP stimulation dataset) was used to automatically identify distinct clusters.

[0110] A spinal bypass was established after complete transection

[0111] In an embodiment of the present invention, by using the neural signals recorded by a device around the spinal cord located above (i.e., closer to the brain) the SCI site as a trigger to generate targeted spinal cord stimulation in a second device below (i.e., further away from the brain) this site, the dual functionality of the recording and stimulation capabilities of the devices described and illustrated above is applied to functionally bypass the site of acute spinal cord injury.

[0112] This method requires implanting two devices 10a, 10b at the T9-10 and T12-L1 levels respectively, as Figure 7 shown in A. Neural monitoring was used to ensure that the implantation did not cause iatrogenic SCI. Intermuscular EMG probes were connected to the lower limbs. MEP was generated using microfilaments placed in the motor cortex. An acute transection of the spinal cord was performed between the two devices 10a, 10b, where the absence of MEP recorded in the peripheral electromyogram (EMG) traces demonstrated the interruption of descending motor signals.

[0113] To generate low-latency communication between the recording and stimulation devices, an ECAP threshold detection method with low computational requirements is used. This allows the proximal recording device 10a to detect the peak of the MEP action potential and trigger the distal stimulation device 10b to achieve hindlimb movement. Figure 7 B schematically shows the workflow: the signal is recorded at T9, amplified, an amplitude threshold is applied and used to stimulate the corresponding electrodes below the lesion.

[0114] The stimulation algorithm can be adjusted to generate hindlimb movements with different muscle activation patterns. The electronic bypass can be easily deactivated and reactivated to allow differentiation between the resting and active states. To make this technique act as a physiological bypass, the delay between the recording and stimulation algorithms should be aimed to be comparable to the delay of neural transmission along the spinal cord.

[0115] Figure 7 C is example data showing the recorded signal from the ventral part of the spinal cord and the quadriceps / calf EMG waveforms after sub-lesion stimulation. The hindlimb angle is calculated based on the markerless kinematic data corresponding to the EMG and MEP waveforms.

[0116] Figure 7 D is a time-series video overlay of the lower limb activation using the above spinal bypass method.

[0117] Figure 7 E shows that the MEP delays calculated before and after spinal cord injury were determined (n = 6).

[0118] After implanting the recording and stimulation devices 10a, 10b, the Intan RHX data acquisition software (Intan Technologies, Los Angeles, California) is used to generate the control signal for the bypass in acute spinal cord injury. By examining the action potentials generated around the spinal cord by motor cortex stimulation and confirming their approximate location on the spinal cord, the selected channels are referenced to reduce ECG and respiratory artifacts, thereby denoising the action potential waveforms. These waveforms can be detected by their rising edges and their amplitudes are amplified using the native Intan RHX software to generate the trigger signal for spinal cord stimulation to reconstruct the interrupted hindlimb movement after acute SCI.

[0119] Migration to humans

[0120] For transfer to humans, the device was scaled to human size to explore the feasibility of surgical implantation. A parylene-C based mock-up device was used, i.e., a device with the same dimensions and mechanical properties as the human-scaled version of the above device, but without any electrical connections required for use as an electrical interface. First, the spinal cord was reconstructed using a saline pump as previously reported

[12] . Laminectomy was performed at the thoracic vertebrae T9 to T10 level. Subsequently, a midline dural incision was made to expose the spinal cord parenchyma. The dura mater was gently retracted using sutures to allow the surgeon to perform the operation in this space. Then the parylene device was passed laterally downwards from the side of the spinal cord, then the device was passed under the ventral side, and then the device was withdrawn from the opposite lateral side. The final result was that the device was wrapped around the spinal cord to complete the operation. When an electroactive device was used at this time, the interface could be used to record signals from the spinal cord or stimulate the spinal cord.

[0121] These devices were also tested in a human cadaver session at the Evelyn Surgery Training Centre in Cambridge (UK). Fresh frozen specimens were used. To achieve spinal reconstruction with saline, the dura mater was first exposed by making an incision above the cervical spinal cord, removing the spinous processes and performing a total laminectomy at the C4-6 vertebrae. The dura mater was incised with an 18-gauge needle, a polyethylene tube (OD = 1.4 mm) was inserted into the incision and secured using a purse-string suture (Prolene 4-0), and a small amount of cyanoacrylate glue (Loctite) was dispensed around the tube to ensure an adequate seal. The tubing was connected to a flow pump (Flowsteady, model 200). The pump was set to 140 mmHg at a rate of 1.5 L / min. After 30 seconds, the dura mater could be seen to expand, and after 60 seconds, fluid could be seen to flow out of the epidural space, indicating a small leak in the dura mater or fluid discharge when the dura mater expanded. The pump was reduced to a pressure of 70 mmHg and 0.5 L / min. The seal was monitored for at least 30 minutes before the first implantation to ensure that the seal was maintained and that the dura mater did not leak or expand further.

[0122] A midline incision was made over the thoracic vertebrae T9 to T10 and dissected down to the bone. Then a wide laminectomy was performed from pedicle to pedicle, followed by a midline dural incision. The dura mater was gently retracted using sutures and access to the dorsolateral spinal cord was allowed. The denticulate ligament was released to allow the device to slide unhindered around the spinal cord. To introduce the device without hitting the spinal cord, microforceps were used to guide the device laterally along the dura mater until it reached the ventral side of the spinal cord. Then the device was advanced until the tip could be seen and retrieved on the opposite side of the spinal cord.

[0123] To evaluate the potential for iatrogenic injury resulting from long-term implantation of a device according to an embodiment of the present invention, a study involving the implantation of an emulated (non-electroactive) device in six rats was conducted. The device was left in place for up to six weeks. During this period, the health and motor function of the animals were monitored. The experimental animals were terminated at six different endpoints. The spinal cord was carefully dissected and examined for signs of trauma or deformation. Microscopic examination showed that the spinal cord retained its native morphology and its staining characteristics, as Figure 8 shown in FIGS. 8A and 8B, while the locomotor ability of the animals remained unimpaired. The circularity index (CI) of the spinal cord of each rat was calculated and summarized in Figure 8 FIG. 8C. There was no significant difference in the average circularity index between the experimental rats and the control rats.

[0124] Discussion and Development

[0125] The above study has demonstrated that docking circumferentially around the spinal cord can comprehensively represent neural signals, generate control signals to trigger targeted spinal cord stimulation, and ultimately facilitate the electronic bypass of acute SCI. The device used is formed of thin and flexible bioelectronics that can be implanted around the spinal cord without causing nerve damage. The use of non-penetrating arrays has made these techniques possible, which limit damage and minimize the foreign body response secondary to long-term implantation compared to current rigid and bulky SCS

[24] .

[0126] Epidural electrodes minimize iatrogenic spinal cord injury

[28] , while recording information from the ascending and descending tracts of the spinal cord located in the periphery of the white matter. Previously, epidural recording of the ventral and ventrolateral tracts has been challenging because anatomical barriers (vertebral bodies, neural arches) prevent the safe insertion of recording probes without causing spinal cord injury, limiting the study of the spinothalamic and ventral corticospinal tracts, the latter being the major motor pathway in humans. The above surgical technique overcomes these anatomical barriers by precisely excising the neural arch to safely slide the device under the ventral spinal cord. While this is a necessary part of the device implantation process at the rodent scale, it may not be necessary for human placement due to the larger size.

[0127] Due to the somatotopic arrangement of the hindlimb muscle groups around the spinal cord, it is possible to achieve selective activation of specific hindlimb muscle groups. This is achieved by directly stimulating the ventral and lateral epidural spinal cord where the ventral corticospinal tract and premotor neurons are located. In addition, the pattern of muscle activation can be adjusted by changing the stimulation parameters to produce gentle, continuous, or bursty movements

[29] . This can enable neurorestorative procedures to achieve movements beyond current walking or stepping goal SCS algorithms, allowing SCI patients to perform specialized activities.

[0128] In the above method, a latency in communication between recording, decoding, and stimulation devices equivalent to physiological neural conduction latency has been achieved. This enables a more natural movement pattern for the user

[30] . In further embodiments, decoding the motor intent may require more complex filtering and classification algorithms, as well as potentially learning models that are trained to be unique to an individual patient

[10] .

[0129] Although acute nerve monitoring has confirmed that the acute implantation of these devices does not cause SCI, the dynamic interaction between the device and the spinal cord may cause injury. To ensure the long-term safety of these circumferential implant devices, this interaction can be mitigated by selecting thin and flexible materials that can accommodate movement within the spinal cord

[31] .

[0130] The above studies were all conducted on rats. However, the somatotopic distribution of the spinal cord tracts in rats is different from that in humans

[32] . Therefore, those skilled in the art will understand that the precise configuration of the device does not need to follow the above examples, and alternative arrangements are preferred for interfacing with the human spinal cord. In fact, considering the comprehensive representation of the spinal cord periphery with a circumferential implant interface, it is expected that a larger spinal cord size will allow for the design of arrays with a higher electrode density to better enhance the degree of selectivity in stimulation

[33] and recording.

[0131] With advancements in biomaterials and signal processing, neuroprosthetics according to embodiments of the present invention can potentially restore volitional motor function in SCI patients and serve as a key tool during rehabilitation or prognosis, especially for those suffering from incomplete injuries.

[0132] The above studies and tests have demonstrated that a trigger signal for targeted spinal cord stimulation can be generated using a low-latency electronic bypass through direct spinal cord recording, thereby restoring hindlimb movement in acute spinal cord injury. This is achieved by circumferentially implanting a thin and flexible device with a high electrode density to allow for a comprehensive representation and stimulation of the spinal cord. Thus, direct docking with the spinal cord represents a viable strategy for bypassing the site of injury in the spinal cord and restoring volitional motor function in SCI patients.

[0133] Alternative configurations

[0134] The above description has focused on devices that sense neural signals from a patient's spinal cord at one site and stimulate nerve bundles in the patient's spinal cord at another site. However, the general principle of bypassing the SCI site can be applied using devices located in different parts of the patient's anatomy, such as directly on an individual nerve.

[0135] Figure 9 Devices that can be used in alternative embodiments and implanted into different anatomical sites, such as nerves, are shown.

[0136] Figure 9 a is a photograph of the flexible nerve cuff array device on a cotton ball, highlighting its flexible and lightweight structure.

[0137] Figure 9 b is a photograph of the flexible nerve cuff array device. The cuffs are made of parylene-C (PaC). These carry electrode connections via PaC connectors, which are joined to a flat flexible cable (FFC) connector through which they are externalized. The shape of the PaC connectors is designed to conform to the shape and motion of the rat's shoulder and includes multiple tabs to facilitate surgical procedures. Those skilled in the art will understand that the design can be varied for implantation in other sites or in other animals.

[0138] Figure 9 c is a diagram showing the cuff and multi-electrode array layout. Each implantable device consists of three cuffs, designed for implantation in three separate nerves (ulnar, median, and radial nerves). Each cuff has a regular array of 100μm×100μm microelectrodes that are wound circumferentially around the nerve during implantation. To seal each cuff, its locking tab is fed through the locking hole of the cuff. In addition to the microelectrodes, each cuff contains larger electrodes that can be used for whole nerve stimulation / recording and for grounding purposes during nerve stimulation.

[0139] Figure 9 d is a diagram showing the naming convention of the microelectrodes (showing the radial nerve cuff). It should be noted that due to the limitation of the number of device channels, the radial nerve cuff lacks the R(1,2) microelectrodes.

[0140] Figure 9 e shows, diagrammatically, an image of the long-term implantation model and the device implanted in the rat brachial plexus. The microelectrode connections of the implanted cuffs are externalized via a subcutaneously positioned FFC, exiting the rat's body through the headcap port. For recording, a ground electrode is implanted in the CSF above the rat's cerebellum.

[0141] Figure 9 f shows, schematically, the structure of the flexible nerve cuff array device. The device has gold tracks encapsulated between two layers of 4μm PaC. The tracks are connected to PEDOT:PSS microelectrodes.

[0142] Figure 9 g shows the impedance of the device microelectrodes after fabrication (n = 343 microelectrodes, N = 12 devices). Scale bars a-b: 10 mm, c: 2 mm.

[0143] Figure 10 Shows Figure 9 The foreign body response analysis of the flexible PaC nerve cuffs described and shown herein, and shows that these cuffs exhibit minimal fibrosis four weeks after implantation.

[0144] Figure 10 a is a collection of photographs of the cuff and nerve removed four weeks after implantation and immunohistochemical cross-sectional images of the nerve. Compared with the stiffer polyethylene (PE) cuff and the soft silicone (PDMS) cuff, the nerve implanted with the ultra-flexible PaC cuff showed little fibrosis (αSMA).

[0145] Figure 10 b provides quantification of fibrosis (αSMA staining) at different depths along the circumference of the nerve. A distance of 0 from the edge represents the tissue in direct contact with the cuff wall. The staining intensity is expressed as the staining ratio relative to the non-fibrotic tissue at the center of the nerve. The flexible PaC cuff induced less fibrosis compared with the PE and PDMS cuffs, and there was little fibrosis above the background level. The lines represent the mean between groups, and the shaded areas represent the standard deviation. N = 5 rats.

[0146] Figure 10 c provides quantification of fibrosis (αSMA staining) in the 25-μm nerve tissue closest to the cuff. The degree of fibrosis induced by the flexible PaC cuff was significantly lower than that of the PE cuff (p = 0.0029, ANOVA and Tukey post hoc test). The bar graph represents the mean between groups, and the circles represent the values of individual animals.

[0147] Figure 11 It is shown that the flexible nerve cuff array can achieve stable recording of the target nerve for a long time.

[0148] Figure 11 a is a series of photographs of the behavioral tasks performed by the animal during electrophysiological recording. The animal was recorded while walking in a transparent tunnel. This allowed the identification of nerve activities related to the swing phase and stance phase of walking of the implanted lower limb.

[0149] Figure 11 b shows representative traces of the recorded activities from the median cuff of a rat within 21 days of implantation. Individual spikes were visible at all time points (the interval marked by the red line in the right figure is enlarged to the trace in the left figure). Scale bars: 100 ms and 10 μV (left), and 5 ms and 10 μV (right).

[0150] Figure 11c shows the microelectrode impedance and survival during implantation. Implantation caused an increase in impedance and rupture of 7% of the microelectrodes, but these values remained stable during the 21-day implantation period. Functional microelectrodes were defined as having an impedance <1 MOhm. n = 96 (before implantation), 89 (days 3 - 7), 88 (day 14), and 87 (day 21) microelectrodes, N = 3 rats. The experiment was specifically designed to terminate at 21 days. One rat was terminated at 14 days post-implantation due to skin damage caused by the device FFC connector.

[0151] Figure 11 d shows representative nerve traces (black) and average RMS values recorded simultaneously from the median, ulnar, and radial nerves 3 days post-implantation. Increased nerve activity can be seen during the stance phase of the implanted limb for the median and ulnar nerves and during the swing phase for the radial nerve.

[0152] Figure 11 e shows in Figure 11 the results of spike sorting performed on the traces in d.

[0153] Figure 12 It shows that the flexible cuff array can achieve sub-nerve recording resolution without penetrating the nerve.

[0154] Figure 12 a is a layout diagram of a row of microelectrodes circumferentially around the radial nerve.

[0155] Figure 12 b shows spike events on four microelectrodes within a row circumferentially along the radial nerve. Nerve recordings (top) are provided for reference. In all microelectrodes, bursty spikes were detected, coinciding with the swing gait phase during rat walking. Some spikes were detected on all microelectrodes, while others were uniquely detected in some of the microelectrodes.

[0156] Figure 12 c is a quantification of spike coincidence across microelectrodes for the activity shown in Figure 12 b. Most nerve spikes were detected simultaneously on two microelectrodes or uniquely by a single microelectrode. These unique spikes were recorded by all four microelectrodes (proportion decomposition as shown). Co-detection was quantified by the number of spikes, rather than by co-detection events (four spikes co-detected on all four microelectrodes were counted as an event of four spikes).

[0157] Figure 12 d is a layout diagram of microelectrodes positioned along the length of the radial nerve.

[0158] Figure 12e shows the neural activity recordings from two columns of microelectrodes. The neural action potential spikes recorded on the two rings exhibit a time delay corresponding to their direction and velocity. This is shown at the bottom of the subfigure.

[0159] Figure 12 f shows the quantification of the radial nerve activity delay between the microelectrodes within the nearest and farthest columns in the cuff array during a 2.7 - second recording. Top panel: Quantification of the cross - correlation values between neural recordings corresponding to two microelectrode columns at different time delays. Bottom panel: Histogram of the inter - spike intervals between the microelectrode columns. Calculated based on the 2 - mm distance between the microelectrode rings, both analyses identify peaks in activity at the delays corresponding to afferent nerves of approximately 1.6 m / s, approximately 12 m / s, and an efferent nerve of 1.35 m / s. Fast afferent / efferent nerve activity is also detected at delays close to 0 milliseconds.

[0160] Figure 12 It shows that the flexible cuff array can achieve sub - neural stimulation resolution without penetrating the nerve.

[0161] Figure 13 a is a diagram of the experimental setup. Under anesthesia, using a single microelectrode to stimulate three implanted nerves results in movements of the wrist and finger of the paw of the rat.

[0162] Figure 13 b is a series of pictures of examples of the movements generated before (left) and during (right) nerve stimulation.

[0163] Figure 13 c is a heatmap of the quantification of the movements generated by microelectrode stimulation in five rats. Although stimulation sometimes elicits simple movements, more commonly, combined (names containing “+”) movements are observed. Some simple movements are never observed. Despite the differences in movements among different implants and animals, more than three movements are observed in each case, indicating a resolution higher than whole - nerve stimulation.

[0164] Figure 13 d is the quantification of the types of movements generated by stimulating three different nerves using microelectrodes.

[0165] Figure 13 e is a plot of the principal components of the kinematic analysis of the various movements generated by Rat 1. Left panel: The movements are labeled according to the microelectrodes that generated them. Clustering can be seen between the groups of microelectrodes within each cuff. Right panel: The movements are labeled by the recognized types of the movements generated. The clustering indicates that many recognized types of movements can be executed with different kinematics, indicating Figure 13 an additional stimulation selectivity not captured in c.

[0166] Figure 13f is the ground truth matrix of the movements of Rat 1 classified using a k-nearest neighbor classifier. WF: Wrist flexion, WE: Wrist extension, FE: Finger extension, UD: Ulnar deviation, SD: Sacral deviation, S: Supination, P: Pronation, WF+FF: Wrist flexion and finger flexion, WF+FE: Wrist flexion and finger extension, WF+UD: Wrist flexion and ulnar deviation, WE+FE: Wrist extension and finger extension, FE+UD: Finger extension and ulnar deviation, Total number of movements: The total number of different movements generated.

[0167] System configuration

[0168] The functional elements of the system according to an embodiment of the present invention are shown in Figure 14 . The system includes a recording interface 32 and a stimulation interface 34. These interfaces can be provided, for example, by implantable devices 10a, 10b having electrode arrays, such as those previously described.

[0169] Interfaces 32, 34 are connected to a controller 30, which can include: one or more processors adapted to run the various processes and methods described herein, and one or more non-volatile memories (configured to store code to cause the processors to run those processes and methods, and / or to store data generated by the operation of the system).

[0170] Controller 30 can have modules that perform specific tasks, such as signal amplification, signal filtering, signal processing, computing, interpretation, pulse generation, etc. Such modules can be provided by separate processors, or combined in any combination in the operation of one or more processors. Controller 30 also has a power management module, which is configured to supply power to the controller itself and interfaces 32, 34. The power management module is connected to a power supply unit 40.

[0171] In some embodiments, power supply unit 40 can be a battery. In other embodiments, power supply unit 40 is a wireless power receiver, which is configured to receive power wirelessly from an external source. This arrangement allows controller 30, interfaces 32, 34, and power supply unit 40 to be implanted inside a patient and receive power from a source outside the patient, thus eliminating the need to replace the battery. In other embodiments, power supply unit 40 provides a direct wired connection to a power source.

[0172] The foregoing description is exemplary in nature, and those skilled in the art will understand that changes and variations to the disclosed embodiments are possible within the scope of the claims. The claims define the invention.

[0173] The project that gave rise to this application has received funding from the European Union's Horizon 2020 research and innovation program (grant agreements No. 685474 and No. 964677).

[0174] References

[0175] 1. Wagner, F. B. et al. Targeted neurotechnology restores walking in humans with spinal cord injury. Nature 563, 65 - 71 (2018).

[0176] 2. Angeli, C. A. et al. Recovery of Over - Ground Walking after Chronic Motor Complete Spinal Cord Injury. N Engl J Med 379, 1244 - 1250 (2018).

[0177] 3. Gill, M. L. et al. Neuromodulation of lumbosacral spinal networks enables independent stepping after complete paraplegia. Nat Med 24, 1677 - 1682 (2018).

[0178] 4. Capogrosso, M. et al. A brain - spine interface alleviating gait deficits after spinal cord injury in primates. Nature 539, 284 - 288 (2016).

[0179] 5. Formento, E. et al. Electrical spinal cord stimulation must preserve proprioception to enable locomotion in humans with spinal cord injury. Nat Neurosci 21, 1728 - 1741 (2018).

[0180] 6. Sharpe, A. N. & Jackson, A. Upper-limb muscle responses to epidural, subdural and intraspinal stimulation of the cervical spinal cord. J. Neural Eng. 11, 016005 (2014).

[0181] 7. Hogan, M. K. et al. A wireless spinal stimulation system for ventral activation of the rat cervical spinal cord. Sci Rep 11, 14900 (2021).

[0182] 8. Guiho, T., Baker, S. N. & Jackson, A. Epidural and transcutaneous spinal cord stimulation facilitates descending inputs to upper-limb motoneurons in monkeys. J. Neural Eng. 18, 046011 (2021).

[0183] 9. Prasad, A. & Sahin, M. Can motor volition be extracted from the spinal cord?J NeuroEngineering Rehabil 9, 41 (2012).

[0184] 10. Fathi, Y. & Erfanian, A. Decoding hindlimb kinematics from descending and ascending neural signals during cat locomotion. J Neural Eng (2021) doi:10.1088 / 1741-2552 / abd82a.

[0185] 11. Song, E., Li, J., Won, S. M., Bai, W. & Rogers, J. A. Materials for flexible bioelectronic systems as chronic neural interfaces. Nature Materials 19, 590 - 603 (2020).

[0186] 12. Woodington, B. J. et al. Electronics with shape actuation for minimally invasive spinal cord stimulation. Science Advances 7, eabg7833 (2021).

[0187] 13. Nelson, M. J., Valtcheva, S. & Venance, L. Magnitude and behavior of cross - talk effects in multichannel electrophysiology experiments. J Neurophysiol 118, 574 - 594 (2017).

[0188] 14. Khodagholy, D. et al. In vivo recordings of brain activity using organic transistors. Nature Communications 4, 1575 (2013).

[0189] 15. Castagnola, V. et al. Parylene - based flexible neural probes with PEDOT coated surface for brain stimulation and recording. Biosensors and Bioelectronics 67, 450 - 457 (2015).

[0190] 16. Thirumala, P. et al. Diagnostic accuracy of evoked potentials for functional impairment after contusive spinal cord injury in adult rats. J Clin Neurosci 25, 122 - 126 (2016).

[0191] 17. Volkov, A. V. et al. Understanding the Capacitance of PEDOT:PSS. Advanced Functional Materials 27, 1700329 (2017).

[0192] 18. Jiang, L., Woodington, B., Carnicer - Lombarte, A., Malliaras, G. & Barone, D. G. Spinal cord bioelectronic interfaces: opportunities in neural recording and clinical challenges. J. Neural Eng. 19, 021003 (2022).

[0193] 19. Webb, A. A. & Muir, G. D. Unilateral dorsal column and rubrospinal tract injuries affect overground locomotion in the unrestrained rat. European Journal of Neuroscience 18, 412 - 422 (2003).

[0194] 20. Raineteau, O., Fouad, K., Bareyre, F. M. & Schwab, M. E. Reorganization of descending motor tracts in the rat spinal cord. Eur J Neurosci 16, 1761 - 1771 (2002).

[0195] 21.Diedrich,A.,Charoensuk,W.,Brychta,R.J.,Ertl,A.C.&Shiavi,R.Analysisof RawMicroneurographic Recordings Based on Wavelet De-Noising Technique andClassificationAlgorithm:Wavelet Analysis in Microneurography.IEEE TransBiomed Eng 50,41-50(2003).

[0196] 22.Strahm,C.,Min,K.,Boos,N.,Ruetsch,Y.&Curt,A.Reliability ofperioperative SSEPrecordings in spine surgery.Spinal Cord 41,483-489(2003).

[0197] 23.Parker,J.L.,Shariati,N.H.&Karantonis,D.M.Electrically evokedcompound actionpotential recording in peripheral nerves.Bioelectronics inMedicine 1,71-83(2018).

[0198] 24.Carnicer-Lombarte,A.,Chen,S.-T.,Malliaras,G.G.&Barone,D.G.ForeignBodyReaction to Implanted Biomaterials and Its Impact in Nerve Neuroprosthetics.Front.Bioeng.Biotechnol.9,(2021).

[0199] 25.Prasad,A.&Sahin,M.Characterization of neural activity recordedfrom the descendingtracts of the rat spinal cord.Front.Neurosci.4,(2010).

[0200] 26. Streeter, K. A. et al. Coupling multielectrode array recordings with silver labeling of recording sites to study cervical spinal network connectivity. J Neurophysiol 117, 1014 - 1029 (2017).

[0201] 27. Barrese, J. C., Aceros, J. & Donoghue, J. P. Scanning electron microscopy of chronically implanted intracortical microelectrode arrays in non - human primates. J Neural Eng 13, 026003(2016).

[0202] 28. Noga, B. et al. Field potential mapping of neurons in the lumbar spinal cord activated following stimulation of the mesencephalic locomotor region. J. Neurosci. 15, 2203 - 2217(1995).

[0203] 29. Ichiyama, R. M., Gerasimenko, Y. P., Zhong, H., Roy, R. R. & Edgerton, V. R. Hindlimb stepping movements in complete spinal rats induced by epidural spinal cord stimulation. Neurosci Lett 383, 339 - 344(2005).

[0204] 30. J.et al.Low Latency Estimation of Motor Intentions to AssistReachingMovements along Multiple Sessions in Chronic Stroke Patients:AFeasibility Study.Frontiers inNeuroscience 11,126(2017).

[0205] 31.Lacour,S.P.,Courtine,G.&Guck,J.Materials and technologies for softimplantableneuroprostheses.Nature Reviews Materials 1,1-14(2016).

[0206] 32.Saliani,A.et al.Axon and Myelin Morphology in Animal and HumanSpinal Cord.Frontiers in Neuroanatomy 11,129(2017).

[0207] 33. O.R.M.,Hyttinen,J.A.K.,Laarne,P.H.&Malmivuo,J.A.Effectofelectrode density and measurement noise on the spatial resolution ofcortical potential distribution.IEEE Trans Biomed Eng 51,1547-1554(2004).

[0208] 34.Rivnay,J.et al.Structural control of mixed ionic and electronictransport in conductingpolymers.Nat Commun 7,11287(2016).

[0209] 35. Sorrenti, V. et al. Understanding the Effects of Anesthesia on Cortical Electrophysiological Recordings: A Scoping Review. Int J Mol Sci 22, (2021).

[0210] 36. Mostany, R. & Portera-Cailliau, C. A Craniotomy Surgery Procedure for Chronic Brain Imaging. J Vis Exp 680 (2008) doi:10.3791 / 680.

[0211] 37. Keast, J.R., Forrest, S.L. & Osborne, P.B. Sciatic nerve injury in adult rats causes distinct changes in the central projections of sensory neurons expressing different glial cell line-derived neurotrophic factor family receptors. J Comp Neurol 518, 3024 - 3045 (2010).

[0212] 38. Batt, J.A.E. & Bain, J.R.T. Tibial Nerve Transection - A Standardized Model for Denervation-induced Skeletal Muscle Atrophy in Mice. J Vis Exp 50657 (2013) doi:10.3791 / 50657.

[0213] 39. Shah, S.B. et al. Does cryoneurolysis result in persistent motor deficits?A controlled study using a rat peroneal nerve injury model. Reg Anesth Pain Med 45, 287 - 292 (2020).

[0214] 40. Gelderd, J. B. & Chopin, S. F. The vertebral level of origin of spinal nerves in the rat. Anat Rec 188, 45 - 47 (1977).

[0215] 41. Pelot, N. A., Thio, B. J. & Grill, W. M. Modeling Current Sources for Neural Stimulation in COMSOL. Frontiers in Computational Neuroscience 12, (2018).

[0216] 42. Brown, A. R. & Martinez, M. Ipsilesional Motor Cortex Plasticity Participates in Spontaneous Hindlimb Recovery after Lateral Hemisection of the Thoracic Spinal Cord in the Rat. J. Neurosci. 38, 9977 - 9988 (2018).

[0217] 43. Watson, M., Sawan, M. & Dancause, N. The Duration of Motor Responses Evoked with Intracortical Microstimulation in Rats Is Primarily Modulated by Stimulus Amplitude and Train Duration. PLOS ONE 11, e0159441 (2016).

[0218] 44. Fehlings, M. G., Tator, C. H., Linden, R. D. & Piper, I. R. Motor evoked potentials recorded from normal and spinal cord - injured rats. Neurosurgery 20, 125 - 130 (1987).

[0219] 45. Donoho, D. L. & Johnstone, I. M. Ideal Spatial Adaptation by Wavelet Shrinkage. Biometrika 81, 425 - 455 (1994).

[0220] 46. Campbell, E., Phinyomark, A. & Scheme, E. Feature Extraction and Selection for Pain Recognition Using Peripheral Physiological Signals. Frontiers in Neuroscience 13, 437 (2019).

[0221] 47. Multivariate Statistical Data Analysis - Principal Component Analysis (PCA). International Journal of Livestock Research http: / / ijlr.org / issue / multivariate - statistical - data - analysis - principal - component - analysis - pca / .

[0222] 48. Jolliffe, I. T. & Cadima, J. Principal component analysis: a review and recent developments. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences 374, 20150202 (2016).

[0223] 49. McInnes, L., Healy, J. & Melville, J. UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction. arXiv:1802.03426 [cs, stat] (2020).

[0224] 50. Pestov, V. Is the k-NN classifier in high dimensions affected by the curse of dimensionality? Computers & Mathematics with Applications 65, 1427-1437 (2013).

[0225] 51. Li, M., Xu, H., Liu, X. & Lu, S. Emotion recognition from multichannel EEG signals using K-nearest neighbor classification. Technol Health Care 26, 509-519.

[0226] 52. Mathis, A. et al. DeepLabCut: markerless pose estimation of user-defined body parts with deep learning. Nat Neurosci 21, 1281-1289 (2018).

[0227] All of the above references are hereby incorporated by reference in their entirety into this text.

Claims

1. A system for treating a patient with a spinal cord injury, the system comprising: A first flexible implantable device having a first plurality of electrodes; A second flexible implantable device having a second plurality of electrodes; And A controller communicatively coupled to the electrodes of the first and second implantable devices and configured to receive an input signal from the electrodes of the first implantable device, process the input signal to generate a stimulation signal, and transmit the stimulation signal to the electrodes of the second implantable device.

2. The system according to claim 1, wherein Either or both of the first and second implantable devices are configured to conform to the patient's spinal cord or nerves when implanted in the patient's body.

3. The system according to claim 2, wherein Either or both of the first and second implantable devices have a gel layer disposed on the surface of the device opposite the surface of the device having the plurality of electrodes.

4. The system according to any one of the preceding claims, wherein, Either or both of the first and second implantable devices are configured to substantially surround the patient's spinal cord or nerves when implanted in the patient's body.

5. The system according to any one of the preceding claims, wherein, The first and second implantable devices each have a distal portion, and each distal portion includes a corresponding plurality of electrodes. Further, the distal portions of the first and second implantable devices are the same.

6. The system according to any one of claims 1 to 4, wherein, The first implantable device is configured for implantation in the patient's brain.

7. The system according to any one of claims 1 to 4 or claim 6, wherein, The second implantable device is configured for implantation in the patient's limb.

8. The system according to any one of the preceding claims, wherein The controller is configured to: Process the input signal by decoding the input signal to determine an expected stimulation of the patient's nerves at the cephalic end of the position of the first device on the spinal cord; And Generate a stimulation signal that, when applied to the patient's spinal cord through the second device, produces the expected stimulation of the patient's nerves.

9. The system according to any one of the preceding claims, wherein, The controller is configured to process the input signal by integrating the input signal in the time domain and the spatial domain.

10. The system according to claim 9, wherein, The controller is further configured to identify unique topographical features from the integrated input signal to identify the major spinal cord tracts that are activated.

11. A method for treating a patient with a spinal cord injury, the method comprising the steps of: Percutaneously implanting a first plurality of electrodes at a position upstream of the site of the spinal cord injury; Percutaneously implanting a second plurality of electrodes at a position downstream of the site of the spinal cord injury; Processing a signal from one or more of the first plurality of electrodes to determine an expected stimulation of the patient's nerves downstream of the spinal cord injury location; Generating a stimulation signal that, when applied to the patient's nerves through one or more of the second plurality of electrodes, produces the expected stimulation of the patient's nerves.

12. The method according to claim 11, wherein, Implanting one or both of the first plurality of electrodes and the second plurality of electrodes adjacent to the patient's spinal cord or nerves.

13. The method according to claim 12, wherein, Implanting one or both of the first plurality of electrodes and the second plurality of electrodes such that they substantially surround the patient's spinal cord or nerves.

14. The method according to any one of claims 11 to 13, wherein, Implanting the first plurality of electrodes in the patient's brain.

15. The method according to any one of claims 11 to 14, wherein Implanting the second plurality of electrodes in the patient's limb.

Images