Muscle probes, systems and methods

Abstract

A muscle probe for obtaining electromyography (EMG) and optical spectroscopy data from muscle tissue is provided. The muscle probe includes an elongated needle having an outer wall surrounding an interior needle portion, the interior needle portion including a core electromyography electrode; one or more optical fibers; the needle is positioned for insertion into a muscle and further configured to detect electrical activity from the muscle; and the one or more optical fibers are positioned to direct incident light from a light source toward a target region of the muscle and further configured to receive scattered light from the target region. The present disclosure aims to provide a muscle probe that improves the diagnostic pathway for patients with neuromuscular disorders by developing a minimally invasive bedside test of muscle health.

Description

[Technical Field]

[0001] The present disclosure relates to muscle probes, and more particularly to muscle probes for obtaining electromyographic and optical spectral data from muscle tissue. [Background technology]

[0002] Neuromuscular disorders result in muscle weakness and cause significant morbidity and mortality, but diagnosis often takes years and requires invasive testing. Better diagnostic pathways are needed. Current primary diagnostic tests include needle electromyography, muscle biopsy, and genetic analysis.

[0003] Many of these diseases require time-consuming diagnostic procedures, invasive biopsies, and are therefore only performed in highly specialized centers, resulting in missed opportunities for intervention and clinical trial participation.

[0004] Electromyography examines the electrical activity generated by muscles. It has low specificity for individual diseases and its interpretation is highly subjective. Due to its subjective nature, variability, and limited quantitative output, electromyography is not typically used to monitor disease progression in neuromuscular diseases. Because electromyography examines only the electrical activity of muscles, it is "blind" to other pathological changes, such as cellular infiltration, that are not normally present in muscles. Electromyography is also a specialized tool that must be performed by skilled and experienced staff, which increases costs and limits appointment availability.

[0005] Traditional muscle biopsies can be used to determine the presence or absence of neuromuscular disorders, but typically only one muscle is biopsied. This muscle may miss areas of pathology, and other muscles may be better targets. Due to its invasive nature, muscle biopsies are not typically used to monitor patients in clinical settings and are rarely used to monitor treatment response in clinical trials.

[0006] It is therefore desirable to provide a more accurate and less invasive alternative to currently available techniques. Summary of the Invention [Problem to be solved by the invention]

[0007] The present invention provides a combination of electromyography (EMG) and optical spectroscopy to improve the diagnostic pathway for patients with neuromuscular disorders by developing a minimally invasive, bedside test for muscle health. The optical spectroscopy of the present invention preferably includes Raman spectroscopy. This solution combines EMG testing of multiple muscles with the molecular specificity typically obtained from traditional muscle biopsies (in this case using optical spectroscopy), but without the invasiveness and single-muscle sampling typically associated with traditional muscle biopsy procedures. This combination also provides improved biomarkers of muscle health and / or disease that can be used to diagnose disease, monitor disease progression, and predict disease prognosis and / or treatment response.

[0008] According to a first aspect of the present disclosure, there is provided a muscle probe comprising an elongated needle having an outer wall surrounding an interior needle portion, the interior needle portion including a core electromyography electrode; and one or more optical fibers; the needle is positioned for insertion into a muscle and further configured to detect electrical activity from the muscle, the one or more optical fibers being positioned to direct incident light from a light source towards a target region of the muscle and further configured to receive scattered light from the target region.

[0009] The needle is preferably intended to function as a typical electromyography needle, as will be understood by those skilled in the art. In particular, the needle is preferably used to detect muscle membrane depolarization, where a potential difference is observed between the outer wall of the needle (acting as a "reference") and the core (acting as the active electrode).

[0010] The use of electromyography in the present probe includes confirmation that the probe is within the muscle, particularly specific muscles that are difficult to accurately target using surface palpation alone. The probe can also be guided in real time to target muscle regions, which may include areas exhibiting electrical (or EMG) abnormalities and areas of muscle that appear normal (as determined by an EMG-trained operator). Once at the target region, optical spectroscopic evaluation can be performed using one or more optical fibers. Thus, in a preferred embodiment, the scattered light includes inelastic scattered light for evaluation using optical spectroscopy. The inelastic scattered light preferably includes one or more of: Raman scattered light; fluorescence scattered light; and Brillouin scattered light. The spectroscopic evaluation preferably determines the molecular composition of the muscle for improved assessment of muscle health and / or neuromuscular disease status, and preferably provides a data fingerprint that includes a characterization of the molecular composition. Thus, the spectroscopic evaluation can be precisely guided to regions of interest within the muscle, and in some cases, to specific muscles that are difficult to accurately target using palpation alone. Furthermore, the probe can be used such that the optical spectrum is used to guide EMG evaluation.

[0011] The use of such real-time EMG-guided biochemical assessment of target muscle regions preferably avoids the need for traditional muscle biopsies, which are painful, time-consuming, and labor-intensive. The improved specificity and accuracy of the combination of the present invention preferably reduces the time required for EMG assessment while simultaneously reducing the number of muscles required for testing, preferably leading to reduced costs and improved patient comfort.

[0012] Determining muscle health and / or disease state, along with prognosis and / or potential treatment response, using EMG assessment alone typically requires a skilled practitioner and involves subjective interpretation of the EMG data. The combination of EMG data with optical spectroscopy capabilities, such as Raman spectroscopy, of the present invention provides an objective assessment of the molecular composition of target muscle regions, and is therefore preferably suitable for use outside of a clinical neurophysiology clinic for diagnostic, triage, and disease monitoring purposes.

[0013] Optical spectroscopic data is potentially sensitive and / or specific, and therefore can be advantageously used to improve diagnosis and monitoring of disease progression, among other things.

[0014] In some embodiments, the muscle probe preferably further comprises a cannula extending along the interior of the needle, with the core EMG electrode and / or one or more optical fibers housed within the cannula. In preferred embodiments, the core EMG electrode is formed from at least a portion of the cannula. In some such embodiments, the cannula and the core EMG electrode are the same. In preferred embodiments, the cannula is electrically insulated from the outer wall of the needle, which may itself form the needle electrode. In some such embodiments, the cannula (which may itself be the core EMG electrode) preferably includes an electrically insulating coating.

[0015] The cannula is preferably positioned to move along the interior of the needle. In this way, the cannula functions as a sealed environment separate from the outer wall of the needle and can move along the inner wall of the needle to engage the target muscle tissue or region. It will be appreciated that in alternative embodiments, any components housed within the cannula, such as the active electrode and / or one or more optical fibers, can move independently within the cannula to engage the target muscle tissue or region. In some embodiments, it may be preferable for the core electromyography electrode to form a coating disposed on at least one of the optical fibers. Preferably, coating one or more of the optical fibers with the active electrode further protects the coated optical fibers and improves the ease of manufacturing the muscle probe.

[0016] The outer wall and / or cannula of the needle can function as a protective layer for the active electrode and / or one or more optical fibers contained therein. The movement of the active electrode and one or more optical fibers relative to the outer wall or cannula of the needle preferably allows the active electrode and / or one or more optical fibers to remain protected within the needle while the needle is moved toward the desired muscle tissue region. The movement then allows the active electrode and / or one or more optical fibers to engage muscle tissue for electromyographic or spectroscopic evaluation.

[0017] In some embodiments, the one or more optical fibers preferably comprise: at least one delivery fiber positioned to direct incident light from the light source toward a target region of the muscle; and at least one collection fiber positioned to receive scattered light from the target region. In preferred embodiments, the one or more optical fibers comprise more collection fibers than delivery fibers. Such embodiments maximize the collection area for receiving scattered light from the target region, thereby optimizing the collection of spectroscopic data. In some preferred embodiments, the one or more optical fibers comprise a single delivery fiber and at least three collection fibers. Such an arrangement can maximize collection capacity while optimizing the form factor for use within the needle. Embodiments in which any suitable number of collection and delivery fibers are used will be understood.

[0018] One or more optical fibers may, in some embodiments, be equipped with an in-line filter, for example, to reduce the effect of complex light (e.g., elastic or Rayleigh scattered light), or indeed scattered light and any associated fluorescence, on at least one delivery fiber, and further reduce the effect of any such complex light on at least one collection fiber. In some embodiments, at least one delivery fiber and / or at least one collection fiber preferably comprises one of: an in-line shortpass filter; an in-line bandpass filter; an in-line longpass filter; or a notch filter. It will be understood that the terms "longpass," "shortpass," "bandpass," and "notch" in the described filters refer to the effect of the filter on light of a particular wavelength, rather than its relationship to wave frequency.

[0019] In certain embodiments, at least one delivery fiber preferably includes an in-line bandpass or shortpass filter. The filter on the at least one delivery fiber is preferably designed to allow transmission of only light of a specific wavelength (e.g., the wavelength of the light source, which in some embodiments may be a laser). Thus, the particular filter used may be selected depending on the light source. In the case of a laser, a filter may be selected to substantially limit transmission of light of wavelengths 785 nm or 830 nm. It will be understood that embodiments in which a filter is used for any suitable wavelength of the light source are suitable. Such a filter preferably reduces the intensity of light inelastically scattered within the fiber and prevents it from reaching the target muscle region where spectroscopic measurements are performed. Such a filter may be an in-line filter located near the end of the at least one delivery fiber or a filter coated directly or near the end of the at least one delivery fiber.

[0020] Filters coated directly on the fibers may be easier to manufacture at scale. In embodiments where the scattered light includes Raman scattered light, bandpass or shortpass filters may be used to observe only the Raman scattered light for longer wavelengths (with a Stokes shift), although embodiments intended to observe Raman scattered light for both longer and / or shorter wavelengths (without a Stokes shift) are also understood, and in such embodiments it may be preferable to include both a bandpass (or shortpass) filter and a notch filter on at least one delivery fiber.

[0021] In certain embodiments, at least one collection fiber preferably includes an in-line long-pass or notch filter. The filter on the at least one collection fiber is preferably designed to reject incident light of a wavelength equal to a particular wavelength (of the light source) and can therefore be selected depending on the light source (e.g., the wavelength of the light source, which in some embodiments may be a laser). The particular filter used can therefore be selected depending on the light source. In the case of a laser, a filter can be selected that substantially limits the transmission of light of wavelengths 785 nm or 830 nm. This reduces the intensity of elastically scattered light in the sample that returns along the at least one collection fiber and may serve as a source of Raman scattering and fluorescence-related scattering from the fiber itself. Such scattering may confound the intended spectroscopic measurement of inelastically scattered light from the target muscle region. Such a filter can be an in-line filter positioned near the end of the at least one collection fiber or a filter coated directly or near the end of the at least one collection fiber. Filters coated directly on the fiber may be easier to fabricate at scale.

[0022] In some embodiments, the outer wall of the needle preferably forms a conductive tube, and one or more optical fibers each have a light-transmitting end and / or a light-receiving end disposed proximate the distal end of the conductive tube. In embodiments in which the needle is a monopolar electromyography needle, the outer wall of the needle is relatively non-conductive. It will be appreciated that the optical fibers are contained within the needle outer wall and / or cannula such that the needle outer wall or cannula encases at least a portion of the optical fiber intended for insertion into the muscle while enabling the light emission and collection necessary for optical spectroscopic assessment of the target muscle region. Thus, the needle outer wall and / or cannula has an open end so that the incident light emitted by the one or more optical fibers is not obstructed. The needle outer wall and / or cannula preferably provides protection for the optical fibers against damage during use; therefore, without the needle outer wall and / or cannula, it would be difficult to obtain such spectral data without significant injury and potential complications. The needle outer wall may comprise any suitable conductive material, and in some embodiments, the needle outer wall preferably comprises steel. In embodiments with monopolar electromyography needles, the outer wall of the needle may comprise any suitable polymer, such as polytetrafluoroethylene (PTFE). The core electrode may comprise any suitable material for performing the function of the active electrode of an electromyography device, and in some embodiments, the core electrode may preferably comprise any of silver, platinum, steel, stainless steel, nickel, chromium, iridium, titanium, or any alloy thereof (e.g., nitinol or nichrome silver). In some embodiments, the muscle is preferably striated or non-striated (smooth) muscle. In the most preferred embodiment, the muscle is striated muscle.

[0023] The ends of the one or more optical fibers, facing the optical transmitting end and / or the optical receiving end, are preferably arranged to communicate with an optical spectrometer. In some embodiments, the one or more optical fibers preferably comprise a silica core having a diameter selected from the range of 50 μm to 200 μm. Some preferred embodiments have a diameter of about 100 μm, e.g., 105 μm. In some embodiments, the one or more optical fibers preferably have a numerical aperture selected from the range of 0.2 to 0.3. Most preferred embodiments have a numerical aperture of about 0.2, e.g., 0.22. Such characteristics preferably combine, for at least one collection fiber, the largest possible core diameter and the largest possible numerical aperture, while taking into account the form factor of the needle outer wall and / or cannula and the physical constraints of any measurement device.

[0024] According to a second aspect of the present disclosure, there is provided a system for obtaining electromyographic data and optical spectroscopy data from a muscle, the system comprising: a muscle probe arranged for insertion into the muscle, the muscle probe comprising a needle and one or more optical fibers; a light source arranged to provide incident light for transmission to a target region of the muscle by the one or more optical fibers; an optical spectrometer arranged to receive scattered light from the one or more optical fibers; and an electromyography device arranged to receive electrical signals from the needle; wherein the needle comprises an outer wall including a needle interior and a core electrode positioned within the needle interior, and the one or more optical fibers are located within the needle interior.

[0025] In some embodiments, the muscle probe may be a muscle probe according to the first aspect and may therefore include any of the features described herein as being suitable for a muscle probe of the first aspect.

[0026] The one or more optical fibers may preferably comprise: at least one delivery fiber positioned to direct incident light from the light source toward a target region of the muscle; and at least one collection fiber positioned to receive scattered light from the target region. In some embodiments, the at least one delivery fiber and the at least one collection fiber each comprise one of: an in-line bandpass filter; an in-line shortpass filter; an in-line longpass filter; or a notch filter. In some specific embodiments, the at least one delivery fiber preferably comprises an in-line bandpass filter or an in-line shortpass filter as described herein, and the at least one collection fiber comprises an in-line longpass filter or a notch filter as described herein.

[0027] In some embodiments, the electromyography device is preferably configured to: use the electrical signals to determine electromyography data (e.g., recordings of motor unit action potentials and other related waveforms); and the optical spectrometer is configured to: use the received scattered light to determine an optical spectrum characteristic of the target region. In some embodiments, the system preferably further comprises a memory arranged to store the optical spectrum and the electromyography data.

[0028] In some embodiments, the system preferably further comprises a processor configured to perform one or more of the following: process the electromyogram data and use the electromyogram data to determine the target region; and / or process the optical spectrum and, optionally, the electromyogram data and use the optical spectrum and, optionally, the electromyogram data to determine a data fingerprint of the target region. It will be understood that the data fingerprint may include raw optical spectroscopy (and, optionally, EMG) data or any suitable processed data, such as data representing an average value or distribution obtained from a series of data samples. Thus, the method provides an EMG-guided biochemical assessment of the target region, preferably reducing the time it takes to accomplish such an assessment, while combining the broad locating strength of EMG with the specificity, objectivity, comparative quantification, and non-invasiveness of spectroscopy. The data fingerprint may include data determined using optical spectroscopy or a combination of electromyography and optical spectroscopy. Such a fingerprint may include raw, preprocessed, or processed optical spectra collected from the target region and / or representative data of said spectra. Additionally, the data fingerprint may include or be determined using metadata, which may include any suitable metadata, such as one or more of the patient's age, sex, symptoms, and laboratory test results. Other suitable metadata will be appreciated by those skilled in the art.

[0029] In some embodiments, the processor is preferably further configured to: compare the data fingerprint of the target region with one or more stored data fingerprints; and use said comparison to determine one or more of: an index of disease state; a prediction of disease state; a predicted disease prognosis; or a predicted response to treatment. Multivariate statistics may be used for such comparison in exemplary embodiments, as will be understood by those skilled in the art. In some embodiments, the processor preferably comprises a machine learning module trained using the plurality of stored fingerprints, the machine learning module being configured to process the data fingerprints and output one or more of: an index of disease state; a prediction of disease state; a predicted disease prognosis; or a predicted response to treatment.

[0030] In some embodiments, the light source is preferably a laser. In some embodiments, the incident light preferably comprises a wavelength selected from the near-infrared spectrum. In some specific embodiments, the wavelength may be selected from the range of 785 nm to 830 nm. These wavelengths preferably produce less fluorescence than visible wavelengths, thus reducing any fluorescence spectrum that may be superimposed on the intended optical spectrum, which may complicate the measurement of certain optical spectroscopic data (e.g., Raman data). Such wavelengths are preferably more biocompatible because the use of shorter UV wavelengths can be mutagenic and cause tissue damage. The use of longer infrared wavelengths may result in less spectroscopic (e.g., Raman) signal and generally require more expensive and less efficient detectors than silicon-based detectors, such as those that may be used in the intended embodiments.

[0031] According to a third aspect of the present disclosure, a computer-implemented method is provided that receives, by a computer, electrical signals indicative of electrical activity in a muscle from an electromyography needle; determines, by the computer, a target muscle site based on the electrical signals; outputs, by the computer, the target muscle site for guiding an optical spectroscopy probe to the target muscle site; and receives, by the computer, optical spectroscopy data characterizing the target muscle site from the optical spectroscopy probe.

[0032] In some muscle pathologies, electrical signals indicative of healthy muscle electrical activity may be observed at the target muscle site, and it will be understood that in some embodiments the electrical signals are indicative of either healthy or diseased muscle electrical activity.

[0033] According to a fourth aspect of the present disclosure, there is provided a computer-implemented method of receiving, by a computer, optical spectroscopy data characterizing a muscle from an optical spectroscopy probe; determining, by the computer, a target muscle site based on the optical spectroscopy data; outputting, by the computer, the target muscle site for guiding an electromyography needle to the target muscle site; and receiving, by the computer, an electrical signal from the electromyography needle indicative of electrical activity in the muscle at the target muscle site.

[0034] In some embodiments, preferably, the method or the third or fourth aspect further comprises: determining, based on the optical spectroscopic data and optionally further based on the electrical signal, one or more of: an indicator of a disease state; a prediction of a disease state; a predicted disease prognosis; or a predicted and / or measured response to treatment. It will be understood that the electrical signal may be processed, for example, by an electromyography device, to provide electromyographic data (e.g., a recording of motor unit action potentials and other related waveforms) prior to use in the determining step. Available processes for performing the determining step may include any suitable process, such as by comparing the optical spectroscopic data and optionally the electrical signal with stored optical spectroscopic data and optionally stored electrical signal data. Other processes may include utilizing a machine learning module trained on stored optical spectroscopic data and optionally stored electrical signal data to perform the determining step. Such determining, in some embodiments, may include generating a digital fingerprint by a processor using the optical spectroscopic data and optionally the electrical signal (or data derived therefrom, such as optical spectra or electromyographic data). Such fingerprints may represent digital biomarkers that characterize one or more neuromuscular diseases; prognosis of muscle and / or associated diseases; or indicators of muscle (and / or associated diseases) response to treatment.

[0035] In some embodiments, the determination is preferably performed by processing the optical spectroscopic data and optionally the electrical signals using a machine learning module trained using the stored optical spectroscopic data and optionally the stored electrical signals.

[0036] The method may in some embodiments be carried out using a muscle probe according to the first aspect or a system according to the second aspect.

[0037] According to a fifth aspect of the present disclosure, there is provided a method for determining a muscle pathology (e.g., acute myopathy, chronic myopathy, inflammatory myopathy; dystrophic myopathy; mitochondrial myopathy; neurogenic muscle pathology, or any suitable pathology as discussed herein) at a target muscle site in a subject, the method comprising: a computer system including at least one processor and a memory storing at least one program for execution by the at least one processor, the at least one program including: instructions for acquiring a dataset in electronic form, the dataset including test optical spectroscopy data samples acquired from a target muscle site in the subject; and instructions for applying the dataset to a machine learning classifier trained using the stored optical spectroscopy data, thereby determining the muscle pathology of the subject.

[0038] In some embodiments, the optical spectroscopic data sample is preferably determined using one or more of: Raman scattering; fluorescence scattering; Brillouin scattering at the target muscle site of the subject.

[0039] In some embodiments, the program preferably further includes instructions for isolating, from the test optical spectroscopic data sample, a spectral region comprising spectral data characterizing at least one protein secondary structure. In some embodiments, the spectral region is preferably obtained from the amide I band. While any suitable spectral region will be understood, the region is preferably associated with muscle proteins, and more preferably associated with protein secondary structure. Limiting the optical spectroscopic data to that associated with muscle, and more specifically, that associated with protein secondary structure, preferably provides more relevant spectral data to the classifier, thereby preferably resulting in an improved false discovery rate.

[0040] The spectral data preferably characterizes at least the amount or proportion (which may be relative or quantitative) of at least one protein secondary structure at the target muscle site, preferably alpha helices at the target muscle site and beta sheets at the target muscle site. In some embodiments, the spectral data can characterize the amount or proportion of disordered secondary structure at the target muscle site.

[0041] In some embodiments, the target muscle site is preferably determined using electrical signals from an electromyography needle, the electrical signals indicative of electrical activity in the muscle. The electrical signals and / or electrical activity are preferably indicative of a muscle pathology. In some muscle pathologies, the electrical signals may be indicative of electrical activity observed in healthy muscle. Accordingly, embodiments are understood in which the electrical signals and / or electrical activity are preferably indicative of healthy muscle. In some embodiments, the predetermined threshold is preferably related to the electrical activity of healthy muscle. In some embodiments, the dataset further includes test electrical signal data from an electromyography needle, the test electrical signal data indicative of electrical activity at the subject's target muscle site; and the machine learning classifier is further trained using the stored electrical signal data. The electrical signal data preferably includes data indicative of motor unit action potentials, such as data indicative of motor unit action potential morphology or configuration, motor unit action potential recruitment, and / or spontaneous muscle cellular activity. Embodiments are understood in which the electrical signal data preferably includes data indicative of compound muscle action potential (CMAP) amplitude at the target muscle site.

[0042] In preferred embodiments, the machine learning classifier includes or is generated using any suitable embedding model or dimensionality reduction technique. In some embodiments, the machine learning classifier includes or is generated using matrix factorization. Matrix factorization has been identified as a powerful machine learning technique for identifying muscle pathologies, preferably improving interpretation of results at the individual sample level rather than requiring population analysis, providing a desirable addition to the diagnostic pathway. As used herein, the term "sample" is understood to mean sample data obtained from a subject and may include a group of data obtained from a subject or a single data entity, such as a single spectrum, from said subject. In particular, matrix factorization provides output that is easier to interpret than traditionally used methods. Suitable matrix factorization methods include, for example, nonnegative matrix factorization or bounded simplex-structured matrix factorization (BSSMF). In some specific embodiments, relevant protein secondary structure information is determined through matrix factorization of a dataset including optical spectroscopic data obtained from multiple training subjects (e.g., including healthy and diseased subjects). In particular, matrix factorization is used to identify dominant spectral patterns within a dataset, with each training data sample preferably being assigned a weight corresponding to each of said spectral patterns, said corresponding weight relating to the relative importance of said spectral pattern to the training data sample. Preferably, each spectral pattern is associated with a corresponding protein secondary structure. Thus, in some embodiments, said corresponding weight is assigned to a relevant spectral data region, e.g., alpha helix (1650-1658 cm), -1 ), β-sheet (1664-1673cm -1 ) and irregular (1630-1640, 1674-1689 and 1700-1710 cm -1 ) The weights can be applied to a model, such as a linear discriminant model, for classification of a test data set.

[0043] In some embodiments, the machine learning classifier includes or is generated using hierarchical modeling. Using hierarchical modeling, for example, to provide an expert system approach, has been shown to improve false discovery rates at the single-sample level while maintaining the power to provide continuous classification, which may provide a more useful tool in clinical settings. This may include using a hierarchical classification model built on multiple continuous classification problems relevant to clinical settings, such as determining the presence of a muscle pathology, such as a myopathy, a myopathy subclass (e.g., acute or chronic myopathy), or a neurogenic muscle pathology. For example, a model may first distinguish "disease" samples from "healthy" samples, then classify the "disease" samples according to disease type, and then classify according to a more specific disease subtype. For example, a model may be configured to first classify "disease" samples from "healthy" samples, then "myopathic" samples from "neuropathic" samples, and then "acute myopathy" samples from "chronic myopathy" samples. It will be understood that any suitable classification can be performed according to the training data used to train the classifier. As an example, consecutive two-class problems can be presented, such as "healthy" versus "disease" (where "disease" encompasses sample data from subjects exhibiting either acute myopathy, chronic myopathy, and neurogenic conditions); "myopathy" (encompassing sample data from subjects exhibiting either acute or chronic myopathy) versus "neurogenic"; and "acute myopathy" versus "chronic myopathy." These hierarchical classifications can be trained using any suitable means, such as using partial least squares discriminant analysis (PLS-DA) models constructed using training data sets for each relevant two-class problem. In some embodiments, training can include feature selection, for example, using variable importance in a projection approach, followed by the selection of a select few variables for inclusion in the model. Thus, a hierarchical modeling approach provides improved classification flexibility while maintaining improved performance in assessing individual samples.Furthermore, the hierarchical modeling approach more closely approximates the decisions required in clinical practice, thereby improving its applicability in clinical settings.

[0044] In some embodiments, the machine learning classifier includes or is generated using multi-block modeling or data fusion modeling, trained by any suitable means as described herein, such as using a PLS-DA model. In preferred such embodiments, the machine learning classifier is further trained using electrical signal data indicative of electrical activity within a target muscle region (e.g., data indicative of motor unit action potentials and / or spontaneous muscle fiber activity), where the test dataset includes the electrical signal data obtained from the target muscle region. In particular, combining optical spectroscopy data with the electrical signal data improves the predictive power of the multi-block modeling or data fusion modeling approach, thereby maintaining an optimal false discovery rate without increasing the amount of training data. Embodiments in which the machine learning classifier is generated using any combination of matrix factorization; principal component analysis; hierarchical modeling; multi-block modeling; and data fusion modeling are understood. In preferred embodiments, the classification may include any suitable disease or disease subtype, such as a muscle myopathy or a neurogenic condition, and the muscle myopathy may be any one selected from: acute muscle myopathy; chronic muscle myopathy; inflammatory myopathy; immune myopathy; dystrophic myopathy; mitochondrial myopathy; hereditary myopathy; congenital myopathy; metabolic myopathy; toxic myopathy; endocrine myopathy; infectious myopathy; severe myopathy; muscular dystrophy; neurogenic condition.

[0045] According to a sixth aspect of the present disclosure, there is provided a digital biomarker determined using either optical spectroscopic data obtained from muscle or a combination of optical spectroscopic data and electromyographic data obtained from muscle, the digital biomarker characterizing one or more of the following: one or more neuromuscular disorders; prognosis of muscle and / or disorders associated therewith; or an indication of response of muscle (and / or disorders associated therewith) to treatment. The digital biomarker is preferably determined using a muscle probe according to the first aspect, a system according to the second aspect, or a method according to the third aspect.

[0046] In preferred embodiments, the optical spectroscopic data may be any data obtained by analyzing scattered light received from the target muscle tissue region. The scattered light is preferably inelastically scattered light. The optical spectroscopic method is preferably one or more of: Raman spectroscopy; fluorescence spectroscopy; or Brillouin spectroscopy.

[0047] It has been observed that neuromuscular disorders and / or myopathies can be identified according to changes in protein secondary structure in affected muscle. Thus, in preferred embodiments, the optical spectroscopic data preferably characterizes the amount or proportion, e.g., relative or quantitative amount or proportion, of any suitable secondary structure or structures in the target muscle region, which may include, for example, alpha helices, beta sheets, and / or disordered secondary structures. In preferred embodiments, the optical spectroscopic data preferably characterizes the proportion of alpha helices and beta sheets in the muscle. Other embodiments will be understood in which the optical spectroscopic data preferably characterizes the proportion of disordered secondary structure relative to alpha helices and / or beta sheets. In many cases, muscle pathologies can be identified according to the proportional increase or decrease of a particular protein secondary structure relative to other protein secondary structures in the target muscle region. In the case of neuromuscular disorders and / or myopathies, a decrease in alpha helices relative to beta sheets is often observed in muscle. Thus, preferably, the ratio of alpha helices to beta sheets is below a predetermined threshold. In some embodiments, the predetermined threshold is the ratio of alpha helices to beta sheets associated with healthy muscle. In some embodiments, the predetermined threshold is a ratio of alpha helix to beta sheet less than 1. A proportional decrease in alpha helix to beta sheet in a target muscle region when compared to healthy muscle has been identified as indicative of a neuromuscular disorder and / or myopathy. Preferably, the predetermined threshold is determined by a machine learning module trained on stored optical spectroscopy data obtained from muscle, or a combination of stored optical spectroscopy data and stored electromyography data obtained from muscle.

[0048] Digital biomarkers may be determined using or include any suitable metadata, which may include, for example, one or more of a patient's age, sex, symptoms, and clinical test results. Other suitable metadata will be appreciated by those skilled in the art.

[0049] According to a seventh aspect of the present invention, there is provided a non-transitory computer readable storage medium having stored thereon a digital biomarker according to the sixth aspect.

[0050] It will be understood that any feature described herein as suitable for incorporation into one or more aspects or embodiments of the present disclosure is intended to be generalizable across any and all aspects and embodiments of the present disclosure.

[0051] Embodiments of the present disclosure will now be described, by way of example only, and with reference to the accompanying drawings, in which: [Brief explanation of the drawings]

[0052] [Figure 1] Figure 1A is a perspective view of an example of a muscle probe according to a first aspect of the present disclosure, Figure 1B is an enlarged cutaway view of the distal end of the needle of the embodiment shown in Figure 1A, and Figure 1C is a perspective cutaway view of the end of the needle of the embodiment shown in Figures 1A and 1B. [Figure 2] Figures 2A to 2D show experimental results demonstrating the electrophysiological functionality of the probe according to the first embodiment in a 90-day-old SOD1G93A mouse model. In particular, Figure 2A shows compound muscle action potential (CMAP) waveforms from both SOD1G93A and non-transgenic (NTg) mice using an optical EMG probe. Figure 2B shows spontaneous EMG activity (positive sharp waves; examples are circled) from the optical EMG probe. Figure 2C shows a comparison of CMAP amplitudes between non-transgenic (NTg) and SOD1G93A mice using an optical EMG probe. Figure 2D shows a comparison of CMAP amplitudes between NTg and SOD1G93A mice using a standard EMG needle. [Figure 3]Figures 3A and 3B show example spectroscopy results (Raman spectroscopy in the illustrated example) obtained from an experiment utilizing a muscle probe according to the first embodiment to obtain Raman spectroscopic data. In particular, Figure 3A shows the average Raman spectra (±standard deviation) of 90-day-old NTg mice (top) and SOD1G93A mice (bottom). Figure 2B shows the difference in the average spectra (SOD1G93A minus NTg). [Figure 4] Figures 4A-4C show the results of multivariate analysis of Raman spectroscopy data obtained using a muscle probe according to the first embodiment. In particular, Figure 4A shows a plot of the linear discriminant loadings. Figure 4B shows that the mean LDF scores for each mouse were significantly different at the group level (nested t-test). Figure 4C shows the receiver operating characteristic curve and classification performance data for the PCA-LDA model. [Figure 5] Figures 5A and 5B show that in vivo intramuscular Raman spectroscopy does not alter CMAP amplitude. In particular, Figure 5A shows an example of the CMAP and Raman spectra recorded from an SOD1G93A mouse using an optical EMG probe. No significant difference was observed in the CMAP amplitude before and after Raman. Figure 5B shows an example of the CMAP and Raman spectra recorded from an NTg mouse. No significant difference was observed in the CMAP amplitude before and after Raman. [Figure 6] Figures 6A and 6B show a comparison of CMAP amplitudes between the muscle probe (optical EMG probe) according to the first embodiment and a standard EMG needle. In particular, Figure 6A shows that the CMAP amplitude was slightly smaller with the standard EMG needle, but this did not reach statistical significance (P = 0.05). Figure 6B shows that the CMAP amplitude was not significantly different in SOD1G93A mice. [Figure 7]This figure shows that preclinical peak fitting of healthy, myopathic, and neuropathic muscles demonstrates a decrease in α-helical conformation in myopathy. Specifically, Figures 7A-C show peak fitting using standard peaks within the amide 1 region of healthy, mdx, and SOD1G93A mice. Figure 7D shows each resolved peak as a percentage of the total area. In the myopathy model (mdx), a decrease in α-helix and a concomitant increase in β-sheet are observed. Figure 7E shows the conformational ratios of different protein secondary structures. A decrease in α-helix / increase in β-sheet is evident in the myopathy mdx model, along with an increase in the β-sheet / disorder ratio. [Figure 8] This figure shows that spectral patterns obtained by nonnegative matrix factorization indicate differences in three-dimensional structures. The unique spectral patterns (left) output from nonnegative matrix factorization demonstrate differences in protein structure (middle) and differences between healthy mice, mdx mice, and SOD1G93A mice (right). The distinct conformational fingerprints and their predominance in different mice are consistent with those observed by peak fitting (*P<0.05, **P<0.01, ***P<0.001, ****P<0.0001). [Figure 9] 10 is a table showing three-group, cross-validated classification performance on in vivo preclinical data using nonnegative matrix factorization mode and linear discriminant models. [Figure 10] Figure 10 shows peak fitting from Raman muscle spectra of human samples obtained from patients with and without myopathy. In particular, Figure 10A and Figure 10B show peak fitting within the amide I region. Figure 10C shows each resolved peak as a percentage of the total area. The myopathy group shows a decrease in α-helix and an increase in other conformations and aromatic amino acids. Figure 10D shows the ratio of different protein conformations. A decrease in α-helix and an increase in β-sheet / disordered structures are evident in the myopathy group. [Figure 11]This figure shows the spectral patterns in human muscle obtained by nonnegative matrix factorization. In particular, the unique spectral patterns output from nonnegative matrix factorization (left) indicate differences in protein structure (center). Significant differences were observed in mode 1 (***P<0.001). [Figure 12] 10 is a table showing the two-group classification performance of human ex vivo samples using non-negative matrix factorization mode and linear discriminant model. [Figure 13] FIG. 10 is a schematic diagram of a hierarchical model suitable for use in a method according to a fifth embodiment. [Figure 14] Confusion matrix showing the classification performance of individual spectra derived from the hierarchical model approach shown schematically in Figure 13 using data obtained from mice (healthy / non-transgenic = healthy; acute myopathy = 30-day-old mdx mice; chronic myopathy = 90-day-old mdx mice; neuropathic = 90-day-old SOD1G93A mice). [Figure 15] 14 is a table showing classification performance statistics for validation test data derived from the hierarchical model approach outlined in FIG. 13. [Figure 16] Figure 16A shows a confusion matrix illustrating the classification performance of a multi-block model suitable for use in a method according to a fifth embodiment, where the multi-block model utilizes Raman spectroscopy data combined with CMAP data, and Figure 16B is a table showing classification performance statistics for the multi-block model approach of Figure 16A. [Figure 17] Figure 17A shows a confusion matrix illustrating the classification performance of a model suitable for use in a method according to the fifth embodiment, which utilizes only Raman spectroscopy data, and Figure 17B is a table showing classification performance statistics for the model approach of Figure 17A. [Figure 18] FIG. 2 is a schematic diagram illustrating an exemplary embodiment of a system according to a second aspect of the present disclosure. [Figure 19] 10 is a flowchart listing the steps of an exemplary embodiment of a method according to a third aspect of the present disclosure. DETAILED DESCRIPTION OF THE INVENTION

[0053] 1A, a perspective view of an exemplary muscle probe 100 according to a first aspect of the present disclosure is shown. The probe 100 comprises a probe body 102 having a needle 104 attached by a standard luer connector 106. The needle 104 comprises an elongated steel outer wall 110 extending from the luer connector 106 and terminating in an open end 108 distal from the probe body 102.

[0054] 1B and 1C each show a close-up cutaway view of a portion of needle 104 proximate its distal end 108. As seen in FIG. 1B, the needle includes a tubular outer wall 110 defining a needle interior 111 therein. In the particular example shown, outer wall 110 is formed from a standard 21G hypodermic needle having an outer diameter of 0.819 mm, an inner diameter of 0.514 mm, and a distal bevel angle of 12 degrees relative to a plane parallel to the longitudinal axis of outer wall 110, providing a needle tip. In use, the needle tip is used to insert outer wall 110 into muscle tissue (not shown).

[0055] The needle interior 111 contains a silver electrode 112 extending therealong, which in the illustrated embodiment forms a tube. The electrode 112 in the illustrated embodiment is coated with a polymer coating 115 which serves to electrically insulate the electrode 112 from the needle outer wall 110. A small distal region of the electrode 112 is left uncoated in order to obtain the required electromyographic signal from the target muscle region in use.

[0056] The needle outer wall 110 has a conductive wire 113 extending therefrom. The conductive wire 113 and each of the electrodes 112 extend along the probe body 102 from the end of the probe body 102 distal from the needle 104 to an electromyograph (not shown). In use, each of the electrodes 112 and conductive wire 113 (connected to the conductive needle outer wall 110) is positioned to transmit electrical signals from the muscle tissue to the electromyograph. The electromyograph processes the electrical signals to record motor unit action potentials and other related waveforms as will be understood by those skilled in the art.

[0057] The tubular electrode 112 contains a plurality of optical fibers 114, 116, which in the illustrated embodiment include an emitting optical fiber 114 and three receiving optical fibers 116. Each optical fiber 114, 116 includes a cladding layer 120 surrounding a light-carrying core 122. In the illustrated embodiment, each of the optical fibers 114, 116 comprises a low-OH fiber having a silica core with a diameter of 105 μm and a numerical aperture (NA) of 0.22. For ease of explanation, the illustrated example 100 includes a single emitting optical fiber 114 and three receiving optical fibers 116. Embodiments with any number of emitting and receiving optical fibers are understood. To maximize the light collection area, embodiments preferably include more receiving optical fibers than emitting optical fibers. Embodiments in which a single optical fiber is used for both emitting and receiving light are also understood.

[0058] The light-emitting optical fiber 114 and the light-receiving optical fiber 116 each extend through the interior of the tubular electrode 112, along the length of the probe body 102, and out the end of the probe body 102 distal to the needle 104. The light-emitting optical fiber 114 extends to a light source, which in the illustrated embodiment is a semiconductor laser (not shown). Proximate the illustrated end of the light-emitting optical fiber 114 (approximately 15 cm from the end in the particular example shown), the light-emitting optical fiber 114 includes a bandpass filter (not shown). The light-receiving optical fiber 116 extends to a Raman spectrometer (not shown). Between the probe body 102 and the Raman spectrometer, proximate the illustrated end of the light-receiving optical fiber 116 (approximately 15 cm from the end in the particular example shown), the light-receiving optical fiber 116 includes a longpass filter (not shown). In use, the semiconductor laser emits a light beam, which in the illustrated embodiment includes a wavelength of 830 nm. The light beam propagates toward the muscle along the core 122 of the light-emitting optical fiber 114. Raman scattered light reflected from the muscle is received by the end of the light-receiving optical fiber 116, which is proximate the distal end 108 of the needle 104. The Raman scattered light propagates along the core 122 of the light-receiving optical fiber 116 and is transmitted to a Raman spectrometer. The Raman spectrometer processes the received light to provide a Raman spectrum that characterizes the molecular composition of the muscle region targeted by the light-emitting optical fiber 114. The Raman spectrum, and optionally recordings of motor unit action potentials and other associated waveforms, can be used as a digital biological fingerprint of the assessed muscle region, which can be used to determine the pathology, prognosis, disease progression, and expected response to treatment, among other appropriate clinical outcome measures.

[0059] Figures 2A-6B, described below, show the results of an experiment using a muscle probe according to the first embodiment and as shown in Figures 1A and 1B. Such a muscle probe is suitable for use in a system according to the second embodiment to carry out a method according to the third embodiment and to provide a digital biomarker according to the fourth embodiment. The following description outlines: 1. the methodology used in collecting the data depicted in Figures 2A-6B; 2. the results obtained from the experiment; and 3. the conclusions drawn from the results.

[0060] methodology Raman spectroscopy For Raman data collection, a probe (hereafter referred to as the "optical EMG probe") was prepared substantially as described above with respect to Figures 1A and 1B and connected to an 830 nm semiconductor laser (Innovative Photonics Solutions) as described above to provide a good signal-to-noise ratio. An in-line filter (Semrock, Inc.) as described with respect to Figures 1A and 1B was used to reduce the effects of the Raman signal and associated fluorescence on the emission optical fiber.

[0061] The laser power was 60 mW at the probe tip. The optical EMG probe was optically adapted to a Raman spectrometer (Raman Explorer Spectrograph, Headwall Photonics, Inc. and iDus 420BR-DD CCD camera, Andor Technology, Ltd.). Raman signals were recorded over a 40-second exposure consisting of 10 averaged 4-second epochs. A spectrum of polytetrafluoroethylene (PTFE) was acquired for wavenumber calibration of the spectrometer.

[0062] electromyography For electrophysiological data collection, the optical EMG probe was connected to a Dantec Keypoint Focus EMG system with standard filter settings (20 Hz–10 kHz). For comparison, compound muscle action potential (CMAP) recordings were also performed using a probe equipped with a standard commercially available concentric EMG needle (Ambu Neuroline, 30G) (hereafter referred to as the "concentric EMG needle").

[0063] In vivo testing Experiments on mice were conducted under a UK Home Office project license (number 70 / 8587) in accordance with the Animals (Scientific Procedures) Act 1986. The project was approved by the University of Sheffield Animal Welfare and Ethical Review Body (AWERB). Mice were housed in standard facilities (12-hour light / dark cycle, room temperature 21°C) and maintained in accordance with the Home Office Code of Practice for Housing and Care of Animals Used in Scientific Procedures. The ARRIVE guidelines were followed in the conduct of this work.

[0064] Transgenic C57BL / 6J-Tg(SOD1G93A)1Gur / J mice were used (originally obtained from Jackson Laboratories). Hemizygous transgenic males were backcrossed to C57BL / 6 females (Harlan UK, C57BL / 6 J OlaHsd substrain) for over 20 generations. Hemizygous females were used for experiments, and nontransgenic (NTg) females served as controls. Transgenic SOD1G93A mice were identified by PCR amplification of genomic DNA extracted from ear clips. These mice are very well characterized, allowing selection of an age at which motor function declines and significant histopathology of hindlimb muscles appear. Raman spectra and / or EMG recordings were collected at 90 days of age. A total of 17 mice (SOD1G93A, n = 10 and NTg, n = 7) were used in this study.

[0065] For Raman spectroscopy and / or EMG recording, mice were anesthetized with 2% isoflurane and placed on a heat pad to maintain body temperature. The hind limbs were shaved, and optical EMG probes were inserted into both gastrocnemius muscles. Thus, Raman data were obtained from two sites (right and left legs) for each mouse. For electrophysiological recordings using the optical EMG probe, electrophysiological data were collected from the probe insertion site in the left medial gastrocnemius muscle.

[0066] Compound muscle action potentials (CMAPs) were elicited by applying a 0.1 ms stimulus to the sciatic notch. Stimulus intensity was adjusted to obtain a supramaximal response, and the amplitude from baseline to the negative peak was recorded. Recordings using a concentric EMG needle probe were performed under anesthesia from the medial gastrocnemius muscle in the same sitting position and using the same methodology as for the probe recordings. Mice were humanely killed after recording.

[0067] Data analysis Raman spectral analysis was performed using custom code in MATLAB (MATLAB R2019b The MathWorks). Raw spectra were first analyzed from 900 cm to 1800 cm. -1The spectra were interpolated to integer wavenumber intervals of 900 cm, then normalized (standard normal variate normalization) and mean-centered. -1 ~1800cm -1 A spectral window of 100 nm was performed to capture information within the biological fingerprint region. Spectra below this window were obscured by silica-related Raman signals generated within the optical fiber; spectra above this window consisted solely of noise. To display the average spectra, background subtraction using the adaptive iteratively reweighted penalized least squares (airPLS) algorithm was performed. However, all SOD1G93A vs. NTg analyses were performed without background subtraction.

[0068] Multivariate analysis involved principal component-linked linear discriminant analysis (PCALDA). Principal components (PCs) (PC1 and PC2) showing significant intergroup differences were used as inputs to the linear discriminant model. The classification performance of the model was validated using leave-one-mouse cross-validation (CV). Data from a given mouse were excluded and treated as the test set. The model was then constructed using the remaining data. The test set was then projected onto the model, performance data were collected, and data from each mouse was excluded once. This process was repeated until the group was predicted by the model. Accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve were reported. Between-group analysis of LDF scores was performed using a nested t-test (scores nested within each mouse) in GraphPad Prism (version 9). Differences in CMAP amplitude between SOD1G93A and NTg mice were analyzed using an unpaired t-test; analysis of CMAP amplitude recorded before and after Raman spectra was performed using a paired t-test.

[0069] result To evaluate both the electrophysiological and Raman spectroscopic capabilities of the optical EMG probe, we performed tests on a 90-day-old SOD1G93A model. Using the optical EMG probe, we were able to record CMAP after sciatic nerve stimulation (Figure 2A). In addition, spontaneous EMG activity in the form of positive sharp waves (PSWs) was also recorded (Figure 2B). Significant differences in CMAP amplitudes recorded from SOD1G93A and NTg mice were observed using both the optical EMG probe and the concentric EMG needle (Figure 2C and Figure 2D). Thus, we were able to record clinically relevant, high-quality electrophysiological data from the optical EMG probe, including PSWs with amplitudes of only 200 µV.

[0070] Immediately after collecting the electrophysiological data, Raman spectra were acquired. These contained muscle-related peaks, and tentative peak assignments were made based on existing literature. The average spectrum showed a particularly prominent peak at 935 cm. -1 (CC stretching, protein α-helix), 1000 cm -1 (phenylalanine), 1448 cm -1 (protein / phospholipid), 1654 cm -1 (amide I, α-helix) (Figure 3A). The spectral difference (average of SOD1G93A minus average of NTg) was 935, 1045, 1448, and 1654 cm -1 We demonstrated increased concentrations of peaks associated with protein structure in NTg mice (Figure 3B). Linear discriminant analysis revealed peaks similar to those of spectral differences (Figure 4A), indicating that these wavenumber / biochemical components are important for distinguishing healthy muscle from SOD1G93A muscle pathology. The mean LDF scores for each mouse were significantly different at the group level (nested t-test, Figure 4B). High classification performance was observed using PCA-LDA (Figure 4C).

[0071] After recording the Raman spectrum, the probe needle was fixed in the same position and the CMAP measurement was repeated (Figure 5A shows the data from the SOD1G93A mouse, and Figure 5B shows the data from the non-Tg mouse). No significant changes in the CMAP amplitude were observed after Raman spectrum acquisition.

[0072] conclusion The above demonstrates the functionality of the combined EMG / Raman spectroscopy probe according to the first embodiment. The results show that the probe is capable of recording high-quality electrophysiological and Raman data in vivo. The data also provide evidence for the utility of optical EMG data as a translational biomarker of muscle health.

[0073] Because Raman spectroscopy has the potential to provide specific molecular information, the combination of EMG and Raman spectroscopy data is an attractive biomarker of muscle health. On a practical level, the combined use of EMG can confirm that the Raman probe is within the muscle of interest. This is useful in pathological conditions that cause muscle weakness and therefore make palpation difficult. Real-time analysis of EMG activity can also be used to target the Raman probe to electrically abnormal (or, if desired, normal) muscle regions, increasing the likelihood of obtaining molecular information from the region of interest.

[0074] The optical EMG probe described herein demonstrated excellent electrophysiological functionality. The method for detecting muscle depolarization is similar to that of a standard concentric EMG needle. Standard EMG needles utilize a potential difference between the outer wall of the needle, typically made of steel (which acts as a "reference"), and the inner wire (called the core) (e.g., silver, platinum, or other suitable material as described herein) (which acts as the active electrode). In the example probe described, the electrode tip is polished to a 15-degree angle, ensuring close proximity between the outer wall and the core, resulting in high-quality signals due to common-mode rejection. The difference in electrophysiological functionality between the optical EMG probe (Figure 6A) and a commercially available concentric EMG needle (Figure 6B) did not result in a statistically significant difference in CMAP amplitude. Any suitable size of the needle's outer wall can be used, e.g., larger or smaller than the sizes described, if necessary. The size of the needle outer wall in preferred embodiments is larger than or equal to the size described for the embodiment of Figures 1A and 1B to maximize the ability to collect inelastically scattered (e.g., Raman scattered) light. A smaller diameter needle outer wall may be used in some embodiments, and in such embodiments, increasing the laser power and / or acquisition time can offset the reduced collection area.

[0075] Prominent protein peaks, likely related to muscle proteins such as myosin and actin, were observed. Therefore, these Raman spectra can distinguish between neurogenic and myopathic pathologies and different disease stages. The molecular information obtained from Raman spectra obtained using the probe according to the present invention cannot be obtained with currently available in vivo techniques. Therefore, the probe can provide disease-specific data that is currently only obtainable through muscle biopsy. A key advantage of the probe is the potential for examining several regions within a muscle and multiple muscles, as is commonly done with routine EMG. This increases the likelihood of obtaining disease-specific information that may be missed by biopsy of a small muscle sample.

[0076] No significant differences in CMAP amplitude were observed after Raman, indicating that laser thermal energy exposure did not adversely affect the depolarization capacity of muscle fibers, highlighting the potential of Raman spectroscopy as a non-destructive technique for tissue analysis.

[0077] Thus, the present invention provides an EMG / Raman spectroscopic assessment technique for muscle tissue that combines electrophysiology and vibrational spectroscopy. The data presented herein, using the SOD1G93A model of ALS, demonstrates the utility of the present invention in detecting neuromuscular disease by showing that optical EMG can provide a sensitive and quantitative measure of disease.

[0078] 7A-12 are described below, which illustrate exemplary training and classification performance results of a machine learning module trained using matrix factorization and applied according to the fifth aspect, where the data is suitable for being acquired by a muscle probe according to the first aspect or a system according to the second aspect, and the method is suitable for providing a digital biomarker according to the sixth aspect. The following description outlines: 1. the methodology used in collecting the data depicted in FIGS. 7A-12, 2. the results obtained from the experiments, and 3. the conclusions drawn from said results.

[0079] method Fiber-optic Raman spectroscopy The fiber-optic Raman system utilized a probe according to the first embodiment. The system followed the second embodiment. Specifically, a 0.5 mm probe housed in a 21-gauge hypodermic needle was used. An 830 nm semiconductor laser (Innovative Photonics Solutions) was used, with two low-OH fibers providing the optical path (delivery and collection). In-line bandpass filters were used to remove inelastically scattered light and fiber-associated fluorescence (Semrock Inc.). The collection fiber was optically coupled to a spectrometer (Raman Explorer Spectrograph, Headwall Photonics, Inc. and an iDus 420BR-DD CCD camera, Andor Technology, Ltd.). The laser power at the end of the probe was 60 mW, and the spectral collection time was 40 seconds for all studies.

[0080] Preclinical records Breeding was performed in a specific pathogen-free environment, and experimental work was carried out in a standard preclinical facility (12-hour light / dark cycle, room temperature 21°C). As a muscle disease model, the mdx model of Duchenne muscular dystrophy was used at both day 30 (n = 16, representing the disease onset phase) and day 90 (n = 16, representing the disease stabilization phase), along with matched wild-type healthy control mice (C57BL / 10ScSnOlaHsd, n = 16 at both day 30 and day 90). SOD1 in motor neuron disease G93A The model was used as a model of "neurogenic" disease (diseases associated with nerve or motor neuron pathology) after 90 days (n=16, established disease phase) along with non-transgenic healthy littermate control mice (n=16).

[0081] Mice were anesthetized with 2% isoflurane and the hair on their hind limbs was removed. Then, needle probes were inserted into the gastrocnemius muscle and optical fibers were placed. Two needle probes were inserted into each muscle (the medial and lateral heads of the gastrocnemius muscle) to examine both legs.

[0082] Recording from human muscle samples Samples from 54 participants were examined. Briefly, these included 10 healthy volunteers without neurological disorders, 17 patients investigated for myopathy but found to have other conditions, and 27 patients ultimately diagnosed with myopathy. Samples were collected either during surgery for joint injuries (healthy volunteers) or by conchotome needle or open biopsy (patients investigated for / diagnosed with myopathy). Samples were flash-frozen and stored at -80°C until use. For spectral acquisition, an optical fiber was gently applied to the muscle sample. A total of 2–6 sites were examined, depending on the sample size.

[0083] analysis Spectral preprocessing, matrix factorization, and classification were performed using custom code in MATLAB (R2023a). Peak fitting was performed using Origin (2023). Spectra were measured from 900 to 1800 cm. -1 To avoid silica-related background from the optical fiber, the lower limit is 900 cm. -1 The spectrum was interpolated to a uniform wavenumber interval, and the amide I region was set to 1590-1720 cm -1 The spectra were then windowed (mean) so that one mouse / human muscle sample would present one spectrum for subsequent analysis. Background removal was performed using a rubber band algorithm, followed by smoothing (a second-order Savitzky-Golay filter with a window width of 5 data points). Vector normalization was then performed; however, this produced spectra that arbitrarily crossed zero, so the minimum spectral intensity was added to all spectra prior to nonnegative matrix factorization to remove negativity.

[0084] For peak fitting, group means were generated and scaled (0-1). A Lorentzian / Gaussian (Voigt) mixture function was used. For preclinical data, the peak at 1601 cm -1 and 1615cm -1 (aromatic amino side chain), 1635 cm -1 (irregular), 1652cm -1(α helix), 1663 cm -1 (β sheet), 1677 cm -1 Six peaks centered around 1705 cm (irregular) were used. For the analysis of human samples, an additional peak at 1705 cm -1 (irregular) peaks were included. In both preclinical and clinical analyses, the onset height of each peak was the spectral intensity of amide I at that wavenumber. The full width at half maximum was up to 30 cm. -1 The proportion of aromatic amino acids and secondary structure components was reported as the proportion of a particular peak relative to all peaks used in the fit.

[0085] The secondary structure ratio was calculated using the percentage of the integrated area under the peak of interest (percentage of all peaks).

[0086] Spectral patterns were derived using a hierarchical alternating least-squares nonnegative matrix factorization algorithm optimized for low-rank solutions. Briefly, nonnegative matrix factorization approximates the original data (A, an n × m matrix, where n is the number of samples and m is the matrix length or number of observations per sample) as the product of two low-rank matrices, A = W H , where W represents the derived nonnegative spectral patterns, and matrix H represents the relative importance (weights or coefficients) of those patterns for each sample. The number of selected spectral patterns (ranks) was determined by calculating the root-mean-square residuals of randomly divided healthy samples in both preclinical and human datasets, since the difference between these two matrices is considered to represent biological noise. To estimate the relative contribution of different secondary structures in each mode, we used the α-helix (1650-1658 cm ). -1 ), β-sheet (1664-1673cm -1 ), irregular (1630-1640, 1674-1689 and 1700-1710 cm -1 The area under the wavenumber region for each mode was integrated. Secondary structure ratios were calculated using the fraction of the integrated area under a peak of interest as a percentage of the total area of ​​that mode. Mode coefficients (weights) were compared using an unpaired t-test (GraphPad Prism, version 9).

[0087] For classification, nonnegative factorization weights were input into a linear discriminant model, and performance statistics were derived (MATLAB) using stratified 10-fold cross-validation to ensure balanced classes in the training set. For the three-group mouse analysis, the area under the receiver operating characteristic curve (AUROC) was calculated using a one-versus-all approach, which reduces the multiclass classification to a set of binary classifications. The average AUROC was calculated using "macro-average" (averaging all one-versus-all binary results).

[0088] result In vivo Raman spectra were collected from the mdx model of Duchenne muscular dystrophy (a type of primary muscle disease) at two stages: disease onset and disease establishment. G93A In vivo spectra were collected from a model ("neurogenic disease resulting from motor neuron loss") during the established disease phase. Spectra were also obtained from age-matched non-transgenic / wild-type healthy mice (96 in total). Amide I peak fitting was performed to explore the secondary structure of the protein through four Gaussian / Lorentzian mixed profiles representing α-helical, β-sheet, and disordered structures (Figure 7). Two additional curves representing aromatic amino acids were included. Detailed peak characteristics are shown in Supplementary Table 1. Notably, the full width at half maximum was within the resolution of the system. Using the percentage of area under each curve, we observed a decrease in α-helical content in the mdx model (48.3% vs. 73.2% in healthy muscle). SOD1 G93A A smaller decrease in α-helix was observed in mdx muscle (68%; Figure 7). Correspondingly, there was an increase in β-sheet content in mdx (19.3% vs. 4.6% in healthy muscle, SOD1 G93A The disorder content was increased in mdx (20.3% compared to 10.8% in normal muscle), and SOD1 G93A The ratios of α-helix to β-sheet and β-sheet to NR were also significant (Figure 7E).

[0089] Figure 7 shows preclinical peak fitting of healthy, myopathic, and neurogenic muscles, demonstrating a reduction in α-helical conformation in myopathy. In particular, Figures 7A-7C show the results of preclinical peak fitting of healthy, mdx, and SOD1 mice. G93A Figure 7D shows peak fitting using standard peaks within the amide 1 region of mouse. Figure 7D shows each resolved peak as a percentage of the total area. A decrease in alpha helix and a concomitant increase in beta sheet are evident in the myopathy model (mdx). Figure 7E shows the conformational ratios of different protein secondary structures. A decrease in alpha helix / increase in beta sheet is evident in the myopathy mdx model, as well as an increase in the beta sheet / disorder ratio.

[0090] Quantification at the individual sample (or mouse) level was performed using non-negative matrix factorization. As evident in Figure 8, the spectral modes dominant in mdx (i.e., mdx had significantly higher scores for those modes) exhibited patterns associated with β-sheet and disordered conformations (e.g., modes 3 and 5). These modes had a low α-helix:β-sheet ratio. In contrast, SOD1 G93A Mode 2, which was more dominant in mdx, lacked β-sheet regions. When these modes were input into a three-group linear discriminant analysis algorithm, the average area under the receiver operating characteristic curve was 0.75 (Figure 9), and mdx was the most successful in identifying mdx.

[0091] Figure 8 shows that the spectral patterns obtained by nonnegative matrix factorization indicate differences in three-dimensional structures. The unique spectral patterns (left) output from nonnegative matrix factorization indicate differences in protein structure (center) and the differences in the structure of healthy mice, mdx mice, and SOD1. G93A The differences between mice (right) are shown. The distinct conformational fingerprints and their dominance in different mice are consistent with those observed by peak fitting (*P<0.05, **P<0.01, ***P<0.001, ****P<0.0001).

[0092] FIG. 9 is a table showing the three-group classification performance of in vivo preclinical data using the nonnegative matrix factorization mode and the linear discriminant model.

[0093] The matrix factorization technique employed constrains the output to a non-negative distribution. This results in profiles that are easier to interpret than those obtained, for example, by principal component analysis (PCA). In this way, the spectral modes provide a straightforward shape for integrating the area under a wavenumber window. This approach is intentionally kept simple, and while the non-negative constraint often reduces the profile to zero, note that some area values ​​will be affected by the relative importance of neighboring wavenumbers. However, the results are consistent with simple visual inspection of the modes and their peaks, as well as more traditional peak-fitting data.

[0094] A preferred advantage of using a non-negative approach is that the importance of each spectral mode for each sample allows quantification at the individual spectral level.

[0095] We next assessed whether a similar conformational fingerprint is evident in human tissues. Because muscle biopsies are rarely performed in patients under investigation for "neurogenic" (e.g., nerve / motor neuron-related) diseases like ALS, we analyzed muscle biopsy samples that could be divided into "non-myopathic" and "myopathic" groups (see Methods). Peak fitting again demonstrated a decreased α-helix:β-sheet ratio in the myopathic group, along with a concomitant decrease in the α-helix:NR ratio (Figure 10). This time, we also uncovered an increase in aromatic amino acid side chains. These residues cluster in the core of folded proteins and play a key role in the stability of protein structure. During modifications such as unfolding, these amino acids become exposed, thus providing an indicator of muscle health.

[0096] Figure 10 shows peak fitting from Raman muscle spectra of human samples from patients with and without myopathy. In particular, Figures 10A and 10B show peak fitting within the amide I region. Figure 10C shows each resolved peak as a percentage of the total area. The myopathy group shows a decrease in α-helix and an increase in other conformations and aromatic amino acids. Figure 10D shows the ratios of different protein conformations. A decrease in α-helix and an increase in β-sheet / disordered structures is evident in myopathy.

[0097] Nonnegative matrix factorization spectral modes were derived, and the most dominant mode in the "non-myopathic" group exhibited a relatively high α-helix:β-sheet ratio. In contrast, mode 3, which appears to be dominant in the myopathic group, exhibited a low α-helix:β-sheet ratio and a low α-helix:irregular ratio (Figure 11). Using these three spectral modes in linear discriminant classification demonstrated a classification accuracy of 78% (Figure 12), comparable to results using the entire spectrum, with the added benefit of making the data more biologically interpretable. The human sample set included a variety of human myopathies, which may have differences in protein structure, making classification more challenging with limited sample numbers. Human muscle pathologies are not uniform within samples (or muscles). Targeting the Raman probe to the desired site, for example, by combining it with electromyography, can also improve disease detection.

[0098] Figure 11 shows the spectral patterns obtained by non-negative matrix factorization in human muscle. In particular, the unique spectral patterns output from non-negative matrix factorization (left) indicate differences in protein structure (center). Significant differences were observed in mode 1 (***P<0.001).

[0099] FIG. 12 is a table showing the two-group classification performance of human ex vivo samples using the non-negative matrix factorization mode and the linear discriminant model.

[0100] conclusion We demonstrated that conformational assessment of the Raman amide I region has the potential to identify muscle pathologies. This approach translates the in vivo preclinical paradigm to ex vivo human tissue using a fiber optic system capable of in vivo human recording. The application of matrix factorization demonstrated that quantitative information on biologically relevant changes in protein secondary structure can be obtained and used to identify muscle pathologies. Thus, conformational fingerprints are shown to be novel, translational biomarkers for neuromuscular diseases, particularly myopathies.

[0101] 13-15 are described below and show exemplary training and classification performance results of a machine learning module trained using hierarchical modeling and applied according to the fifth aspect, where the data is suitable to be acquired by a muscle probe according to the first aspect or a system according to the second aspect, and the method is suitable to provide a digital biomarker according to the sixth aspect.

[0102] Methods and Results Fiber-optic Raman spectroscopy A fiber optic Raman system was used according to the studies of Figures 7A to 12.

[0103] Preclinical records Breeding was performed in a specific pathogen-free environment, and experimental work was carried out in a standard preclinical facility (12-hour light / dark cycle, room temperature 21°C). As a muscle disease model, the mdx model of Duchenne muscular dystrophy was used after both 30 days (n = 16, representing "acute disease") and 90 days (n = 16, representing "chronic disease"), along with matched wild-type healthy control mice (n = 16 at both 30 and 90 days). SOD1 in motor neuron disease G93A The model was used after 90 days (n=16) as a model of "neurogenic" disease (diseases involving nerve or motor neurons). C57BL / 10ScSnOlaHsd (C57BL / 10) mice (matched to the mdx mouse background) after 30 and 90 days also expressed SOD1.G93A Non-transgenic healthy littermate control mice from the colony were also used. These wild-type and non-transgenic mice were then pooled into a single "healthy" category.

[0104] Mice were anesthetized with 2% isoflurane and the hair on their hind limbs was removed. Then, needle probes were inserted into the gastrocnemius muscle and optical fibers were placed. Two needle probes were inserted into each muscle (the medial and lateral heads of the gastrocnemius muscle) to examine both legs.

[0105] Data analysis A custom MATLAB code was used to analyze the spectra from 900-1800 cm -1 All analyses involved interpolation to integer wavenumber intervals, followed by background subtraction using an adaptive iteratively reweighted penalized least-squares algorithm, Savitzky-Golay smoothing (order 2, frame length 5, standard normal variate normalization). A rolling window (window length: 20–25 wavenumbers) was then used to remove outlying spectra (>3 standard deviations from the group mean).

[0106] Using the PLS Toolbox (Eigenvector Research Inc, USA), the Kennard Stone algorithm was used to separate the spectra into training and test sets (70% training / 30% test). Because each mouse could have up to four spectra, this training / test division was performed while keeping all data from individual mice in either the training or test group (thus, data from individual mice did not span training / test groups).

[0107] Training models included normal mice (n = 42, total spectra = 96), acute myopathy mice (n = 17, total spectra = 39), chronic myopathy mice (n = 26, total spectra = 52), and neuropathic mice (n = 20, total spectra = 41). Test models included normal mice (n = 17, total spectra = 60), acute myopathy mice (n = 7, total spectra = 22), chronic myopathy mice (n = 11, total spectra = 33), and neuropathic mice (n = 7, total spectra = 19).

[0108] Next, partial least squares discriminant analysis (PLSA) models were constructed using the training dataset for the following two-class problems: healthy vs. "disease" (combination of acute myopathy, chronic myopathy, and neurogenic groups), myopathy (combination of acute and chronic myopathy) vs. neurogenic, and acute vs. chronic myopathy. Prior to generating the PLS-DA models, feature selection was performed using a variable importance in projection approach to reduce the number of variables included in the PLS-DA models. Venetian blind cross-validation (10 data folds, blinding depth 1) was used to construct the PLSDA models.

[0109] Next, we constructed a hierarchical classification model using an expert system approach based on the clinical decision-making process (schematically shown in Figure 13). In this hierarchical approach, we first separated healthy and diseased samples. Samples classified as diseased were passed to the next level, at which point they were classified as either neuropathic or myopathic. Finally, samples selected as myopathic were divided into acute and chronic myopathic groups.

[0110] The training dataset samples were then input into the hierarchical model and classification performance statistics for the individual spectra were derived, as can be seen in Figure 14. Performance statistics for the validation test dataset are shown in Figure 15 and demonstrate high specificity and accuracy in classifying both myopathy subtypes: acute and chronic myopathy.

[0111] 16A-17B are described below and show exemplary training and classification performance results of a machine learning module trained using m-modeling and applied according to the fifth aspect, where the data is suitable for being acquired by a muscle probe according to the first aspect or a system according to the second aspect, and the method is suitable for providing a digital biomarker according to the sixth aspect.

[0112] Methods and Results Raman spectroscopy Raman spectra and CMAP amplitudes were obtained as described herein in connection with Figures 2A-6B.

[0113] Multiblock modeling of Raman data combined with CMAP amplitudes Raman spectral analysis was performed using custom code in MATLAB (MATLAB 15 R2019b The MathWorks). Raw spectra were first interpolated to the integer wavenumber interval of 900–1800 cm-1 and then normalized (standard normal variate normalization). Raman spectra and CMAP amplitudes were subjected to block variance scaling and combined into a new dataset containing 902 variables per mouse (901 spectral wavenumbers and 1 CMAP amplitude). A partial least squares discriminant model was constructed using three latent variables with Venetian blind cross-validation (10 data folds, 1 blinding depth).

[0114] The confusion matrix and performance statistics are shown in Figures 16A and 16B, respectively, and are based on SOD1 G93A These results demonstrate high sensitivity and specificity in classifying mouse pairs and transgenic mice. These results represent an improvement over using Raman spectra alone (described below and shown in Figures 17A and 17B).

[0115] Multi-block modeling of Raman data alone Raman spectral analysis was performed using custom code in MATLAB (MATLAB 15 R2019b The MathWorks). Raw spectra were first analyzed from 900 to 1800 cm -1 The data were interpolated to integer wavenumber intervals, followed by normalization (standard normal variate normalization) and mean-centering. A partial least squares discriminant model was constructed using one latent variable with Venetian blind cross-validation (10 data folds, blinding depth 1).

[0116] The confusion matrix and performance statistics are shown in Figures 17A and 17B, respectively, and compared to the combined modeling of both Raman and CMAP data (shown in Figures 16A and 16B), SOD1 G93A It can be seen that the sensitivity and specificity for classifying mice from non-transgenic mice is slightly lower.

[0117] The comparative improvement when CMAP data is combined with data obtained using Raman spectroscopy, shown in Figure 16, demonstrates the improved classification performance of the model when Raman data is combined with CMAP data.

[0118] 18 , a schematic diagram of an exemplary embodiment of a system 700 according to a second aspect of the present disclosure is shown. The system 700 comprises a muscle probe 702, substantially as described herein, including an EMG needle 704 positioned for insertion into muscle tissue, the EMG needle having an outer wall 710 positioned to transmit electrical signals 712 obtained from the muscle tissue to an EMG device 714. The needle 704 further comprises a tubular core electrode 716 housed within the outer wall 710, the core positioned to transmit electrical signals 718 from the muscle tissue to the EMG device 714. Upon receiving the electrical signals 712, 718, the EMG is configured to output EMG data 720 characterizing the electrophysiological activity of the muscle tissue for storage in a memory 724 of a processing unit 722.

[0119] The probe 702 further comprises a plurality of optical fibers 706 enclosed within the tubular core electrode 716, the plurality of optical fibers 706 including an emitting optical fiber 726 positioned to receive light 728 from a light source 730 and transmit the light 728 toward muscle tissue. The plurality of optical fibers 706 further comprises three receiving optical fibers 732 positioned to receive Raman scattered light from the muscle tissue and transmit inelastically scattered light 734 to a spectrometer 736. The spectrometer 736 is positioned to generate spectral data from the Raman scattered light 734 and transmit the spectral data 740 to a memory 724 of the processing unit 722. A processor 742 of the processing unit is positioned to access the spectral data and electromyogram data 744 from the memory and process the data. The processed data 746 may be transmitted to the memory 742 for storage. It will be understood that the system 700 can be used in any manner within the scope described herein. For example, electromyogram data may be acquired prior to acquiring spectral data, and the electromyogram data may also be used to inform positioning of the probe 702 for acquiring the spectral data. In this manner, the electromyogram data may guide placement of the probe in a desired location for acquiring the spectral data (and optionally further electromyogram data) for use in determining a clinical outcome as described herein or for generating a digital biological fingerprint according to the fourth aspect.

[0120] 19 , a flowchart listing steps of an exemplary embodiment of a method 800 according to a third aspect of the present disclosure is shown. The method includes: receiving an electrical signal from an electromyography needle, the electrical signal indicating muscle electrical activity 802; determining a target muscle site using the electrical signal 804; directing a Raman spectroscopy probe to the target muscle site 806; and receiving Raman spectroscopic data from the Raman spectroscopy probe, the Raman spectroscopic data characterizing the target muscle site 808. It will be appreciated that in embodiments, spectroscopic data may be obtained instead of an electrical signal in step 802, and the spectroscopic data may be used to determine the target muscle site in step 804. Thus, step 806 may instead include directing an electromyography needle to the target site to receive electromyography data and characterize the target muscle site in step 808.

[0121] In the particular example shown, the method further includes step 810 of using the Raman spectroscopic data and optionally the electrical signals to determine one or more of an indicator of a disease state; a prediction of a disease state; a predicted disease prognosis; or a predicted response to treatment. It will be appreciated that the electrical signals may be processed, for example, by an electromyograph, to provide a record of motor unit action potentials and other related waveforms prior to use in said determining step 810. Possible processes for making said determination may include any suitable process, such as comparing the Raman spectroscopic data and optionally the electrical signals with stored Raman spectroscopic data and optionally stored electrical signal data. Other processes include utilizing a machine learning module trained on stored Raman spectroscopic data and optionally stored electrical signal data to perform said determining step. In some embodiments, such determination may include generating a digital fingerprint by a processor using the Raman spectroscopic data and optionally the electrical signals (or data derived therefrom, such as Raman spectra or records of motor unit action potentials and other related waveforms). Such fingerprints may represent digital biomarkers that characterize one or more of the following: one or more neuromuscular diseases; prognosis of muscle and / or related diseases; or an indication of response to treatment of muscle and / or related diseases.

[0122] Within the scope of the present disclosure, further embodiments not described above may be envisioned without departing from the scope of the appended claims. For example, the specific example described utilizes Raman spectroscopy data. It will be appreciated that optical fiber can be used to provide any suitable optical spectroscopic assessment (such as fluorescence or Brillouin spectroscopy), and that the optical fiber allows for the combination of electromyography data with optical spectroscopic data, providing a muscle probe that improves the diagnostic pathway for patients with neuromuscular disorders and constitutes minimally invasive bedside testing of muscle health. Furthermore, in the specific example described, the muscle probe includes a tubular electrode that houses an optical fiber therein. It will be appreciated that embodiments in which the electrode includes a wire that extends inside the needle alongside the optical fiber are also understood. It will also be appreciated that other embodiments in which the electrode forms a coating on at least one of the optical fibers are also understood. In all embodiments in which the outer wall of the needle forms the electrode of the electromyography needle, the active electrode housed therein is electrically insulated therefrom.

Claims

1. 1. A muscle probe comprising an elongated needle having an outer wall surrounding an interior of the needle, the interior of the needle comprising: Core EMG electrodes and; one or more optical fibers; the needle is positioned to be inserted into a muscle and further positioned to detect electrical activity from the muscle; A muscle probe, wherein the one or more optical fibers are positioned to direct incident light from a light source toward a target area of ​​the muscle and further positioned to receive scattered light from the target area.

2. 10. The muscle probe of claim 1, wherein the scattered light comprises inelastically scattered light for evaluation using optical spectroscopy.

3. 3. The muscle probe of claim 2, wherein the inelastic scattered light comprises one or more of: Raman scattered light; fluorescent scattered light; and Brillouin scattered light.

4. 4. The muscle probe of claim 1, claim 2 or claim 3, further comprising a cannula extending along the needle interior, the cannula containing the core electromyography electrode and / or one or more optical fibers.

5. The muscle probe of claim 4 , wherein the core electromyography electrode is formed from at least a portion of the cannula.

6. 6. The muscle probe of claim 4 or claim 5, wherein the cannula and / or one or more optical fibers are arranged to move along the interior of the needle.

7. 7. The muscle probe according to claim 1, wherein the core electromyographic electrode forms a coating disposed on at least one of the optical fibers.

8. the one or more optical fibers: at least one delivery fiber positioned to direct incident light from said light source towards a target area of ​​said muscle; and at least one collecting fiber positioned to receive scattered light from the target area.

9. The muscle probe of claim 8 , wherein the one or more optical fibers comprise more collection fibers than delivery fibers.

10. 10. The muscle probe of claim 8 or claim 9, wherein the at least one delivery fiber and / or the at least one collection fiber is equipped with one of: an in-line short-pass filter; an in-line band-pass filter; an in-line long-pass filter; a notch filter.

11. 1. A system for acquiring electromyographic and optical spectroscopy data from a muscle, comprising: a muscle probe positioned for insertion into a muscle, the muscle probe comprising a needle and one or more optical fibers; a light source positioned to provide incident light for transmission by the one or more optical fibers to a target area of ​​the muscle; an optical spectrometer positioned to receive scattered light from the one or more optical fibers; an electromyography device positioned to receive electrical signals from the needles; The system, wherein the needle comprises an outer wall containing a needle interior and a core electrode disposed within the needle interior, and the one or more optical fibers are located within the needle interior.

12. the one or more optical fibers: at least one delivery fiber positioned to direct incident light from said light source towards a target area of ​​said muscle; and at least one collection fiber positioned to receive scattered light from the target area.

13. 13. The system of claim 12, wherein each of the at least one delivery fiber and / or the at least one collection fiber comprises one of: an in-line bandpass filter; an in-line shortpass filter; an in-line longpass filter; or a notch filter.

14. the electromyography device configured to: determine electromyography data using the electrical signals; 14. The system of claim 11, claim 12 or claim 13, wherein the optical spectrometer is configured to: use the received scattered light to determine an optical spectrum characteristic of the target area.

15. 15. The system of claim 14, further comprising a memory arranged to store the optical spectrum and electromyogram data.

16. 16. The system of claim 15, further comprising a processor, the processor comprising: processing the electromyogram data and using the electromyogram data to determine the target region; and / or processing the optical spectrum and optionally the electromyogram data; and determining a data fingerprint of the target region using the optical spectrum and optionally the electromyogram data.

17. The processor further comprises: comparing the data fingerprint of the target region to one or more stored data fingerprints; b.

17. The system of claim 16, wherein the system is arranged to use the comparison to determine one or more of: an indicator of a disease state; a prediction of a disease state; a predicted disease prognosis; or a predicted response to treatment.

18. 18. The system of claim 16 or claim 17, wherein the processor comprises a machine learning module trained using a plurality of stored fingerprints, the machine learning module being arranged to process the data fingerprints and output one or more of: an index of disease state; a prediction of disease state; a predicted disease prognosis; or a predicted response to treatment.

19. The system of any one of claims 11 to 18, wherein the light source is a laser.

20. 20. The system of claim 19, wherein the incident light comprises a wavelength selected from the near-infrared spectrum.

21. receiving, by a computer, electrical signals from the electromyography needles indicative of electrical activity in the muscles; determining, by the computer, a target muscle site based on the electrical signal; outputting the target muscle site by the computer for directing an optical spectroscopic probe to the target muscle site; A computer-implemented method comprising receiving, by the computer, optical spectroscopic data characterizing the target muscle site from the optical spectroscopic probe.

22. receiving, by a computer, optical spectroscopy data characterizing the muscle from the optical spectroscopy probe; determining, by the computer, a target muscle site based on the optical spectroscopic data; outputting the target muscle site by the computer for directing an electromyography needle to the target muscle site; A computer-implemented method, wherein the computer receives, from an electromyography needle, an electrical signal indicative of electrical activity within the muscle at the target muscle site.

23. 23. The method of claim 21 or claim 22, further comprising determining, by the computer, based on the optical spectroscopic data, and optionally further based on the electrical signal, one or more of: an indicator of a disease state; a prediction of a disease state; a predicted disease prognosis; a predicted and / or measured response to treatment.

24. 24. The method of claim 23, wherein the determining is performed by processing the optical spectroscopic data and optionally the electrical signals using a machine learning module trained using stored optical spectroscopic data and optionally stored electrical signals.

25. 1. A method for determining muscle pathology at a target muscle site in a subject, comprising:

1. A computer system including at least one processor and a memory storing at least one program for execution by the at least one processor, the at least one program comprising: instructions for acquiring a data set in electronic form, the data set including test optical spectroscopy data samples acquired from target muscle sites of the subject; and applying the dataset to a machine learning classifier trained using stored optical spectroscopy data, thereby determining a muscle pathology in the subject.

26. 26. The method of claim 25, wherein the optical spectroscopic data sample is determined using one or more of: Raman scattering; fluorescence scattering; Brillouin scattering at a target muscle site of the subject.

27. 27. The method of claim 25 or claim 26, wherein the program further comprises instructions for: isolating, from the test optical spectroscopic data sample, a spectral region that includes spectral data characterizing at least one protein secondary structure.

28. 28. The method of claim 27, wherein the spectral region is obtained from the amide I band.

29. 29. The method of claim 27 or claim 28, wherein the spectral data characterizes at least a proportion of alpha helices in the target muscle site and at least a proportion of beta sheets in the target muscle site.

30. 30. The method of any one of claims 25 to 29, wherein the target muscle site is determined using electrical signals from an electromyography needle, the electrical signals indicative of electrical activity in the muscle.

31. the dataset further includes test electrical signal data from an electromyography needle, the test electrical signal data indicative of electrical activity at a target muscle site of the subject; 31. The method of any one of claims 25 to 30, optionally further wherein the machine learning classifier is further trained using stored electrical signal data.

32. 32. The method of claim 31 , wherein the electrical signal data includes data indicative of one or more of: motor unit action potentials at the target muscle site; motor unit action potential morphology at the target muscle site; motor unit action potential configuration at the target muscle site; motor unit action potential recruitment at the target muscle site; spontaneous muscle activity at the target muscle site; and compound muscle action potential (CMAP) amplitude at the target muscle site.

33. 33. The method of any one of claims 25 to 32, wherein the machine learning classifier is generated using at least one selected from: matrix factorization; hierarchical modeling; multi-block modeling; data fusion modeling; principal component analysis. A

34. 34. The method of any one of claims 25 to 33, wherein the muscle pathology is selected from one of: acute myomyopathies; chronic myomyopathies; inflammatory myopathies; immune-mediated myopathies; dystrophic myopathies; mitochondrial myopathies; hereditary myopathies; congenital myopathies; metabolic myopathies; toxic myopathies; endocrine myopathies; infectious myopathies; severe myopathies; muscular dystrophies; neurogenic conditions.

35. 1. A digital biomarker determined using either optical spectroscopy data obtained from muscle or a combination of optical spectroscopy data and electromyography data obtained from muscle, comprising: one or more neuromuscular disorders; prognosis of said muscle and / or diseases associated therewith; A digital biomarker characterizing one or more indicators of response to treatment of said muscle and / or a disease associated therewith.

36. 36. The digital biomarker of claim 35, wherein the optical spectroscopic data characterizes the muscle's alpha helix fraction and beta sheet fraction.

37. 37. The digital biomarker of claim 36, wherein the ratio of alpha helices to the ratio of beta sheets is less than a predetermined threshold.

38. 38. The digital biomarker of claim 37, wherein the predetermined threshold is the ratio of alpha helices to beta sheet ratio associated with healthy muscle.

39. 39. The digital biomarker of claim 37 or claim 38, wherein the predetermined threshold is determined by a machine learning module trained on stored optical spectroscopic data obtained from muscle or a combination of stored optical spectroscopic data and stored electromyographic data obtained from muscle.

40. 40. The digital biomarker of any one of claims 35 to 39, determined using a muscle probe of any one of claims 1 to 10, a system of any one of claims 11 to 20 or a method of any one of claims 21 to 34.

41. A non-transitory computer-readable storage medium storing the digital biomarker of any one of claims 35 to 40.

Images