Characterization of tinnitus using functional near-infrared spectroscopy

Abstract

The present disclosure discloses a method for characterizing tinnitus in a subject using functional near-infrared spectroscopy (fNIRS). The method includes receiving, at a processing device, data including an fNIRS signal indicative of cortical activity in one or more regions of the subject's brain. The received data is processed using the processor device by inputting one or more features including one or more characteristics of the received data into a model. The model is configured to provide one or more classification results indicative of at least one characteristic of the subject's tinnitus based on the one or more features. The present disclosure also discloses a system for applying the disclosed method.

Description

Technical Field

[0001] The present disclosure relates to methods and systems for characterizing tinnitus.

Background Art

[0002] Tinnitus is a symptom characterized by the perception of unwanted sounds that do not exist externally within the ear. Chronic tinnitus affects 6 - 20% of adults and can be a harmful health condition that significantly impacts their quality of life. Approximately 20% of adults suffering from tinnitus experience severe tinnitus along with related symptoms such as depression, cognitive impairment, and stress.

[0003] Despite its high prevalence, there are currently no objective tests clinically used to evaluate tinnitus. Generally, the clinical evaluation of tinnitus relies on subjective feedback from individuals and can be inaccurate.

[0004] Any discussion of documents, operations, materials, devices, articles, etc. included in this specification should not be regarded as admitting that any or all of these matters form part of the basis of the prior art or are common general knowledge in the relevant field of the present invention, on the grounds that any or all of these matters existed prior to the priority date of each claim of the present application.

Summary of the Invention

Means for Solving the Problems

[0005] According to one aspect of the present disclosure, a method for characterizing a subject's tinnitus using functional near - infrared spectroscopy (fNIRS), comprising: receiving, in a processor device, data including an fNIRS signal indicating cortical activity in one or more regions of the subject's brain; processing the data received using the processor device, including inputting one or more feature quantities including one or more features of the received data into a model. A method is provided which includes the step of configuring the model to provide one or more classification results that show at least one feature of the subject's tinnitus based on the one or more features.

[0006] Functional near-infrared spectroscopy (fNIRS, also known as optical tomography) is a non-invasive optical imaging technique that can be used to measure changes in hemoglobin (Hb) concentration in cortical regions of the brain. Cortical brain activity can be inferred from these measurements. fNIRS signals may include signals indicating changes in oxyhemoglobin (O2Hb) concentration and / or deoxyhemoglobin (HHb) concentration in the subject's brain.

[0007] In some embodiments, the fNIRS signal may indicate activity in a subject's brain region including one or more of the frontal, left temporal, right temporal, and occipital cortical regions. In some embodiments, the subject's brain region includes each of the frontal, left temporal, right temporal, and occipital cortical regions of the subject's brain.

[0008] In some embodiments, the classification result may include the presence or absence of tinnitus in the subject. Additionally or alternatively, the classification result may include an assessment of the severity of the subject's tinnitus. In some embodiments, the severity assessment may classify tinnitus into mild or low tinnitus, or moderate or severe tinnitus. In other embodiments, the severity assessment may be selected from a wider range of categories. For example, possible assessments may include mild tinnitus, low tinnitus, moderate tinnitus and severe tinnitus, or other categories. In some embodiments, the severity assessment may be selected from a set of severity levels. For example, the severity assessment may be expressed on a numerical scale.

[0009] In some embodiments, the above model can further provide quantification of the loudness of tinnitus and / or the irritation caused by tinnitus (e.g., as perceived by the subject). Providing quantification of loudness and / or irritation may be useful, for example, when evaluating the impact of tinnitus on the subject's quality of life. Quantification of loudness and / or irritation may also be useful in defining parameters for developing treatments for tinnitus and for evaluating the relative success of such treatments.

[0010] Features can be extracted from received data using one or more methods. In some embodiments, features are extracted from received data using information gain. Information gain is a measure of entropy in the data that enables the identification of channel and O2Hb / HHb features using the most relevant classification information. For example, information gain may be used to select the most relevant features by ranking these features based on their weight or importance in classification. In other embodiments, alternative feature selection methods are available. Features may be extracted using, for example, the Gini exponent, SVM (Support Vector Machine) weights, wrapper methods, or one or more other suitable methods (e.g., different entropy methods).

[0011] In some embodiments, the model may include an algorithm. In some embodiments, the model may include a trained model. The model may be trained using an artificial intelligence (AI) algorithm based on, for example, one or more of the above features. The above one or more of the above features may be mapped to subjective measures of tinnitus features.

[0012] This model can provide classification results using a classification algorithm. The classification algorithm may be selected from, for example, Naive Bayes, K-Nearest Neighbors (KNN), Rule Induction, and Artificial Neural Networks (ANN). In other embodiments, alternative or customized classification algorithms may be used.

[0013] In some embodiments, the classification algorithm may be a multilevel hierarchical classification using a binary classifier at each level. This classification method can offer advantages over a single-level classifier by improving flexibility in selecting the most relevant features for each component binary classification module. Furthermore, this method can provide improved classification performance compared to a single-level multi-class classification.

[0014] In some embodiments, the method may further include the steps of applying a treatment for tinnitus and processing received data to detect changes in one or more characteristics of tinnitus as a result of applying the treatment. The changes detected as a result of applying the treatment may be used to evaluate the effectiveness of the applied treatment, etc.

[0015] In some embodiments, the model may be configured to provide a prognostic scale indicating whether a proposed treatment for tinnitus is effective in treating a subject's tinnitus. For example, the model may be configured to provide a prognostic scale indicating whether cochlear implantation is effective as a treatment for a subject's tinnitus.

[0016] In some embodiments, this method may include determining the quality of each fNIRS signal and removing signals of poor quality before processing the received data. The quality of each fNIRS signal can be determined, for example, based on the level of signal gain. Channels with high gain may indicate insufficient detected light intensity. Alternatively or additionally, the quality of each fNIRS signal may be determined based on the level of cardiac signal components (as an indicator of the degree of contact between the fNIRS device's recorded optode and the subject's scalp).

[0017] In some embodiments, preprocessing steps may be performed on the fNIRS signal to improve signal quality before processing the received data. In some embodiments, before processing the received data, an algorithm (e.g., a motion reduction algorithm) may be used to detect and remove motion artifacts from the fNIRS signal. Additionally or alternatively, signals containing motion artifacts exceeding a maximum threshold may be determined to be of poor quality and removed before processing the received data.

[0018] In any aspect of this specification, the fNIRS signal may include a plurality of signals obtained using an fNIRS system that measures cortical activity levels in one or more regions of a subject's brain. The fNIRS system may include a multi-channel fNIRS system. Each channel may be defined by a light source-detector pair. The fNIRS system may include a continuous-wave fNIRS system.

[0019] In some embodiments, the fNIRS signal may include signals from multiple channels located over different regions of the subject's brain. In some embodiments, at least one fNIRS signal may represent a systemic signal from at least one surface layer of the subject's head, including the scalp and / or skull. Such systemic signals may be used to preprocess the received data, for example, to remove unwanted artifacts or noise.

[0020] In some embodiments, the received data may include resting-state data. Resting-state data may include fNIRS signals indicating cortical activity in two or more regions of the subject's brain when the subject is in a resting state. In such embodiments, processing the data may further include determining at least one resting-state functional connectivity measure between at least two regions of the subject's brain, based on the resting-state data. Resting-state functional connectivity is a measure (or an indication of this level of coordination) between different neural populations. One or more features may comprise one or more features of at least one resting-state functional connectivity measure.

[0021] Resting-state data may include fNIRS signals obtained during periods when the subject is not receiving any stimulation (e.g., auditory or visual stimulation) or before any stimulation (e.g., auditory or visual stimulation) is supplied to the subject. In one example, resting-state data may include fNIRS signals acquired over a period of approximately 6 minutes. In other embodiments, resting-state data may be acquired over periods having shorter or longer durations.

[0022] In some embodiments, a resting-state functional coupling measure can be determined from resting-state data using seed analysis. Seed analysis may include selecting at least one region of the subject's brain as a seed region and correlating the seed fNIRS signal from the seed region with at least one fNIRS signal from at least one other region of the subject's brain. For example, in some embodiments, the temporal cortex of the subject's brain may be selected as a seed region. In other embodiments, other regions of the subject's brain may be selected as a seed region. The seed fNIRS signal from the seed region can be correlated with at least one fNIRS signal from other regions of the subject's brain. For example, the seed fNIRS signal can be correlated with the prefrontal cortex, occipital cortex, and / or contralateral temporal region of the subject's brain, or other regions of the subject's brain.

[0023] In some embodiments, two or more seed regions may be selected. For example, to determine each of the resting-state functional connectivity metrics for the left and right sides, seed regions on the left and right sides of the subject's brain may be selected. In some such embodiments, both the left and right temporal cortices of the subject's brain are selected as the left and right seed regions, respectively.

[0024] In some embodiments, the seed fNIRS signal may include the average of a plurality of fNIRS signals from the seed region. The seed fNIRS signal can be correlated with values obtained at channels in another region of the subject's brain. The correlation with each channel may be evaluated individually or averaged (e.g., across the entire region). The correlation can be performed on fNIRS signals including O2Hb concentration measurements and / or fNIRS signals including HHb concentration measurements.

[0025] In other embodiments, additional or alternative other connectivity analysis methods may be used to determine at least one resting-state functional connectivity metric. For example, other time domain and / or frequency domain methods for analyzing resting-state connectivity may be used. As an example, resting-state functional connectivity can be determined using independent component analysis (ICA). ICA enables the analysis of multiple networks in the brain and generates a large number of spatially independent components from the fNIRS signals. These components can separate the resting-state networks from each other and from noise. As another example, the resting-state functional connectivity metric can be determined using graph connectivity analysis. In graph connectivity analysis, the correlation between selected nodes is calculated, and this correlation is represented by an edge between the nodes. The way the nodes are connected and the strength of the connection can be compared between groups and between tinnitus patients with different severity levels.

[0026] In some embodiments, the received data includes evoked response data including fNIRS signals indicative of cortical activity in at least one region of the subject's brain due to at least one stimulus provided to the subject. The evoked response data may include fNIRS signals recorded during the provision of each stimulus. The evoked response data may include fNIRS signals recorded during and / or after the provision of at least one stimulus.

[0027] In such embodiments, one or more feature quantities may include one or more features of the evoked response data. In some embodiments, the feature quantity may include one or more amplitudes of the evoked response data. For example, the feature quantity may include peak amplitude, absolute peak amplitude, or average amplitude.

[0028] Alternatively or additionally, the feature quantity may include alternative features of the evoked response data. For example, the feature quantity may include one or more of variance, area under the curve, absolute area under the curve, peak power amplitude, entropy of the waveform, temporal content of the waveform, and spectral content of the waveform. As another example, the feature quantity may include the "principal components" of the response waveform calculated using principal component analysis. In principal component analysis, all evoked responses to one or more stimuli are considered, and the principal components of the responses are calculated. And many of these components (e.g., the first 10 components, or the components covering most of the variation in the data) can be used as feature quantities. In some cases, the feature quantity may be determined over a predefined period. This period can be, for example, the duration of the stimulus, 0 seconds to 5 seconds after the start of an auditory stimulus, 10 seconds to 15 seconds after the start of a visual stimulus, or other appropriate periods.

[0029] In some embodiments, a general linear model (GLM) can be used to compare evoked response data with a model of a typical hemodynamic response function. Based on this comparison, a coefficient can be generated indicating whether or not a response was detected. This coefficient can also function as a feature of the evoked response data. Additionally or alternatively, the features of the evoked response data can be compared with corresponding features of the model's hemodynamic response to determine the correlation between the evoked response data and the typical hemodynamic response.

[0030] At least one stimulus may include an auditory stimulus. In some embodiments, the auditory stimulus may include pink noise. In some embodiments, the auditory stimulus may be provided at a sound pressure level of approximately 65 dB. Additionally or alternatively, other forms of auditory stimuli may be used.

[0031] At least one stimulus may include a visual stimulus. The visual stimulus may be configured to elicit a strong cortical response from the subject. In some embodiments, the visual stimulus may include the display of a pattern, such as a black and white pattern (e.g., a checkerboard pattern). The checkerboard pattern may include a radial configuration containing concentric rings divided into sectors, where adjacent sectors are opposite colors. The visual stimulus may include repeated inversion (or flickering) of the pattern. The inversion of the pattern can be performed at a time frequency of about 7.5 Hz (i.e., about 15 inversions per second). Additionally or alternatively, other forms of visual stimuli may be used.

[0032] In some embodiments, at least one stimulus may include a plurality of individual stimuli that are supplied sequentially to the subject. At least one stimulus may include a plurality of auditory stimuli, a plurality of visual stimuli, or a combination of at least one auditory stimulus and at least one visual stimulus. In some embodiments, this method may include supplying a plurality of auditory stimuli and a plurality of visual stimuli sequentially. For example, the plurality of auditory stimuli and the plurality of visual stimuli may be supplied in an alternating order (e.g., alternating one auditory stimulus with one visual stimulus, or alternating one or more auditory stimuli with one or more visual stimuli). Alternatively, the plurality of auditory stimuli and the plurality of visual stimuli may be supplied in a different predetermined order, or in a substantially random order. In some embodiments, the plurality of auditory stimuli and the plurality of visual stimuli may be supplied (e.g., pseudo-randomly) such that the same type of stimulus (e.g., visual or auditory) does not occur more than twice consecutively.

[0033] Each stimulus may have a set duration. For example, in some embodiments, each stimulus may have a duration of approximately 15 seconds. However, stimuli with other durations may also be used. In some embodiments, each stimulus may have substantially the same duration. In other embodiments, the duration of the stimuli may vary.

[0034] Each stimulus may be followed by a pause period in which no stimulus is supplied. Each pause period may have a predetermined duration. For example, each pause period may have a duration of about 20 seconds to about 30 seconds, or longer. For example, each pause period may have a duration of about 20 seconds, about 25 seconds, about 30 seconds, or longer. In some embodiments, each pause period may have substantially the same duration. In other embodiments, the duration of the pause periods may vary. In some embodiments, the duration of the pause periods may be randomly selected. According to one aspect of the present disclosure, a non-temporary machine-readable storage medium is provided which includes instructions configured to cause a processor device to execute the method according to the present disclosure.

[0035] According to another aspect of the present invention, a system for characterizing a subject's tinnitus using functional near-infrared spectroscopy (fNIRS), The process includes receiving data containing fNIRS signals that show cortical activity in one or more regions of the subject's brain, The process of processing the received data, This includes inputting one or more feature quantities, each containing one or more features of the received data, into the model. A system is provided which includes a processor device configured to perform the steps of: providing one or more classification results that indicate at least one feature of the subject's tinnitus based on the one or more features.

[0036] In some embodiments, the received data may include resting-state data, which includes fNIRS signals indicating cortical activity in at least one of the subject's brain regions when the subject is at rest. In such embodiments, processing the data may further include determining at least one resting-state functional coupling measure between at least two regions of the subject's brain, based on the resting-state data. One or more features may comprise one or more features of the at least one resting-state functional coupling measure.

[0037] In some embodiments, the received data includes evoked response data, which includes fNIRS signals indicating cortical activity in at least one region of the subject's brain resulting from at least one stimulus supplied to the subject. In such embodiments, one or more features may include one or more features of the evoked response data. Generally, the system may be configured to perform one or more of the method steps described in the embodiments described above, including those relating to preprocessing of the received data, processing of the received data, or other processing.

[0038] This system may further include an fNIRS system configured to measure cortical activity levels in at least two regions of a subject's brain. The fNIRS system may include a multi-channel system. For example, the fNIRS system may include multiple channels configured to be positioned over each of at least two regions of the subject's brain. In one example, the fNIRS system may include multiple channels configured to be positioned over each of the frontal, left temporal, right temporal, and occipital regions of the subject's brain.

[0039] This system may further include an auditory stimulator configured to deliver auditory stimuli to a subject. This system may further include a visual stimulator configured to deliver visual stimuli to a subject.

[0040] In some embodiments, the system may further include a display configured to show one or more classification results. In some embodiments, the system may further include a user input module.

[0041] Generally, it will be recognized that a processor device according to an embodiment of the present disclosure may include one or more processing components for performing the processing steps relating to the present disclosure, and may also include one or more storage devices for storing data such as resting data and / or evoked response data. The processing components and / or storage devices may be located in one place, distributed across multiple locations, and interconnected via one or more communication links.

[0042] Throughout this specification, variations of the phrase "comprise," "comprises," or "comprising" shall be understood to indicate that the specified element, integer, or step, or group of elements, integers, or steps, is included, but not to indicate that other elements, integers, or steps, or other groups of elements, integers, or steps, are excluded. [Brief explanation of the drawing]

[0043] Although this is merely an example, embodiments of the present disclosure will be described below with reference to the attached drawings.

[0044] [Figure 1] This is a flowchart of the steps for a method of characterizing a subject's tinnitus using fNIRS according to one embodiment of the present disclosure. [Figure 2] This is a flowchart of the steps for a method of characterizing a subject's tinnitus using fNIRS, according to another embodiment of the present disclosure. [Figure 3] This is a schematic diagram of a system for characterizing tinnitus using fNIRS according to an embodiment of the present disclosure. [Figure 4] This figure shows an example of a pure-tone audiometry test performed on subjects prior to the fNIRS test. [Figure 5] This figure shows the detailed layout of the light source-detector channels in the multi-channel fNIRS system within the system shown in Figure 3. [Figure 6] This figure shows an example timeline of an fNIRS test session for acquiring data according to the method in Figure 2. [Figure 7] This figure shows a comparison of resting functional connectivity measures between temporal seeds with frontal channels and temporal seeds with occipital channels. [Figure 8A] This figure shows group-averaged O2Hb fNIRS auditory evoked responses recorded from channels in the upper left and right temporal cortices, respectively. [Figure 8B] This figure shows group-averaged O2Hb fNIRS auditory evoked responses recorded from channels in the upper left and right temporal cortices, respectively. [Figure 8C] This figure shows group-averaged O2Hb fNIRS auditory evoked responses recorded from channels in the upper left and right temporal cortices, respectively. [Figure 8D] This figure shows group-averaged O2Hb fNIRS auditory evoked responses recorded from channels in the upper left and right temporal cortices, respectively. [Figure 8E] This figure shows group-averaged HHb fNIRS auditory evoked responses recorded from channels in the upper left and right temporal cortices, respectively. [Figure 8F] This figure shows group-averaged HHb fNIRS auditory evoked responses recorded from channels in the upper left and right temporal cortices, respectively. [Figure 8G] This figure shows group-averaged HHb fNIRS auditory evoked responses recorded from channels in the upper left and right temporal cortices, respectively. [Figure 8H] This figure shows group-averaged HHb fNIRS auditory evoked responses recorded from channels in the upper left and right temporal cortices, respectively. [Figure 9A] This figure shows the O2Hb and HHb visually evoked responses recorded from the wedge region and the occipital channels above the superior occipital gyrus, respectively. [Figure 9B] This figure shows the O2Hb and HHb visually evoked responses recorded from the wedge region and the occipital channels above the superior occipital gyrus, respectively. [Figure 10A] This figure shows the group-averaged O2Hb and HHb auditory and visual evoked responses, respectively, averaged across the auditory and visual domains of interest. [Figure 10B] This figure shows the group-averaged O2Hb and HHb auditory and visual evoked responses, respectively, averaged across the auditory and visual domains of interest. [Figure 11] This figure shows the changes in the temporofrontal junction originating from O2Hb in relation to the duration of tinnitus (Panel A), and the changes in the temporo-occipital junction originating from HHb in relation to the subjective evaluation of sound intensity (Panel B). [Figure 12] This figure shows typical visual evoked responses from two cochlear implant wearers to a 15-second visual stimulus when switching their cochlear implants on and off. [Figure 13] This figure shows the temporofrontal junction of the right seed when the subject's cochlear implant was switched off, and the change in volume perception when the cochlear implant was switched on and off. [Figure 14] This figure shows O2Hb visually evoked responses recorded from the wedge and occipital channels in the superior occipital gyrus. [Figure 15A] This figure shows group-averaged HHb fNIRS auditory evoked responses recorded from channels in the upper left and right temporal cortices, respectively. [Figure 15B] This figure shows group-averaged HHb fNIRS auditory evoked responses recorded from channels in the upper left and right temporal cortices, respectively. [Figure 16A] This figure compares channel 24 fNIRS data from tinnitus subjects with channel 24 fNIRS data from control subjects, using the area under the absolute curve over 10-15 seconds as a feature of the visually evoked response. [Figure 16B] This figure compares channel 11 fNIRS data from tinnitus subjects and channel 11 fNIRS data from control subjects, using the average amplitude over 0-5 seconds as a feature of the auditory evoked response. [Figure 16C] This figure compares channel 24 fNIRS data from tinnitus subjects with channel 24 fNIRS data from control subjects, using absolute peak amplitude over a period of 0 to 15 seconds as a feature of the visually evoked response. [Figure 16D] This figure compares channel 21 fNIRS data from tinnitus subjects with channel 21 fNIRS data from control subjects, using the variance of epochs over a period of 0 to 15 seconds as a feature of the visually evoked response. [Figure 16E] This figure compares channel 34 fNIRS data from tinnitus subjects and channel 34 fNIRS data from control subjects, using the maximum power over a period of 0 to 15 seconds as a feature of the auditory evoked response. [Modes for carrying out the invention]

[0045] The following describes a method for characterizing a subject's tinnitus using functional near-infrared spectroscopy (fNIRS) according to embodiments of the present disclosure.

[0046] Referring to flowchart 100 in Figure 1, a method for characterizing a subject's tinnitus according to one embodiment of the present disclosure is shown. This method includes receiving data 110 in a processor device 130. The received data 110 includes fNIRS signals indicating cortical activity in one or more regions of the subject's brain.

[0047] fNIRS signals may include signals indicating changes in deoxyhemoglobin (HHb) and / or oxyhemoglobin (O2Hb) concentrations in cortical regions of the subject's brain. Cortical brain activity in the measured region can be inferred from these measurements.

[0048] fNIRS signals may be preprocessed by filtering, downsampling, or other methods. In some embodiments, fNIRS signals may be preprocessed to remove signals of poor quality. For example, the quality of each signal may be determined before further processing of the received data, and signals of poor quality may be removed from the data. Alternatively or additionally, before further processing of the received data, undesirable artifacts in the fNIRS signal (e.g., due to motion or other interference) may be filtered from the signal. This is described in more detail in Example 1 below. In another embodiment, for example, an embodiment using an AI algorithm, signals of poor quality and / or undesirable artifacts may be left in the signal, and the algorithm is trained to ignore data from these signals / artifacts.

[0049] Then, one or more feature quantities 150, each containing one or more features extracted from the received data 110, are input into the model 160.

[0050] Model 160 may include a pre-trained model. For example, Model 160 may be trained using an artificial intelligence (AI) algorithm based on one or more prior features mapped to subjective measures of tinnitus characteristics. Based on one or more features 160, Model 160 may be configured to provide one or more classification results 170 that represent at least one feature of a subject's tinnitus. The classification results 170 may include, for example, the presence or absence of tinnitus in the subject, the severity of the subject's tinnitus, a quantification of the loudness of the tinnitus, and / or a quantification of the irritability caused by the tinnitus. The classification results can be determined using an appropriate classification algorithm such as Naive Bayes, K-Nearest Neighbors (KNN), Rule Induction, Artificial Neural Networks (ANN), or Multilevel Hierarchical Classification.

[0051] As shown in Figure 2, the data may include resting data 111 and / or evoked response data 112 received by the processor device 130.

[0052] Resting-state data 111 includes fNIRS signals showing cortical activity in two or more regions of the subject's brain while the subject is in a resting state. Based on the resting-state data 111 (after any preprocessing steps have been performed), at least one resting-state functional coupling scale 140 is determined between at least two regions of the subject's brain. The resting-state functional coupling scale 140 can be determined, for example, using seed analysis (described in more detail below), but other methods for determining coupling may also be used as needed.

[0053] The evoked response data 112 includes fNIRS signals indicating cortical activity in at least one region of the subject's brain resulting from at least one stimulus supplied to the subject. The method may include the steps of supplying at least one stimulus to the subject and recording the evoked response data 112. The evoked response data 112 may correspond to fNIRS signals recorded during and / or after the supply of the stimulus.

[0054] One or more features 150 may include one or more features of the resting functional connectivity scale 140 and / or one or more features of the evoked response data 112.

[0055] One or more features 150 of the evoked response data 112 may include peak amplitude, absolute peak amplitude, or mean amplitude over a predefined period. Alternatively or additionally, features 150 may include one or more of the following: variance, area under the curve, absolute area under the curve, peak power amplitude, waveform entropy, waveform temporal content, waveform spectral content, "principal components" of the response waveform (e.g., calculated using principal component analysis), or other features of the evoked response data 112. The features of the auditory and visual evoked response data 112 may differ depending on the model 160 and the classification algorithm used. For example, Figures 16A–16E show various examples of comparisons of features extracted from fNIRS data from each channel of the auditory evoked response data or visual evoked response data. In each illustrated example, the selected combination of features and channels may show the overall group difference between the evoked responses of tinnitus subjects and control subjects. Model 320 determines the optimal combination of channels, features, and classification algorithms to provide classification results.

[0056] In some embodiments, this method can be used in the context of applying a treatment to address tinnitus and detecting subsequent changes in one or more characteristics of the tinnitus as a result of applying this treatment. In some embodiments, Model 160 may be configured to provide a prognostic scale indicating whether a proposed treatment to address tinnitus is effective in treating a subject's tinnitus. For example, the use of the method to derive a prognostic scale regarding whether cochlear implantation may alleviate symptoms of tinnitus is described in Example 2.

[0057] Figure 3 shows a schematic diagram of a system for characterizing a subject's tinnitus using fNIRS according to an embodiment of the present disclosure. The system includes a processor device 310 configured to receive data including resting-state data and / or evoked response data, which include fNIRS signals representing the subject's cortical brain activity. The processor device 310 is configured to process the resting-state data to determine at least one resting-state functional coupling measure between at least two regions of the subject's brain based on the resting-state data.

[0058] The processor device 310 is further configured to input one or more features of at least one resting functional coupling measure and / or one or more features of evoked response data into the model 320. The model 320 is configured to provide one or more classification results that represent at least one feature of the subject's tinnitus based on one or more features. The model 320 may be a trained model or other models. For example, the model 320 may be trained using an artificial intelligence (AI) algorithm.

[0059] Referring again to Figure 3, the fNIRS signal can be acquired using an fNIRS system 330 configured to measure cortical activity levels in at least two areas of the subject's brain. As shown in Figure 3, the processor device 310 can directly receive data containing the fNIRS signal from the fNIRS system 330. Alternatively or additionally, the fNIRS signal data may first be received, stored in an intermediate data acquisition device, and then received by the processor device.

[0060] The fNIRS system 330 may include a multi-channel fNIRS system in which each fNIRS channel is defined by a light source-detector pair. One representative fNIRS channel is illustrated in Figure 3 by a light source 331 and a detector 332. In some embodiments, the fNIRS system 330 may include multiple channels configured to be positioned over different regions of the subject's brain. In one such embodiment, as shown in Figure 5 and described in more detail in Example 1 below, the fNIRS system 330 includes multiple channels configured to be positioned over the frontal, left temporal, right temporal, and occipital regions, respectively.

[0061] The processor unit 310 can directly or indirectly control the operation of the fNIRS system 330. For example, the processor unit 310 may include an optical output module for controlling each light source 331 of the fNIRS system and a data input module for receiving fNIRS signals from each detector 332 of the fNIRS system 330. Alternatively, the fNIRS system 330 may be controlled by an fNIRS control unit located separately from the processor unit 310.

[0062] In embodiments where the received data includes evoked response data, the system 300 may include at least one stimulator for supplying at least one stimulus to the subject. For example, the at least one stimulus may include at least one auditory stimulus and / or at least one visual stimulus. Thus, the system may include an auditory stimulator 340 and / or a visual stimulator 350 configured to supply the respective auditory and / or visual stimuli to the subject.

[0063] As shown by the dashed lines in Figure 3, the auditory stimulator 340 and the visual stimulator 350 may be controlled directly or indirectly by the processor unit 310. For example, the processor unit 310 may include an audio stimulus output module configured to supply auditory stimuli to the auditory stimulator 340, and / or a visual stimulus output module configured to supply visual stimuli to the visual stimulator 350. Alternatively, the auditory stimulator 340 and / or the visual stimulator 350 may be controlled manually or in other ways by one or more controllers different from the processor unit 310 (together or independently).

[0064] Optionally, system 300 may further include a display 360. The display 360 may be configured to display at least one classification result. In some embodiments, the display 360 (or an alternative display) may be configured to display information related to the operation of the fNIRS system 330. In other embodiments, the visual stimulus device 350 can operate as a display that displays at least one classification result and / or information related to the operation of the fNIRS system 330. System 300 may also include one or more user input modules 370 to facilitate interaction between the user and system 300.

[0065] Example 1

[0066] We recruited 25 subjects with chronic subjective tinnitus (23 of whom experienced bilateral tinnitus) and 21 healthy adults with no history of tinnitus, neurological or hearing impairment to participate in this study. Data from three healthy subjects were excluded: two had long hair and poor signal quality, and one had a technical issue. Each subject participated in one test session.

[0067] Figure 4 shows the results of pure-tone audiometry tests performed on all subjects at frequencies of 0.25 kHz, 0.5 kHz, 1 kHz, 3 kHz, 4 kHz, 6 kHz, and 8 kHz to assess hearing thresholds prior to the start of the fNIRS test. There were no significant differences in average hearing thresholds across all frequencies for each ear between the groups. There were no significant differences in mean age between the tinnitus group and the non-tinnitus group.

[0068] The severity of tinnitus in each subject was assessed using the Tinnitus Handicap Inventory (THI). The THI is a 25-item test that quantifies the perceived severity of tinnitus on a scale of 0 to 100. The score range is associated with different severity levels (e.g., 0-16 for mild tinnitus, 58-76 for severe tinnitus). Prior to each record, participants with tinnitus were asked to rate the loudness and irritability of their tinnitus on a scale of 1 to 10. Demographic and clinical data are shown in Table 1 below.

[0069] [Table 1]

[0070] Data were acquired using a multi-channel continuous-wave fNIRS system 500 (NIRScout, NIRx Medical Technologies LLC) operating at 760nm and 850nm. Figure 5 shows a montage of channel locations relative to the subjects' heads. For each subject, a total of 16 light sources 510 (oblique parallel line markers) and a total of 16 detectors 520 (black markers) were positioned over regions of the subject's head 400 corresponding to the frontal, temporal, and occipital cortical regions of the subject's brain, by positioning each of them relative to the nasion 401 and inion 402 of the subject's head 400. White markers indicate potential light source / detector locations not utilized in this study. Each light source 510-detector 520 pair defines fNIRS channels 530, 531, respectively.

[0071] Using an ICBM-152 head model and NIRSite software (NIRxMedical Technologies LLC) that enables the export of MNI coordinates corresponding to channel locations, the light source 510 and detector 520 were positioned. Subsequently, these coordinates were used in an open-source Matlab script named "AtlasViewer" to identify the brain regions corresponding to each channel location, and it was confirmed that the auditory and visual cortices were covered in particular.

[0072] Of the most light source-detector pairs, the light source 510 and detector 520 were positioned 30 mm apart, forming 36 "long" channels 530 (shown by the black connecting lines between the light source 510 and detector 520 in Figure 5). In each of the four cortical regions (frontal, left and right temporal, and occipital), one "short" channel 531 (shown by an asterisk in Figure 5) was defined by positioning the light source 11 mm apart from the detector. This short channel 531 was configured to detect and record systemic signals from the superficial layers of the subject's head (including scalp and skull) that could interfere with the detection of deeper cortical signals. The recordings from the short channels were used to remove systemic artifacts from the fNIRS signals received from the long channels.

[0073] The following numbers were assigned to channels 530 and 531: Frontal region - channel numbers 1, 2, 3, 4, 5, 6, 7, 8, 26, 27, 28, 29; Left temporal region - channel numbers 9, 10, 11, 13, 14, 16, 17, 18; Right temporal region - channel numbers 30, 31, 32, 34, 35, 37, 38, 39; Occipital region - channel numbers 19, 20, 21, 22, 23, 24, 25, 40, 41, 42.

[0074] The estimated anatomical regions covered by each temporal channel are shown in Table 2 below. The supraoccipital channel covered the cuneiform region and the superior occipital gyrus.

[0075] [Table 2]

[0076] In this study, by using features from all long channels 530, the feature extraction algorithm was able to select the most relevant channels, and the evoked responses were mainly obtained from the relevant anatomical regions (e.g., auditory response features were obtained from auditory channels). However, in other embodiments of this method, fewer channels 530 may be used, which may allow for the use of a simplified test setup and / or faster computation time. Alternatively, in some embodiments, a larger number of channels may be used.

[0077] Inside a soundproof booth, presentation software (Neurobehavioral Systems, USA) is used to deliver multiple individual auditory stimuli to in-ear earphones used for hearing tests (ER-3A in-ear earphones, EA-RTONE). TM The samples were delivered to each subject binaurally by 165 GOLD (USA). Each auditory stimulus consisted of a 15-second pink noise segment, calibrated using a Norsonic sound level meter (Norsonic SA, Norway), and delivered at a sound pressure level (SPL) of 65 dB. The power of the pink noise was inversely proportional to the signal frequency, with the same power at different octaves (i.e., double the frequency). This is similar to how the human auditory system perceives sound.

[0078] Multiple visual stimuli were presented to each subject as a reversed display of a circular checkerboard pattern that reversed at a time frequency of 7.5 Hz (15 reversals per second). This pattern elicited a strong cortical response in individuals with good visual acuity. These images were essentially radial, composed of rings, divided into multiple sectors, with adjacent sectors being opposite colors (black and white).

[0079] Figure 6 shows an exemplary timeline of an fNIRS test session. This test session included three recording sessions 611, 612, and 613, with short breaks 621 and 622 in between. In this example, the breaks were 3 to 5 minutes long, but other durations were also possible.

[0080] The first recording session 611 included a 6-minute resting recording. During this recording, the subject was instructed to sit still with their eyes closed and not to fall asleep. No auditory or visual stimuli were provided to the subject during this session 611.

[0081] The second recording session 612 and the third recording session 613 each contained a series of evoked response recordings. In these recording sessions 612 and 613, the subject was given multiple 15-second stimuli 630 in sequence. In this example, individual auditory and visual stimuli were given in a random order, ensuring that no two consecutive stimuli of the same type occurred. However, other arrangements of stimuli were also possible.

[0082] Each stimulus 630 was followed by a non-stimulation interval 640. In this example, randomly selected non-stimulation intervals were 20 or 25 seconds. In other embodiments, alternative durations for the non-stimulation intervals were also available. Generally, the duration of each non-stimulation interval 640 can be chosen to allow sufficient time for any elicited response to subside and for cortical activity to return to baseline. In Figure 6, the dotted lines indicate repetition of the stimulus delivery pattern. In this example, each stimulus type (i.e., auditory or visual) was repeated a total of 10 times (6 times in the second recording, 4 times in the third recording, with a break in between), but other numbers of stimuli were also available in other examples. The total recording time (excluding breaks) was approximately 20 minutes.

[0083] In this example, fNIRS data was recorded at a sampling rate of 7.8125 Hz. However, other suitable sampling rates could also be used. In each recording session, data (i.e., from the fNIRS signal) was recorded continuously. In the evoked response recording sessions 612 and 613, the continuous data recordings were later correlated with the supply time of each stimulus 630, and data portions corresponding to each evoked response were extracted.

[0084] Data processing was performed using Matlab 2019 A (Mathworks, USA). fNIRS signals were preprocessed using the "NIRS Brain AnalyzIR" toolbox and a custom-created Matlab script. Channels with low signal quality were selected and excluded from further analysis according to the following criteria: First, channels with a gain exceeding 7 and indicating insufficient detected light intensity were excluded. The above gain was calculated using the NIRx instrument in the calibration procedure performed prior to each experiment. In addition, the cardiac signal component of the channels was confirmed using the SCI (scalp coupling index), which was calculated by bandpass filtering two detection signals at 760 nm and 850 nm at 0.2–2.5 Hz (22). This provided an index indicating the degree of contact between the optode (detector) and the scalp. Signals from optodes with good skin contact are highly correlated because they mainly contain heart rate data. Channels with an SCI value of less than 0.75 were excluded. On average, 13% of channels were excluded.

[0085] For the remaining channels, the following preprocessing steps were performed. For resting recordings, the unfiltered raw signals from each channel were downsampled to 1 Hz and converted to optical density. For evoked response recordings, the conversion to optical density was performed at the original sampling rate. Short channel correction was applied to the optical density data using the NIRS toolbox function ntbxSSR.m (parameter task set to 0). The corrected optical density for each long channel was calculated by subtracting a portion of the nearest short channel. This subtraction removed two sources of interference: variability measured from the scalp and overall variability such as systemic responses and respiration. Next, the modified Beer-Lambert law was used to estimate the concentration changes of oxygenated and deoxygenated hemoglobin (O2Hb and HHb, respectively).

[0086] We used seed analysis to confirm resting-state functional connectivity. In seed analysis, we selected cortical regions as seeds and confirmed connectivity with other brain regions by determining the correlation between the seed regions and other brain regions. In this example, we used two channels above the temporal cortex on both sides of the head as seed channels. We estimated that channels 9 and 10 on the left and channels 30 and 31 on the right covered the superior temporal gyrus and Heschl's gyrus (shown in Table 2). Then, on both sides, we averaged the signals from the two channels to obtain the left and right seeds, respectively. We calculated the correlation between the seed channels and other channels using whitening correlation (NIRS toolbox function nirs.sFC.ar_corr.m) (27). This is a robust correlation method that solves the sensitivity of fNIRS to spurious correlations due to slow hemodynamic signals, systemic physiological noise such as heart rate and respiration (serial correlations), and motion artifacts that may introduce abnormal noise structures. We then averaged the obtained values ​​for channels containing the frontal and occipital regions of interest (ROIs) for statistical analysis. The frontal ROI included channels over the superior frontal gyrus, medial frontal gyrus, medial orbit, and middle frontal gyrus (channels 1, 3, 4, 5, 6, 7, 8, 26, 27, 28, 29). The selected occipital ROI covered the cuneiform and superior occipital gyrus (channels 20, 21, 23, 24, 25, 41, 42). Whitening correlations were derived from both O2Hb and HHb signals and compared between groups.

[0087] To analyze evoked responses, motion artifacts were removed using the function "WaveletFilter" (outlier threshold set to 3). The signals were band-pass filtered between 0.01 and 0.12 Hz by applying zero-phase 8th-order Butterworth high-pass (0.01 Hz) and low-pass (0.12 Hz) filters, respectively. O2Hb and HHb concentrations were then estimated using the modified Lambert-Beer law. For each channel, the O2Hb and HHb signals were epoched from t=5 to t=30 seconds relative to the stimulus onset using the "EpochExtraction" function, which removes linear trends and baseline-corrected epochs by subtracting the baseline mean. Based on outlier detection, epochs with amplitudes 2.5 standard deviations larger than the epoch mean were excluded. For each condition recording auditory and visual responses, the mean values ​​of O2Hb and HHb activation were calculated over time frame intervals of 0–5 seconds (auditory response) and 10–15 seconds (visual). For statistical analysis, visually evoked responses were averaged across the occipital lobe channel, and auditory responses were averaged separately across the left and right temporal lobe channels.

[0088] Machine learning methods, including feature selection and classifiers, were used to combine resting-state features with evoked response signals from fNIRS channels over different cortical regions. The features input to these algorithms included the aforementioned auditory and visual response amplitudes, and frontal and occipital lobe connectivity measures. Here, by using features from all channels as individual inputs (without averaging across ROIs), the feature selection algorithms were able to automatically select the channels that best distinguished the groups. Features derived from O2Hb and HHb were used. Information gain was used to select the most relevant features by ranking these features based on their weight or importance in classification. Information gain is a measure of entropy in the data, enabling the identification of channels and / O2Hb / HHb features using the most relevant classification information.

[0089] Next, the selected characteristics were used to classify subjects into control groups or those experiencing tinnitus using four different classification methods. The classifier was also used to differentiate subjects based on the severity of their tinnitus. The tinnitus experienced by subjects was classified into mild / slight or moderate / severe (based on the THI scale). In the analyses described later, the data was divided into only two groups to increase the sample size; however, in other examples, more categories could have been used. For example, the severity classification could include mild, slight, moderate, and severe tinnitus as separate assessments. In other examples, other severity assessments could have been used to classify subjects into more groups.

[0090] The four classifiers evaluated were Naive Bayes, K-Nearest Neighbors (KNN), Rule Induction, and Artificial Neural Networks (ANN). In other examples, other suitable classification algorithms, such as multilevel hierarchical classification, could also be used. Tenfold cross-validation was used to evaluate the performance of these algorithms. This validation method involved randomly dividing the dataset into 10 subsets. One subset was kept for testing, and the remaining nine were used for training. This process was repeated across all 10 subsets (using one of the 10 subsets for testing each time), and the mean sensitivity (true positive rate), specificity (true negative rate), and accuracy of the classifiers were calculated. Classification accuracy or predictive performance was calculated as the number of correctly predicted samples relative to the total number of samples.

[0091] Figure 7 shows the difference in connectivity between temporal seeds with frontal channels and temporal seeds with occipital channels. Connectivity between left and right seeds with frontal O2Hb signals was higher in the tinnitus group, and the difference in right seeds was significant. The occipital connectivity value of right seeds obtained from HHb signals was significantly higher in the tinnitus group. This was not observed in connectivity of left seeds.

[0092] Figures 8A–8D show group-averaged O2Hb auditory evoked responses recorded from channels in the upper left and right temporal cortices. Figures 8E–8H show group-averaged HHb auditory evoked responses recorded from channels in the upper left and right temporal cortices. Channel numbers are indicated above each graph. Vertical dashed lines indicate stimulus onset and offset times at 0 and 15 seconds. The amplitude of auditory responses was averaged over the first 5 seconds after stimulus onset and compared between the left and right auditory regions using paired t-tests, and between groups using independent sample t-tests. This period was selected to capture the rise time or onset of responses that lasted an average of 5 seconds. There were no significant differences in left and right auditory responses. On average bilaterally, the tinnitus group had smaller auditory responses. This group difference was not observed in HHb responses.

[0093] Figures 9A and 9B show O2Hb and HHb visually evoked responses recorded from occipital channels above the cuneiform and superior occipital gyrus, respectively. Channel numbers are indicated above each graph. Vertical lines indicate the onset and offset times of the stimulus at 0 and 15 seconds. Visual responses lasted longer after stimulus onset than auditory responses. The average response amplitude over 10–15 seconds after stimulus onset was significantly larger in the control group.

[0094] Figures 10A and 10B show group-averaged O2Hb and HHb auditory and visual responses, averaged over the auditory and visual regions of interest (ROIs), respectively. The dotted lines represent the Standard Error of Mean (SEM). Vertical dotted lines indicate the onset and offset times of the stimulus at 0 and 15 seconds. This figure allows for a visual comparison of the waveforms averaged across the auditory and visual channels. The auditory response showed a pronounced onset or rise that lasted approximately 5 seconds after the stimulus. The visual response showed a more gradual rise that lasted approximately 15 seconds after the stimulus and had a longer duration.

[0095] Changes in fNIRS measurements associated with THI score, age, duration of tinnitus, hearing thresholds at 4 kHz and 8 kHz, and tinnitus severity assessed by subjective loudness and irritability were evaluated using multiple linear regression. Figure 11 shows changes in O2Hb-derived temporoporal connectivity with duration of tinnitus (Panel A) and changes in HHb-derived temporoporal connectivity with subjective loudness assessment (Panel B). O2Hb-derived connectivity between the left and right seeds and the frontal channel increased with duration of tinnitus, and the correlation on the right side became significantly closer. HHb-derived connectivity between the left and right seeds and the occipital channel significantly increased with subjective loudness assessment.

[0096] The feature sets of auditory, visual, and resting-state fNIRS for individual channels (rather than an averaged ROI) were used individually or in combination with a classifier. Features were weighted (or ranked) by using information gain as the feature extraction method. The results obtained using various classifiers are shown in Table 3 below.

[0097] [Table 3]

[0098] Using a Naive Bayes classifier, a single set of auditory-only features with weights greater than 0.45 was able to separate tinnitus subjects from controls with 78.3% accuracy (Table 3). Based on the weighting criteria, 36 auditory features were used (20 O2Hb-derived auditory response amplitudes and 16 HHb-derived auditory response amplitudes). Combinations of auditory, visual, and resting-state features with weights greater than 0.56, using rule induction, Naive Bayes, and neural network classifiers, also achieved accuracy of over 70%. The features used included 19 auditory features, 17 visual features, and 22 resting-state combined measures. Of these 58 features in total, 35 were derived from O2Hb signals and 23 from HHb signals. The combined measures in the selected features included more right-seed features than left-seed features, and more temporo-occipital features than temporo-frontal features. Using a Naive Bayes classifier with auditory features, we achieved the highest accuracy in classifying tinnitus subjects from controls. Using a Naive Bayes classifier with features from all three selected conditions using information gain, we achieved the highest sensitivity. An artificial neural network algorithm achieved similar sensitivity (71.41%) and specificity (74.62%). Furthermore, while KNN was used to classify tinnitus subjects from controls, the accuracy decreased (~60%).

[0099] Table 4 shows the classification results for distinguishing between mild / slight (n=18) and moderate / severe (n=7) tinnitus. Neural networks, KNNs, and rule-inducted classifiers were used to classify these tinnitus subjects, achieving the highest accuracy (over 75%) with combined scales having weights greater than 0.45 (Table 4). A total of 48 features (23 O2Hb-derived auditory response amplitudes and 25 HHb-derived auditory response amplitudes) were included in the majority of features from the right-seeded HHb temporofrontal and temporo-occipital scales. Using a neural network classifier resulted in a low specificity of 51.23%, but achieved the highest sensitivity (accurately predicting moderate / severe tinnitus patients) and accuracy.

[0100] [Table 4]

[0101] This study revealed that fNIRS can be used to distinguish subjects with tinnitus from control subjects and to identify fNIRS features associated with subjective assessments of tinnitus severity. Furthermore, the results of this study suggest that tinnitus characteristics such as loudness and perceived loudness can be individually measured using fNIRS.

[0102] In another study, the number of participants was increased to 52 individuals with chronic subjective tinnitus and 31 healthy adults with no history of tinnitus, neurological or hearing impairment. The demographic characteristics of the patients in the updated study are shown in Table 5 below. Patients were matched for age and hearing level.

[0103] [Table 5]

[0104] Figure 14 shows O2Hb visually evoked response data recorded from the wedge and occipital channels above the superior occipital gyrus of this enlarged group of subjects. Channel numbers are indicated above each graph. Vertical lines indicate the onset and offset times of the stimulus at 0 and 15 seconds. Visual responses lasted longer after stimulus onset than auditory responses. The average response amplitude over 10–15 seconds after stimulus onset was significantly larger in the control group. From this data, it was found that the channels in the visual domain that best distinguished tinnitus patients from controls were channels 24, 25, 41, and 42.

[0105] Figures 15A and 15B show auditory response data from the left and right temporal lobes of the enlarged subject groups, respectively. Channel numbers are indicated above each graph. Vertical lines indicate the onset and offset times of the stimulus at 0 and 15 seconds. In the left temporal lobe, channels 10, 11, 14, 16, 17, and 18 showed significant differences between tinnitus patients and controls. Of these, channels 10 and 11, 14 and 16, and 17 and 18 showed similar differences. In some cases, only one of these channel pairs may be necessary. The same channel pair (or region of interest in the brain) can also be used in the right temporal lobe. However, in the right temporal lobe, channels 35 and 37 showed significant differences between the groups. This difference between the left and right temporal lobes is thought to be due to the asymmetry of brain activity commonly seen in tinnitus patients.

[0106] Example 2

[0107] Cochlear implants (CIs) are devices that provide hearing impairments with the sensation of sound, and in some cases, can also suppress tinnitus. However, the mechanism by which cochlear implants affect tinnitus is unclear. In 4-26% of cases, tinnitus has been reported to worsen after cochlear implantation.

[0108] This study involved 10 cochlear implant recipients who experienced tinnitus and whose perception of tinnitus (i.e., perceived loudness or noise) had changed due to cochlear implant use. Resting data was recorded with the cochlear implant switched on and off. Evoked response data to visual stimuli was recorded for 15 seconds with the cochlear implant switched on and off. Auditory responses were not included in this protocol because they could not be recorded with the cochlear implant switched off.

[0109] Figure 12 shows representative visual evoked responses from two cochlear implant users to a 15-second visual stimulus when the cochlear implant was switched on and off. For each person, the mean value (solid line) and standard error (dashed line) of 10 averaged responses are shown. The first subject (TCI008, shown in panel A) reported that tinnitus was completely suppressed by the use of the cochlear implant. The second subject (TCI009, shown in panel B) experienced the opposite effect, that tinnitus sounded louder with the use of the cochlear implant. These conflicting effects were reflected in the onset (t=0~5s) of the visual response. When the cochlear implant suppressed tinnitus (TCI008), the onset response amplitude when the cochlear implant was switched on (gray trace) was smaller than the onset response amplitude when the cochlear implant was switched off (black trace). The opposite effect was observed when tinnitus worsened after cochlear implantation during this onset period (TCI009).

[0110] By comparing resting time recordings obtained with the cochlear implant switched on and resting time recordings obtained with the cochlear implant switched off, it became clear that the data obtained from fNIRS when the cochlear implant was switched off could predict whether or not an active cochlear implant is effective in alleviating tinnitus symptoms.

[0111] Figure 13 shows the temporofrontal connectivity of the right seed when the subjects' cochlear implants were switched off, and the changes in perceived volume when the cochlear implants were switched on and off. Positive numbers on the x-axis represent a decrease in perceived volume when the CI (Cochlear Interventional Index) was switched on. A resting functional connectivity score greater than 0.5 is associated with a decrease in the loudness of tinnitus when the cochlear implant was switched on. In this study, individuals with a connectivity score of less than 0.5 experienced a decrease in the loudness of tinnitus with cochlear implantation. However, in these individuals, the severity of tinnitus, as measured based on the Tinnitus Disorders Questionnaire, was low (THI: 8), making cochlear implantation unlikely as a treatment option.

[0112] Based on these findings, fNIRS signals recorded before cochlear implantation can provide a suitable prognostic measure for determining whether the cochlear implant is effective in suppressing tinnitus.

[0113] Those skilled in the art will understand that many changes and / or modifications can be made to the embodiments described above without departing from the broad general scope of this disclosure. Therefore, these embodiments are considered illustrative rather than restrictive in all respects.

Claims

1. A method for characterizing tinnitus in a subject using functional near-infrared spectroscopy (fNIRS), The process of receiving data by a processor device, which includes evoked response data including fNIRS signals that show cortical activity in one or more regions of the subject's brain caused by individual stimuli supplied to the subject, The process of processing the received data using the aforementioned processor device, This includes inputting one or more feature quantities, each containing one or more features of the received data, into the model. The process includes configuring the model to provide one or more classification results that show at least one feature of the subject's tinnitus based on the one or more feature quantities, The one or more feature quantities include one or more features of the evoked response data, The method wherein the individual stimuli include auditory stimuli and visual stimuli.

2. The method according to claim 1, wherein the classification result includes one or more of the following: the presence or absence of tinnitus in the subject, an assessment of the severity of the tinnitus in the subject, quantification of the loudness of the tinnitus sound, and quantification of the annoyance caused by the tinnitus.

3. The method according to claim 1 or 2, wherein the model includes a trained model, the model being trained using an artificial intelligence (AI) algorithm based on one or more prior features mapped to subjective measures of tinnitus characteristics.

4. The method according to claim 3, wherein the model provides classification results using a classification algorithm selected from the group consisting of Naive Bayes, K-Nearest Neighbors (KNN), Rule Induction, Artificial Neural Networks (ANN), and Multilevel Hierarchical Classification.

5. The method according to any one of claims 1 to 4, further comprising the step of detecting a change in at least one characteristic of tinnitus by processing data received by the processor device.

6. The method according to any one of claims 1 to 5, wherein the model is configured to provide a prognostic scale indicating whether a proposed treatment for treating tinnitus in the subject is effective.

7. The method according to any one of claims 1 to 6, comprising determining the quality of each fNIRS signal and removing signals of insufficient quality before processing the received data.

8. The method according to claim 7, wherein the quality of each fNIRS signal is determined based on one or more of the signal gain level and the level of the cardiac signal component.

9. The fNIRS signal is the oxyhemoglobin (O) in the subject's brain. ２ The method according to any one of claims 1 to 8, comprising a signal indicating a change in Hb concentration and / or a signal indicating a change in deoxyhemoglobin (HHb) concentration in the brain of the subject.

10. The received data is, The method according to any one of claims 1 to 9, comprising resting data including fNIRS signals indicating cortical activity in two or more regions of the subject's brain when the subject is in a resting state, wherein processing of the data further comprises determining at least one resting functional coupling measure between at least two regions of the subject's brain based on the resting data, and the one or more features include one or more features of the at least one resting functional coupling measure.

11. The method according to claim 10, wherein at least one stimulus comprises a plurality of individual stimuli that are sequentially supplied to the subject.

12. A non-temporary, machine-readable storage medium comprising instructions configured to cause a processor device to perform the method according to any one of claims 1 to 11.

13. A system for characterizing a subject's tinnitus using functional near-infrared spectroscopy (fNIRS), A step of receiving data including evoked response data, which includes fNIRS signals indicating cortical activity in one or more regions of the subject's brain caused by individual stimuli supplied to the subject, The process of processing the received data, The model includes inputting one or more feature quantities that include one or more features of the received data, and one or more feature quantities that include one or more features of the evoked response data, The processor device is configured to perform the steps of: providing one or more classification results that indicate at least one feature of the subject's tinnitus based on one or more feature quantities; The aforementioned individual stimuli include auditory and visual stimuli, forming a system.

14. The received data is, The system according to claim 13, comprising resting data including fNIRS signals indicating cortical activity in two or more regions of the subject's brain when the subject is in a resting state, wherein processing of the data further comprises determining at least one resting functional coupling measure between at least two regions of the subject's brain based on the resting data, and the one or more features comprising one or more features of the at least one resting functional coupling measure.

15. The system according to claim 13 or 14, further comprising an fNIRS system configured to measure cortical activity levels in at least two regions of the subject's brain.

16. The system according to claim 15, wherein the fNIRS system includes a multi-channel fNIRS system in which each channel is defined by a light source-detector pair.

17. The system according to any one of claims 14 to 16, further comprising an auditory stimulator configured to supply the auditory stimuli to the subject and / or a visual stimulator configured to supply the visual stimuli to the subject.

18. The system according to claim 16, comprising one or more channels configured to be positioned over the frontal, left temporal, right temporal, and occipital regions of the subject's brain.

Images