Method and device for analyzing intracranial dynamics

Abstract

A method for measuring intracranial dynamics by performing optoelectronic measurements of intracranial neuronal and fluid dynamics, brain tissue pulsation, and glymphatic activity through the skull with an optoelectronic measurement arrangement, comprising: applying, by a data processing unit, at least one of the following analyses to electrical signals received from the optoelectronic measurement arrangement and conveying information about said dynamics: at least one of the dynamic pulses is decomposed for processing using a relationship between intrinsic pulses; moments of a statistical analysis of said dynamic pulses are determined; cerebral fluid, particularly water-hemodynamic coupling, is determined from said dynamic pulses based on the definition that the sum of the volumes of brain tissue, cerebrospinal fluid, and intracranial blood is constant; a ratio between the power spectral density of a portion of the entire measured frequency band and the power spectral density of the entire measured frequency band is determined; and an entropy associated with the electrical signal is determined. A G-index is determined based on the above analysis of the electrical signal and a standard based on an analysis of a control group with known and / or estimated intracranial dynamics, the G-index representing the relative brain dynamics.

Description

[Technical Field]

[0001] The present invention relates to an analysis method and apparatus for intracranial dynamics. [Background technology]

[0002] The glymphatic brain clearance system is responsible for convection of both central nervous system metabolites and waste products in both the brain and spinal cord. The glymphatic system maintains brain homeostasis, particularly during nighttime sleep, as it balances metabolite concentrations and removes waste products that accumulate throughout daily function. Deviations from its proper activity and normal function are known to be associated with both brain health and disease states, respectively. Significant changes in the glymphatic system can lead to major cognitive changes.

[0003] Decreased glymphatic activity has been linked to neurodegeneration, epilepsy, narcolepsy, invasive brain tumors, ischemic stroke, and several other major brain diseases. Furthermore, mood, stress, and sleep disorders can affect the function of the glymphatic system, implicating the glymphatic system in brain health.

[0004] However, there are no wearable methods to monitor and analyze glymphatic activity, and therefore measurement and analysis are in need of improvement. Summary of the Invention [Problem to be solved by the invention]

[0005] The present invention seeks to provide improvements in measurement. [Means for solving the problem]

[0006] The invention is defined by the independent claims. Embodiments are defined in the dependent claims.

[0007] In the event that one or more of the embodiments are considered to fall outside the scope of the independent claims, such one or more embodiments may still be useful for understanding the features of the present invention.

[0008] Exemplary embodiments of the present invention will now be described, by way of example only, with reference to the accompanying drawings, in which: [Brief explanation of the drawings]

[0009] [Figure 1] FIG. 1 illustrates an example of a device for measuring intracranial dynamics related to neuronal and fluid dynamic states. [Figure 2] FIG. 1 illustrates an example of an optical measurement arrangement. [Figure 3] FIG. 10 shows examples of possible fit lines for the G-index as a function of time (age). [Figure 4] FIG. 4 shows an example of G-index as a function of age and / or neurodegeneration based on analysis derived from the fit lines of FIG. 3. [Figure 5] FIG. 10 shows an example of a possible fit line for the G-index as a function of time (days). [Figure 6] FIG. 1 shows examples of Gaussian decomposition of brain pulsations for different brains. [Figure 7] FIG. 1 shows examples of Gaussian decomposition of brain pulsations for different brains. [Figure 8] FIG. 1 shows an example of pulse decomposition analysis. [Figure 9] FIG. 10 shows an example of a flow diagram of an analysis based on inter-pulse distance. [Figure 10A] FIG. 10 is a diagram illustrating an example of fluctuations around the average. [Figure 10B] FIG. 10 is a diagram illustrating an example of fluctuations around the average. [Figure 10C] FIG. 10 is a diagram illustrating an example of fluctuations around the average. [Figure 10D] FIG. 10 is a diagram illustrating an example of fluctuations around the average. [Figure 11A] FIG. 1 shows an example of water-hemodynamic coupling. [Figure 11B] FIG. 1 shows an example of water-hemodynamic coupling. [Figure 12] FIG. 1 shows an example of a flow diagram for the analysis of water-hemodynamic coupling. [Figure 13] FIG. 10 illustrates an example flow diagram of fractional amplitude of physiological variability analysis. [Figure 14] FIG. 10 illustrates an example of a flow diagram for sample entropy analysis. [Figure 15] FIG. 10 is a diagram illustrating an example of a flow chart of a spectral entropy analysis. [Figure 16] FIG. 1 shows an example of a flow diagram for cardiopulmonary envelope modulation analysis. [Figure 17] FIG. 1 shows an example of a flow chart of phase transfer entropy analysis. [Figure 18A] FIG. 1 illustrates an example of signal processing for spike rejection and smoothing. [Figure 18B] FIG. 1 illustrates an example of signal processing for spike rejection and smoothing. [Figure 18C] FIG. 1 illustrates an example of signal processing for spike rejection and smoothing. [Figure 18D] FIG. 1 illustrates an example of signal processing for spike rejection and smoothing. [Figure 18E] FIG. 1 illustrates an example of signal processing for spike rejection and smoothing. [Figure 19A] FIG. 1 shows an example of baseline correction and signal normalization. [Figure 19B] FIG. 1 shows an example of baseline correction and signal normalization. [Figure 19C] FIG. 1 shows an example of baseline correction and signal normalization. [Figure 20] FIG. 1 shows an example of a flow diagram for separating individual pulses from a brain signal. [Figure 21] FIG. 2 illustrates an example of a data processing unit. [Figure 22] FIG. 10 is a diagram illustrating an example of a flowchart of an analysis method. DETAILED DESCRIPTION OF THE INVENTION

[0010] The following embodiments are merely examples. Although this specification refers to "an" embodiment in several places, this does not necessarily mean that such references refer to the same embodiment or that a feature applies only to a single embodiment.

[0011] As used herein, the articles "a" and "an" give a general meaning of entity, structure, component, composition, operation, function, connection, etc. Note also that the singular can include the plural.

[0012] Individual features of different embodiments can also be combined to provide other embodiments. Furthermore, the words "comprising" and "including" should be understood not to limit the described embodiments to embodiments consisting only of the recited features; such embodiments may also include features / structures not expressly stated. All combinations of embodiments are considered possible, provided that such combinations do not result in structural or logical contradictions.

[0013] The term "about" means that a quantity or any numerical value is not, and typically does not have to be, exact. This may be because of tolerances, resolution, measurement errors, or rounding, or it may be that the nature of the solutions herein requires that such quantity or numerical value only be approximately such an order of magnitude. Real-world quantities and numerical values ​​always include a certain tolerance.

[0014] It should be noted that the figures show different embodiments, but are schematic diagrams showing only some structural and / or functional entities. The connecting lines shown in the figures may represent logical or physical connections. It is clear to those skilled in the art that the described device may include functions and structures other than those shown in the figures and text. It should be recognized that the details of some functions, structures, and signals used for measurement and / or control are irrelevant to the actual invention. Therefore, it is not necessary to discuss these details in detail in this specification.

[0015] The terms "comprise" (and grammatical variations thereof) and "include" shall be read as "comprise without limitation" and "include without limitation," respectively.

[0016] In this application, the term "determining" in its various grammatical forms can mean calculating, computing, processing data to derive a result, or database consultation, etc. Consequently, "determining" can also mean selecting or choosing, etc.

[0017] The abbreviation HbO refers to oxyhemoglobin measured in the blood. The abbreviation HbR refers to reduced oxygen hemoglobin measured in the blood. The abbreviation HbT refers to total oxygen or total hemoglobin in the blood. Those skilled in the art will be familiar with these terms themselves.

[0018] The term "very low frequency (VLF) band" refers to the frequency band from about 0.008 Hz to about 0.1 Hz.

[0019] The term "respiration band" refers to the frequency band for recording respiration. The respiration band is from about 0.1 Hz to about 0.6 Hz.

[0020] The term "cardiac band" refers to the frequency band for recording heartbeats. The cardiac band is from about 0.6 Hz to about 5 Hz, along with harmonics of the main cardiac frequency. These given frequency ranges refer to the typical physiological frequency range in humans.

[0021] The glymphatic system, which contains cerebral fluid, is responsible for cleansing the brain from metabolic waste products and must function properly to maintain brain health. The activity and state of the glymphatic system are related to brain health and disease states, and therefore to neurodegenerative diseases (NDDs). NDDs are primarily seen in older adults, which means that brain function is affected by NDDs and vice versa. Significant changes in brain glymphatic activity are responsible for cognitive changes. Specifically, the glymphatic activity of the brains of older adults with NDDs should differ from that of older adults without NDDs. Normal aging, without evidence of mild cognitive impairment (MCI) or dementia, also alters cognitive function; these changes generally begin gradually after about age 40, when the brain begins to lose volume, weight, and tissue elasticity. The prefrontal cortex is one of the brain regions affected by aging and is responsible for memory storage. Two other major areas are the posterior association neocortex and the medial temporal lobe. That is, two different age groups should have different glymphatic activity in the brain, which may manifest as different brain health.

[0022] Mood disorders and sleep disorders resulting in stress are also linked to the glymphatic system, i.e. glymphatic activity reflects mental health as well as brain health.

[0023] Glymphatic system activity, which reflects clearance mechanisms, can be monitored using the devices described herein, which use neurofluids or neural fluids (i.e., blood, cerebrospinal fluid, and brain tissue interstitial fluid) and brain mechanical pulsation to assess glymphatic activity and brain clearance. The glymphatic system removes metabolic waste products and balances brain metabolites, which decrease especially during sleep. Inadequate glymphatic convection can reduce the health of mammals, such as humans. In addition, impaired glymphatic convection has been linked to several serious brain disorders, including, for example, Alzheimer's disease, frontotemporal dementia, vascular dementia, normal pressure hydrocephalus, multiple sclerosis, epilepsy, trauma, stroke, tumors, hydrocephalus, Chiari malformation, syringomyelia, pseudotumor cerebri, cerebral vasospasm, glaucoma, and cerebral aneurysms. The dynamics of cerebral and neural fluids within the skull can be measured and monitored using the devices described herein, but determination and / or diagnosis of any potential disease based on the measurements is expressly at the discretion of the medical professional. That is, dynamic measurements may be based on heart rate, respiration, and / or low frequency vasomotor waves (Mayer waves and Traube-Hering waves).

[0024] The optoelectronic measurement system described herein can measure and monitor intracranial brain and cerebral fluid volume dynamics. Fluids within the skull include interstitial fluid, CSF (cerebrospinal fluid), and blood. Furthermore, measurements of the permeability movement of these fluids between blood, perivascular glymphatic CSF, and brain tissue interstitial fluid, involving electrophysiological activity from about 0 Hz to about 1000 Hz, can be performed using full-bandwidth direct current (DC) electroencephalography (EEG). The optoelectronic measurement system and EEG can be used in combination to quantify fluid dynamics. The study of the glymphatic system and neurofluid dynamics is a fairly new field within neurophysiology, and its recordings, processed as described herein, can be used to generate parameters or index values ​​representative of brain function, or more specifically, what can be referred to as cerebral glymphatic activity, brain clearance activity, cerebral fluid dynamics, neurofluid dynamics, or neurofluid dynamics. Overall, the problem can be viewed as a problem of intracranial dynamics, including neuronal and fluid dynamics, brain tissue pulsation, and glymphatic activity related to the central neurological system.

[0025] By using a light source with at least three wavelengths in the near-infrared (NIR) range (e.g., one less than 800 nm, one between 800 nm and 940 nm, and one greater than 940 nm) and at least one photodetector for these wavelengths, brain tissue and / or fluid dynamics can be measured. Furthermore, blood / hemoglobin dynamics can be measured. Brain tissue water or neurofluid dynamics can be interpreted as pulsatile volume, concentration, or fluid flow changes primarily caused by three major physiological pulsatile sources: cardiopulmonary and vasomotor pulsations. Each pulsatile source has a different, non-overlapping frequency range, which can be used to identify the pulsatile source. This identification can further be used to identify the measured brain fluid compartment and, more importantly, to identify failure of the glymphatic clearance mechanism. Additionally, head movement or vibration can be used as a pulsatile source, and these movements can be measured, for example, using an acceleration sensor. Various near-infrared (NIR) light wavelengths result in specific intensity variations or pulse shapes of the detected light that can be used both to distinguish neurofluid compartments and to continuously estimate changes in brain stiffness or elasticity and intracranial pressure, all of which may affect glymphatic activity.

[0026] Additionally, direct current EEG (DC-EEG, electroencephalography) configurations with at least two electrodes can be used to measure potential and time-domain signals between the brain tissue interstitium and blood for direct information on the permeability movement of glymphatic fluid (or glymphatic water) / electrolytes from the glymphatic perivascular space, i.e., beyond the glial interface between blood and brain tissue. The interactions between the collected signals, particularly their specific phase differences and amplitude changes, reflect the activation of the glymphatic system, blood-brain barrier (BBB) ​​permeability, and CSF fluid flow dynamics between CSF and brain tissue. Furthermore, the direction of information flow between these signals can be used to quantify the connection of consciousness to neurohumoral changes.

[0027] The devices herein can be used to measure and analyze signals from the brain and their interactions. The devices herein can be wearable and can be used for long-term monitoring of glymphatic activity and for examining brain health.

[0028] Convection and clearance of glymphatic brain fluid occurs before some chronic brain diseases, such as Alzheimer's disease and chronic traumatic encephalopathy, and in tumors and focal epilepsy, scar formation prevents normal CSF convection. After hemorrhage or ischemic occlusion of an artery, CSF (or brain fluid) can flow into brain tissue, inducing edema and high pressure. The glymphatic systems of mammals, including humans and mice, are similar, as evidenced by magnetic resonance imaging using invasive intrathecal injection of gadolinium (MRI GD3+) followed by contrast imaging lasting 24–48 hours.

[0029] Currently, there is no system for determining glymphatic activity in the brain of a subject that would allow results to be compared across groups of people.

[0030] What is presented herein allows for easy monitoring of sleep quality, can quantify improvements in brain clearance related to glymphatic improvement during sleep, and, more importantly, can monitor subjects in an upright position during daytime activities. Furthermore, it provides a method for studying and diagnosing how physiological exercise, various treatments, etc. affect the glymphatic system and brain clearance, as well as the health status of mammals such as humans.

[0031] The monitoring can be performed within a short measurement period of, for example, up to 10 minutes while awake or overnight while asleep, and can be easily performed using a wearable support structure 102 having at least the sensing device 100. Furthermore, the measurements can be used to diagnose brain diseases such as, for example, Alzheimer's disease, stroke, Parkinson's disease, epilepsy, tumors, which is the job of a medical professional.

[0032] FIG. 1 shows an example of a measuring apparatus including at least one sensing device 100 for measuring intracranial dynamics, i.e., the dynamics of the central nervous system and / or intracranial neuronal and fluid dynamics, as well as glymphatic activity. The apparatus for measuring intracranial dynamics may include a support structure 102 to which at least one sensing device 100 is attached. The support structure 102 may be, for example, but not limited to, a headband or a head cap. The support structure 102 with at least one sensing device 100 is easily worn on the head, making the apparatus truly wearable. The at least one sensing device 100 transmits information about the intracranial dynamics via wired or wireless means to a data processing unit 150, which may be located at a certain distance from the mammal 10, such as a human, on whose head it is attached.

[0033] The at least one sensing device 100 is an optical measurement arrangement 120, an example of which is shown in FIG. 2. The optical measurement arrangement 120 may include at least one visual radiation source 122 and at least one visual radiation detector 124. Each of the at least one visual radiation source 122 directs visual radiation through the skull toward the brain, and the at least one detector 124 receives visual radiation reflected and / or scattered from the brain. The at least one visual radiation source 122 may be in contact with the skull's epidermis. The at least one visual radiation detector 124 may be in contact with the skull's epidermis. Alternatively, a non-zero distance may be provided between the at least one visual radiation source 122 and the skull's epidermis. Correspondingly, a non-zero distance may be provided between the at least one visual radiation detector 122 and the skull's epidermis.

[0034] When light penetrates tissue, it begins to scatter and attenuate. This phenomenon is also influenced by partially irregular physiological events such as breathing, muscle movement, and blood flow. As a result, the intensity changes in the frequency spectrum of the received light exhibit variations that may be constant and / or nonlinear. Most importantly, these intensity variations also include wavelength-dependent variations that can be utilized to monitor metabolism and brain activity. Oxygenated and deoxygenated hemoglobin may be detected.

[0035] Functional near-infrared spectroscopy (fNIRS) signals recorded from the brain exhibit pulsations related to cardiac activity, similar to PPG (photoplethysmography) signals recorded from the skin or finger, with a steep upslope and a slowly decreasing downslope. Each pulse represents a single cardiac beat, and its shape varies from beat to beat. This pulse is sometimes called a brain pulse because the entire brain beats due to cardiac activity as blood volume constantly changes within the blood vessels, resulting in changes in pulse shape due to blood flow and pressure propagation through the vessels. Importantly, therefore, pulse shape also depends on tissue properties. As the brain shrinks due to loss of volume and weight, the properties of brain tissue also change, affecting the elasticity and pulsation of brain tissue and blood vessels.

[0036] The data processing unit 150 is capable of applying one or more analyses to the electrical signals conveying information about the dynamics of the neurological central nervous system in the form of pulses of brain pulsation received from the optical measurement device 120 .

[0037] The analysis may be based on time-domain analysis and / or frequency-domain analysis. Examples of time-domain analysis include pulse shape analysis (or decomposition analysis), analysis of water-hemodynamic coupling and water-EEG coupling (interconnectivity), and sample entropy. Time-domain analysis, for example, can determine the distance between the hemodynamic pulse, water pulse, and EEG pulse (neural activation) for each cardiac pulse. Herein, intracranial water (HO) or brain water is also referred to as cerebrospinal fluid (CSF). Furthermore, differences between these couplings, particularly interhemispheric correlations, between various measurement locations in the brain can be analyzed. For example, interhemispheric balance in the 0.6-5 Hz cardiac band range for HbO-HO and HbR-HO concentrations measured from prefrontal cortex sites in the resting state provides good separation between stroke patients and controls.

[0038] Furthermore, when hemodynamic-water coupling is cross-correlated within a sliding time window, blood-water dynamics in the infrasonic (0.008-0.1 Hz) band, the respiratory (0.1-0.6 Hz) band, and the cardiovascular pulsation (0.6-5 Hz) band provide good separation between AD and control subjects. This time difference forms a series that can be subjected to statistical measures as a feature to distinguish AD patients from controls with a significance of 0.05 based on a t-test. In particular, the SD and IQR of the series in the cardiac band provide the best separation, characterized by an IQR with a p-value between 0.0002 and 0.0018.

[0039] Frequency domain-based analyses include fractional amplitude of physiological fluctuations (fAPF), high- and low-band power spectral density, spectral entropy, and cardiorespiratory envelope modulation (CREM).

[0040] In addition to this, the following analyses can be performed: tissue oxygen saturation, phase synchrony, phase transfer entropy, and EEG signal analysis.

[0041] The phase transfer entropy forms the dPTE (normalized value) and PTE value. Pairs such as HbO-HbR, HbO-HO, HbR-HO, HbT-HO, and EEG-HO can be evaluated. Furthermore, simultaneous multidirectional analysis between dcEEG and HO and HbO and HbR can provide more accurate information about the causal relationship between BBB water permeability and neuronal interactions.

[0042] Using dPTE (normalized value), both directions of movement give the same p-value.

[0043] Using PTE, two-way moves may have different p-values, which may further change from significant to non-significant and vice versa. Significant results from 10Hz JPEG2025534268000002.jpg32170

[0044] Mathematically, the above can be expressed as: dPTE x1→x2 =nPTE x1→x2 -nPTE x2→x1 In the above, the symbol The non-integer after JPEG2025534268000003.jpg6170 is the p-value, which indicates the probability of error. When the p-value is less than 0.05, the result is significant.

[0045] It should also be noted that fNIRS signal analysis and EEG signal analysis can be performed together to form combined results.

[0046] Additionally, power spectral density (PSD) was calculated using periodograms in selected bands: VLF (0.008-0.1 Hz), respiration (0.1-0.6 Hz), and heart rate (0.6-5 Hz). Then, fAPF features for each band were calculated by dividing the PSD from the corresponding band by the PSD from the entire band. fAPF features were calculated for HbO, HbR, and HbT. Similarly, spectral entropy (SE) was calculated from all concentrations of HbO, HbR, and HbT. SE features from all concentrations in the VLF band provided good separation between AD patients and controls. Furthermore, SE features from the VLF band provided very low p-values, indicating significant differences in these features between patients and controls.

[0047] fNIRS-measured EEG pulses can be analyzed using appropriate statistical methods to distinguish subjects from each other when they have different glymphatic activity. Additionally or alternatively, fNIRS pulses can be decomposed into several intrinsic pulses, each of which, when added together, reconstructs the original pulse. Among the many intrinsic pulse options, a log-normal pulse can be selected. A log-normal pulse consists of three parameters: amplitude, position, and width. A single fNIRS pulse can be decomposed into characteristic log-normal pulses, the number of which can be adjusted. For example, by decomposing a single fNIRS pulse into five characteristic log-normal pulses, 15 parameters are acquired. This pulse decomposition analysis can be used, for example, to classify subjects based on age group, obesity level, hypertension, vascular and brain stiffness. Intrinsic pulses are sometimes referred to as standard pulses.

[0048] The concentration changes of oxygenated and deoxygenated hemoglobin, HbO and HbR, and water can be calculated from the fNIRS signal using the modified Beer-Lambert law (MBLL), which is a known method. Total hemoglobin (HbT) can be formed as the sum of HbO and HbR, which represent the concentration changes under the fNIRS sensor. A G-index can be formed from either the raw fNIRS signal or HbO, HbR, HbT, and water.

[0049] Another example of the at least one sensing device 100 is an electrode of an electroencephalography (EEG) electrode arrangement including a support 102 and at least two electrodes. EEG measurements can be used in addition to optical measurements. The at least one sensing device 100 is in electrical contact with the epidermis of the skull of a mammal 10, e.g., a human. The sensing devices of the electroencephalography electrode arrangement receive and sense direct current (DC) electroencephalogram (EEG) signals from the brain of the mammal 10. While the electroencephalography electrode arrangement itself is familiar to those skilled in the art, the bandwidth used for DC EEG is unusual, ranging from about 0 Hz to about 0.5 Hz.

[0050] In an embodiment, at least one optical radiation source 122 can be connected to a pigtail such that optical radiation is guided from the at least one optical source 122 to the skull epidermis through the optical fiber. In an embodiment, at least one optical radiation detector 124 can be connected to a pigtail such that optical radiation is guided from the skull epidermis to the at least one optical radiation detector 124. The end of the optical fiber can be in contact with the skull epidermis or at a non-zero distance from the epidermis. The optical fiber is not separately illustrated in the figures.

[0051] In an embodiment, all of the sensing devices 100 may be optical radiation detectors. In an embodiment, the sensing devices 100 may include a combination of at least two of different types of sensing devices 100. The combination may include at least one electroencephalography sensing device 100 and at least one optical sensing device 100.

[0052] The data processing unit 150 receives the electrical signals from at least one sensing device 100. The electrical signals can be filtered and their baselines corrected and / or stabilized. Existing filtering and baseline correction / stabilization methods are familiar to those skilled in the art.

[0053] The data processing arrangement 150 then processes these electrical signals to form the G-index that represents the glymphatic activity of the brain referred to earlier in this specification.

[0054] Calculation of G-indicators The G-index is a representation of glymphatic activity of the brain as mentioned earlier in this specification. The G-index is based on one or a combination of various analyses of the brain signals performed by the data processing arrangement 150. The G-index can be calculated from fNIRS measurements alone or from a combination of EEG and fNIRS measurements. These calculations can utilize unimodal and / or multimodal approaches, which include multiple analyses of brain signals, either fNIRS measurements or a combination of fNIRS and EEG measurements.

[0055] One type of multimodal analysis is the coupling of various G-index values ​​from various analyses, because in principle, any two analyses relate to different physiological phenomena. Therefore, it is believed that the relationship between several phenomena may reflect interesting physiological conditions. This type of multimodal approach can provide a better and broader range of G-index values ​​than a single modality that provides a G-index based on a single analytical method. However, a single modality and a single feature or algorithm of the signal can be used as the G-index value.

[0056] Rates of change and trends may vary depending on the subject or condition, and therefore more than one model may be used, with models based on known glymphatic activities. Similarly, different models allow for different methods of analysis.

[0057] Possible approximation lines for the G-index as a function of time are shown in Figure 3. In Figure 3, the time axis is age. The rate of change and trend may vary depending on the subject or condition, and therefore several models have been proposed. The approximation lines can be used to predict the G-index during the G-index model development process. For example, a linear function (-kx) and an index function (e) that model the G-index can be used to find a good match with the measurement results when x is time, such as age, or is related to age. -kx ) can be adjusted. This matching can be performed, for example, using least squares error. The slope is negative as a function of age, especially in adults.

[0058] Both single and multi-modality analytical models can be used. Brain signals can be measured in subjects with brain pathology and in subjects with healthy brains. These experiments can utilize brain signals measured from both Alzheimer's disease (AD) patients and control groups (healthy individuals in the same age range) with the aim of selecting an appropriate index for the G-index. The control group can be used as a reference, but either a group with or without brain pathology can be selected as the reference. A basic guideline is needed to determine which index is a good candidate. The following guideline can be used to evaluate whether the proposed method can predict successful results for further treatment: 1. The methods must be capable of separating patients from age-matched controls. 2. The methods must be able to separate healthy control groups with the same age but different lifestyle habits, e.g., differences in sleep quality, stress, smoking, feeding / diet, etc. 3. The methods must allow good separation between subjects of different ages.

[0059] All proposed methods should be evaluated using a simple t-test at the 0.05 significance level. These methods can be expected to perform well when they result in the three separations mentioned above.

[0060] There are several ways to satisfy this criterion. It can be determined that differences in G-index exist, that they can be observed, and that there are several modalities that can utilize at least one of these methods. In embodiments, two or more indices from various methods can be combined to estimate the G-index. Combining various indices can be done by assigning a certain weight value to the corresponding indices. For example, the G-index can be equal to 100 or any other desired value, representing the best brain health or most desirable glymphatic activity, while zero or any other desired value can be the worst; see Figures 3 and 4 in this regard. The G-index is inversely proportional to age. Furthermore, the G-index of a healthy individual of a certain age should be higher than the G-index of a patient suffering from a pathological condition affecting the brain. Even among healthy individuals, the G-index from various lifestyles may differ. Experiments to develop a G-index model may be based on these assumptions.

[0061] The data processing unit 150 processes the received electrical signals. The electrical signals convey information about the brain, particularly about glymphatic activity, during the pulse. Each pulse can be processed using decomposition, statistical methods, and power distribution. Alternatively or additionally, these pulses can be decomposed into groups of pulses, and a G-index can be formed based on one or more properties of the decomposed pulses. Additionally or alternatively, the processing in the data processing unit 150 can utilize one or more moments of a statistical analysis of the decomposed or non-decomposed pulses. The statistical analysis can be based on the mean, variance, skewness, and kurtosis. The data processing unit 150 can then form a G-index based thereon. The non-decomposed pulses can be so-called unprocessed pulses, or can be filtered to reduce noise, interference, and / or artifacts.

[0062] The mathematical processes of the analysis methods are known per se, but in this case these mathematical processes are applied specifically to electrical signals received from optical measurements of the brain, and these mathematical processes can be applied to EEG measurements.

[0063] Brain pulsations are modeled for different groups of people with different brain functions, i.e., different G-indexes. The different groups can have different age ranges, and within an age group, people can be divided into groups with healthy brains and groups with one or more brain pathologies, such as Alzheimer's and / or mental health problems. These pathologies can be further subdivided based on their severity, etc.

[0064] Modeling Phase 1 To verify which features have a good contribution, it is necessary to perform feature analysis, for example, using principal component analysis (PCA). Feature analysis also serves as a dimensionality reduction to speed up the calculation. Then, a simple machine learning classifier, such as kNN (k-nearest neighbor) or multi-class support vector machine, can be used as the learning algorithm.

[0065] Data collection can be performed as follows: Data will be collected from healthy subjects aged 20 years or older. The measurement time for each session was 10 minutes for the purpose of analysis in the VLF band (0.008-0.1 Hz). · Because the G-index is dynamic but can be assumed to lie within a certain range, measure subjects over several sessions over several different weeks. -Measure the subject in a sitting position at rest. Subjects must abstain from certain medications and foods that may alter the measurements.

[0066] Data Processing: All signals can be passed through standard signal processing procedures, e.g., to reduce artifacts, select bands of interest, etc. · Extract features using promising methods. Generate groups based on age, for example, 20-29, 30-39, 40-49, 50-59, older than 59. Each group can now be assigned a label for classification. · Applying machine learning analysis protocols using learning algorithms such as k-nearest neighbor (kNN), support vector machines (SVM), and neural networks (NN) for classification purposes. Validation such as k-fold cross-validation or nested cross-validation can be used, but the excerpt data are from several subjects from the same week and of different ages. The excerpt assay dataset consists of subjects with unknown measurement data for the trained model.

[0067] Evaluation criteria: Further studies will benefit from model sets that provide high accuracy or similarly high performance along with other appropriate performance measures such as F1 score, sensitivity, specificity, or balanced accuracy. Feature sets suitable for further analysis can be selected according to the models that achieve high performance with the models being developed or by appropriate feature importance analysis.

[0068] During feature extraction, many more or less advantageous features are obtained, among which the appropriate ones are selected to adequately support the goal.

[0069] Phase 2 Based on the G-index model from Phase 1, a regression model can be developed in Phase 2. The goal of this phase is not to classify subjects into corresponding age groups, but to achieve a true value for each subject. Model development begins with the simplest regression, i.e., linear regression. Other types of regression, such as logistic regression and logarithmic regression, can be evaluated.

[0070] Data collection: No data collection occurs in this phase, as all necessary signals have been collected during Phase 1.

[0071] Data processing can proceed, for example, as follows. The feature set selected in Phase 1 is used for this phase. Assign the following "expected" G-index values ​​for each group. It is considered a good idea to have a high G-index for young subjects. The correct values ​​for each group are not yet known, so temporarily use the "expected" values. These expected scores are derived based on linear approximation. o20~29 years old: 70 o30~39 years old:60 o40~49 years old:50 50-59 years old: 40 o Older than 59 years old: 30 Apply machine learning algorithms that yield true values ​​instead of classes or labels, for example Bayesian networks, support vector regression (linear, logarithmic, logistic) etc. Models can be validated based on data consisting of a selection of subjects of different ages from a given week. The selection assay dataset consists of subjects whose measurements are unknown to the training model. Furthermore, the "predicted" G-index can be applied to other approximation lines (see Figure 3) and other types of regression algorithms. Parameters on the approximation lines can be changed as needed to achieve reliable and robust results.

[0072] Evaluation criteria: The results should be similar to those described in Figure 4. Mean squared error (MSE) or other suitable performance measures can be used to test for good performance. The variation between different measurements on the same subject should be within, for example, the mean ± two standard deviations.

[0073] Expected result: A rough idea of ​​the approximation line for the G-index curve and parameters.

[0074] Phase 3 Children and teenagers can be included in Phase 3. The process is the same as for the G-index model in Phase 2, and can be carried out, for example, as follows:

[0075] Data collection: Subjects Teens: 13-19 years old Children: 7-12 years old Toddlers: 2-6 years old · Data collection is performed using the same protocol.

[0076] Data Processing: · The signals from these subjects are combined with signals from previous data collection in Phase 1. Use the same procedure as in Phase 2. Assign the following "expected" G-indicators from the linear approximation: o2~6 years old: 100 7-12 years old: 90 o13~19 years old: 80 Applying the "forecast" G-index using other approximation lines, see Figure 3.

[0077] Evaluation criteria: See Phase 2.

[0078] Expected result: A rough idea of ​​the approximation lines for the G-index curves and parameters after adding data from children and teenagers.

[0079] Phase 4 Phase 4 involves patients with certain NDDs, and the process follows Phases 1 and 2. This phase requires more data collection from patients undergoing standard NDD assessments. At this point, the severity of the NDD and age are ignored; that is, as long as the patient has a low G-index, it does not matter. The G-index from healthy subjects should perform similarly to that in Phase 3. Phase 4 can be performed, for example, as follows:

[0080] Data collection: Collect data from patients with NDD. Examining patients to determine how serious their health condition is. · Perform data collection using the same protocol.

[0081] Data Processing: Data from patients is used in two scenarios: a scenario where all patients are treated as one entity (patients with NDDs) and a scenario where data is treated based on NDDs. The expected G-index for this group is either 10 or 20, based on a linear approximation. These options will be explored in this phase. Combine data from patients with data from Phase 2 and Phase 3 studies. Use the same procedure as in Phase 2.

[0082] Evaluation criteria: See Phase 2.

[0083] Expected result: G-index curve and parameter fit lines after adding patient data

[0084] Phase 5 In Phase 5, the severity of NDD is included. Patients are classified based on the severity of NDD. In this experiment, the patient's age is ignored. These data are combined with the data used in Phase 4 to construct the next model. These steps follow Phase 4. Phase 5 can be performed, for example, as follows:

[0085] Data collection: Data collection is not performed in this phase as this has been done using the previous phase.

[0086] Data Processing: · Data from patients will be divided into three groups due to varying levels of severity. The "expected" G-index values ​​are 10, 5, and 0 based on a linear approximation line. Repeat the procedure in Phase 1 for each of the NDDs, but only for the data from the NDD.

[0087] Evaluation criteria: See Phase 1. If good performance can be predicted, the steps in Phase 2 can be implemented.

[0088] Expected result: Fitted lines for G-index curves and parameters with gradual changes indicating the severity level of NDD.

[0089] In general, the development of a G-index model involves the following algorithms: ·Classification kNN using various distance measures, e.g., Euclidean, Mahalanobis, Chebyshev, or city blocks. Supervised learning algorithms such as Support Vector Machines (SVMs with various kernels, e.g., linear, polynomial, radial basis function, and sigmoid). Decision trees and random forests Regression o Regression Support Vector Regression Bayesian Networks

[0090] One or more of these models can be used by the data processing unit 150 to represent the glymphatic activity of a person's brain or brain health as a reference for forming a G-index as a function of age or any other parameter. This reference can be, for example, the G-index of a normal, healthy brain. Alternatively, the reference can be the G-index of a group of people with a brain pathology or deviation from normal brain function caused by, for example, lifestyle, medication, illness, obesity, stress, mental health, pregnancy, injury, nutrition, etc.

[0091] Measurement signal processing In embodiments, the dynamics of the neurological central nervous system can be measured in a frequency range of about 0 Hz to about 0.5 Hz. In embodiments, the dynamics can be measured in a frequency range of about 0 Hz to about 2.5 Hz. In embodiments, the dynamics can be measured in a frequency range of about 0 Hz to about 4 Hz. In embodiments, the dynamics can be measured in a frequency range of about 0 Hz to about 5 Hz. The dynamics of the glymphatic system refers to the pulsatile activity related to its clearance function.

[0092] Decomposition Examples of analysis based on decomposition are illustrated in Figures 6, 7, and 9. The waveform of the received pulse can be decomposed based on an integral transform. In an embodiment, the integral transform can be a Fourier transform or a Laplace transform, which are familiar to those skilled in the art. The data processing unit 150 can decompose at least one of the pulses of the dynamics of the neurological central nervous system into multiple intrinsic pulses to process at least one relationship between them, for example, the temporal distance between them.

[0093] In general, pulses can be transformed into pulses that look like Gaussian pulses. Without limitation, logarithmic transformations can be performed using natural logarithms. Since negative values ​​are not allowed (logarithms have no real value for negative numbers), scaling of variables may be necessary. Scaling can be performed, for example, together with filtering, if necessary.

[0094] A transform, one example of which may be a Gaussian decomposition, can be used to extract, for example, three to five characteristic individual lognormal pulses or waves. The number of unique pulses can be made adjustable for the measurement. As shown in Figures 6 and 7, there are five characteristic lognormal pulses. The Gaussian decomposition can be mathematically expressed as:

number

[0095] An alternative to the Gaussian transform itself is available: a reciprocal transform can be used to make the pulse look like a Gaussian pulse.

[0096] An index transform can be used to make a pulse look like a Gaussian pulse. Common indices are the square (square transform), the cube (cube transform), and the corresponding roots.

[0097] Yet another transform is the Box-Cox transform, which can also be used to make pulses look like Gaussian pulses. A modification of the Box-Cox transform is the Yaw-Johnson transform, which can also be used to make pulses look like Gaussian pulses. These examples show that measured glymphatic pulsations of the brain can be decomposed in various ways to form a G-index.

[0098] Pulse decomposition analysis (PDA) can be applied to a single pulse measured from the brain to extract several unique pulses that can be used to reconstruct the original pulse. The use of PDA in fNIRS (functional near-infrared spectroscopy) signal analysis requires a sampling frequency sufficient to achieve good temporal resolution with respect to parameters such as center position and pulse width. The sampling frequency should be higher than, for example, about 100 Hz. For example, a sampling frequency of about 500 Hz or higher can be used.

[0099] To obtain five log-normal pulses for each pulse together with the 15 parameters of the Gaussian transformation, which is an example here, the PDA can be applied using the following algorithm: 1. The algorithm guesses 15 random numbers as starting points for the optimization algorithm (different starting points may result in different estimated parameters). 2. Based on the starting point, a constrained minimization or optimization function finds the appropriate parameters subject to the constraints. 3. The estimated parameters are evaluated based on the error and correlation with the original pulse. 4. If the evaluation criteria are not met, the entire process is repeated.

[0100] The Gaussian transform shown in Figures 6 and 7 can be used to obtain five pulses that, when combined, resemble the original pulse measured from the brain. The first and third of these five unique pulses represent the systolic and diastolic pulse pressures, while the second and fourth pulses are reflections of these pulse pressures due to bifurcation. The distance between the first and second pulses can be advantageous for estimating the stiffness of the brain, since the reflected pulse from the stiffened artery arrives earlier than that from the elastic artery. This assessment based on optical measurements of the brain can be performed, for example, on segmented signals to achieve results periodically, e.g., over the last approximately two minutes.

[0101] Because some individual pulses may be useless for pulse resolution, pulse selection can be applied to remove unqualified pulses. Based on pulse rate variability (such as HRV in an ECG), pulses outside the mean ± 3x standard deviation boundary can be discarded. Additional tasks, such as correlating pulses with reference pulses, can be used to collect reference pulses that may differ between age groups. Because neurodegenerative diseases (NDDs) and other issues affect tissue stiffness, the distance between unique pulses can be used to quantify the severity of NDDs, for example. Instead of NDDs, some other reasons that are not necessarily pathological, such as age, nutrition, and / or lifestyle habits, may affect tissue stiffness.

[0102] 6 and 7, the decomposed first, second, third, fourth, and fifth pulses have different heights and widths. The fourth and fifth pulses are lower in the case of a rigid brain than in the case of an elastic brain, for example, at the trailing edge. Because there are various possibilities for how the waveform relates to brain stiffness, in embodiments, the recognition of brain stiffness from the waveform can be based on, for example, artificial intelligence, neural networks, and / or machine learning methods. Alternatively or additionally, brain stiffness can be determined based on image processing of the waveform curve.

[0103] Use of statistical moments deviation The fluctuations around the mean (or center) of blood oxygen level-dependent (BOLD) signals, i.e., the standard deviation (SD), can be used to separate patients with brain pathologies such as Alzheimer's disease (AD) from controls with healthy brains. Figures 8A through 8D show examples of fluctuations around the mean. The SDs of people with brain pathologies (AD patients) are consistently larger than those of healthy controls in the very low frequency (VLF, 0.008-0.1 Hz), respiration (0.1-0.6 Hz), and heart rate (0.6-5 Hz) bands. A similar procedure can be applied to oxygenated hemoglobin (HbO) (see Figure 10A), deoxygenated hemoglobin (HbR) (see Figure 10B), and water (H2O) (see Figure 10C) measurements. This method uses oxygenated hemoglobin (HbO) and total hemoglobin (HbT) at various bands to separate patients with brain pathologies such as AD from controls with healthy brains. Figure 10D provides an example in which it is possible to use a light source with only a single wavelength (830 nm in this example) to yield successful analyses. All measurements in Figures 10A through 10D clearly show that AD patients have greater variation from baseline to baseline than controls without AD.

[0104] The variation around the mean can be found as follows: 1. Use MBLL to estimate changes in HbO, HbR, and water concentrations, or the voltage V signal in dcEEG. 2. Apply various bandpass filters to select the VLF (0.008-0.1 Hz) band, the respiration (0.1-0.6 Hz) band, and the heart rate (0.6-5 Hz) band. 3. Calculate the standard deviation (SD) or variance from each concentration. 4. Normalize the SD of the coefficient of variation (CV) = SD / mean.

[0105] water-hemodynamic coupling Figures 11A, 11B, and 12 show an example of the analysis of water-hemodynamic coupling. Based on the Monroe-Kelly law, the sum of the brain, CSF, and intracranial blood volumes is constant. That is, there is a relationship between hemodynamics and water within the skull.

[0106] It is hypothesized that if AD, other brain pathologies, or non-pathological conditions of the brain result in problems with CSF and hemodynamic coupling, correlation analysis will shed light on these problems.

[0107] Since blood also contains water, CSF or free water is calculated by subtracting approximately 80% of HbT from the water concentration. By examining the derivative -d(BOLD) / dt and CSF, they are anti-correlated at certain time differences. This analysis also includes first differences from both sides in the VLF band, respiratory band, and cardiac band.

[0108] Water-hemodynamic coupling can be determined in the following manner: 1. HbO, HbR, HbT, and water (H2O), as well as dHbO, dHbR, dHbT, and dH2O can be used. 2. The bands of interest can be the VLF band, the respiratory band, and the cardiac band. 3. The pairs HbO-H2O, HbR-H2O, HbT-H2O, dHbO-H2O, dHbR-H2O, dHbT-H2O, HbO-dH2O, HbR-dH2O, and HbT-dH2O can also be used with dcEEG in full band (VLF, respiration, heart rate, delta (0.2-4 Hz), theta (4-7 Hz), alpha (8-12 Hz), beta (20-40 Hz), gamma (40-70)). 4. Cross-correlation can be applied to each pair to obtain segmented signals using a predetermined window, where the minimum width in each band is the longest period, while the maximum is approximately 1.5 times the minimum. a.VLF=VLF is, for example, 120 seconds, 140 seconds, 160 seconds, 180 seconds. b. Breaths are, for example, 10 seconds, 12 seconds, 14 seconds, 16 seconds. c. Heartbeats are, for example, 2 seconds, 2.4 seconds, 2.6 seconds, and 3 seconds. 5. The windows can be shifted so that they have approximately 25%, 50%, and 75% overlap. 6. The time lag values ​​from each window with the best correlation coefficient can form the time series data for further analysis. 7. This time series can be subjected to statistical analyses such as mean, SD, median, IQR (interquartile range), skewness, and kurtosis.

[0109] In addition, the SD and IQR of the HbR lag variability profiles are consistent with Mini-Mental State Examination (MMSE) scores, indicating that healthy-brain controls have greater variability in the lag profiles than AD patients and those with other brain pathologies.

[0110] Fractional Amplitude of Physiological Fluctuations (fAPF) An example of a flow chart of fAPF is shown in Figure 13. In this method, the ratio of PSD (power spectral density) between the low frequency band and the entire band is compared. The method of calculating the ratio between a part of a band and the entire band is known per se, for example, in the field of functional magnetic resonance imaging, but its use or applicability in optical brain measurements is less known to those skilled in the art. Instead of using the proposed low frequency band, fAPF (fractional amplitude of physiological fluctuations) uses the very low frequency (VLF, approximately 0.008 Hz to approximately 0.1 Hz) band, the respiratory (approximately 0.1 Hz to approximately 0.6 Hz) band, and the cardiac (approximately 0.6 Hz to approximately 5 Hz) band. The VLF band reflects the vasomotor response related to vascular contraction and dilation.

[0111] Mathematically, this reaction can be expressed as follows:

number

[0112] The algorithm for measuring fAPF may be as follows: 1. Calculate the power spectral density (PSD) of HbO, HbR, and HbT from the VLF, respiratory, cardiac, and total bands used in the measurements. 2. Calculate the area under the curve (AUC) of the power spectral density using, for example, trapezoidal integration. 3. Divide the power spectral density of the VLF band, respiratory band, and cardiac band by the power spectral density of the total band to obtain the fAPF from the corresponding band.

[0113] Experiments on fNIRS signals measured from both AD patients and / or patients with various other brain pathologies and control groups with healthy brains can be expected to yield successful results. The fAPF of HbT in the cardiac band can be used to separate patients with brain pathologies from control groups with healthy brains. Again, as elsewhere herein, brain pathologies can refer to brain diseases such as AD, mental health issues, lifestyle effects on the brain, and / or injury.

[0114] Ratio of PSD in the high band to the low band An example flow diagram of the PSD ratio of the high and low bands is shown in Figure 15. Another application of power spectral density relates to the characterization of wideband irregular signals. In the fNIRS domain, signal amplitudes vary significantly between subjects, and comparing (power spectral density) values ​​between subjects can sometimes be inaccurate. Therefore, in an embodiment, the ratio between the PSD of the high frequency band and the PSD of the low frequency band is a suitable solution to normalize this value. This method is inspired by HRV analysis in the frequency domain.

[0115] Mathematically, this reaction can be expressed as follows:

number

[0116] Below are examples of high and low PSDs for VLF. The high and low respiratory and cardiac bands are determined in a corresponding manner.

[0117] The PSD ratio of the high frequency band to the low frequency band can be: 1. Calculate thresholds to separate low and high bands within the VLF, respiratory, and cardiac bands. a. A logarithmic scale can be used instead of a linear scale. b. There are two values ​​for the frequency limits for each band. c. The center point can be calculated as the average of log(lower limit) and log(upper limit). d. Convert the logarithmic scale to a linear scale using a central power of 10. VLF band example i. The lower and upper limits are 0.008 Hz and 0.1 Hz, respectively. ii. Center point = (log(.008)+log(.1)) / 2=-1.5485 iii. On a linear scale, the center point is approximately 0.0283 Hz. iii. Calculate the PSD from each frequency band. 3. Using the center point, each band can be divided into a low frequency band and a high frequency band. 4. To achieve PSD low band and PSD high band, the AUC of the low frequency band and high frequency band respectively can be calculated using, for example, trapezoidal integration. 5. Divide the PSD of the high frequency band by the PSD of the low frequency band to obtain the features.

[0118] Sample Entropy An example flow diagram for sample entropy is shown in Figure 14. Entropy measures the irregularity of time series data without prior knowledge of the data generator source. Therefore, entropy can be used to estimate the regularity of the data. Unfortunately, entropy estimation requires the entire time series, which is not usually obtainable. However, this type of estimate of a partial signal can be formed using an algorithm called approximate entropy, which itself is familiar to those skilled in the art. Furthermore, sample entropy can be used to improve approximate entropy algorithms.

[0119] The sample entropy can be: 1. Sample entropy can be applied to HbO, HbR, and HbT in various bands along with the Chebyshev distance measure. 2. The sample entropy values ​​of different age groups can be compared.

[0120] Experiments on HbO, HbR, and HbT using sample entropy in conjunction with the Chebyshev distance measure can be expected to yield favorable results. Sample entropy scores increase gradually and proportionally with the age of the group, except for HbT at ages younger than 30 years. This may indicate that the older a person is, the more irregular the changes in Hb, HbR, and HbT become.

[0121] Spectral Entropy An example of a flow diagram for spectral entropy is shown in Figure 15. Spectral entropy is a measure of the spectral power distribution of a signal based on Shannon entropy. Therefore, it applies Shannon entropy to the normalized power distribution of the signal. Spectral entropy analysis of MREG, EEG, and fNIRS signals revealed electrophysiological and hemodynamic changes in drug-resistant epilepsy.

[0122] The calculation of spectral entropy begins with the normalized power spectral density (PSD). There are many ways to achieve the PSD, such as FFT (Fast Fourier Transform) and periodogram. Spectral entropy simply applies Shannon entropy to the probability distribution of the normalized PSD. Spectral entropy can be applied to raw fNIRS signals and concentration changes.

[0123] The spectral entropy can be: 1. Use a signal sampled at 10Hz instead of 800Hz. 2. A band pass filter (BPF) can be applied to select the band of interest, i.e. VLF, respiration, or heart rate. 3. Spectral entropy can be calculated using the pentropy function in Matlab. 4. Spectral entropy values ​​can be compared between patients with brain problems and a control group with healthy brains using, for example, a t-test at a significance level of 0.05.

[0124] Again, brain problems can refer to brain diseases such as AD, mental health problems, lifestyle effects on the brain, and / or injuries.

[0125] Cardiorespiratory Envelope Modulation (CREM) An example flow diagram of CREM is shown in Figure 16. Analysis on the blood oxygen level dependent (BOLD) signal shows a respiratory band modulation signal in the cardiac band. The same method can be applied to HbO, HbR, and HbT.

[0126] The algorithm of CREM can be as follows: 1. Apply BPF (band pass filtering) to select respiratory and cardiac bands for analysis. 2. Estimate the upper envelope of the heart rate signal. 3. Apply FFT (Fast Fourier Transform) to the upper envelope of the respiration signal and the heart rate signal. 4. Compute the absolute value of the FFT from step 3 for the respiration signal and the upper envelope of the heart rate signal. 6. Limit results in the upper envelope of the breathing band and / or heart rate signal to about 0.1 Hz to about 0.6 Hz. 7. Calculate the correlation between the absolute value of the breathing band limited to about 0.1 Hz to about 0.6 Hz and the absolute value of the upper envelope of the heart rate signal.

[0127] Phase transfer entropy An example of a phase transfer entropy flow diagram is shown in Figure 17. Phase transfer entropy uses the average of the phase transfer entropy to obtain the differential phase transfer entropy dPTE (normalized) value and the PTE value. Binding is evaluated for the following pairs: HbO-HbR, HbO-H2O, HbR-H2O, and HbT-H2O. 2. Using dPTE (normalized value), both directions of travel give the same p-value. 3. By using PTE, transfers from both directions can have different p-values, which can also change from significant to non-significant and vice versa. 4. Significant results from 10Hz: JPEG2025534268000007.jpg32170

[0128] Mathematically, the phase transfer entropy can be expressed as: dPTE x1→x2 =nPTE x1→x2 -nPTE x2→x1 In the above, the symbol The non-integer at the end of JPEG2025534268000008.jpg6170 is the p-value, which indicates the probability of error. The result is significant when the p-value is less than 0.05.

[0129] The acceleration signal recorded by the accelerometer is sensitive to vibrations, and therefore the raw signal may appear noisy. For this reason, a low-pass filter can be used to remove unwanted spikes due to this property. This signal can then be utilized.

[0130] By placing a 3D accelerometer on the body, especially on the waist, it is possible to estimate body movement. 3D accelerometers can also be placed on the wrist or even on the fingers. For sleep studies where body posture is restricted, the easiest way to estimate body posture is to place the sensor on the chest. However, since the optical measurement arrangement 120 and possible EEG measurements are worn on the head, sometimes on the forehead, placement on the chest is not practical for G-index measurements anyway. For this reason, it is more practical to place the accelerometer on the forehead.

[0131] Alternative body positions during sleep are supine with the head facing towards the centre, supine with the head facing to the left or right, and left or right lateral position, i.e. there are five possible head positions during sleep, which correspond to three body positions.

[0132] By using a 3D accelerometer, three different axes of movement can be evaluated. The axis parallel to the left-right movement of the head movement may be the dominant signal, while the other axes act as secondary axes. Another secondary signal is the single signal intensity area of ​​the 3D acceleration signal. When a person is in a lateral position, the forehead is almost perpendicular to the vertical axis. This situation is slightly different from the supine position, where the head is turned to the left or right. The angular difference should be sufficient to separate the lateral position from the supine position. To make the prediction more reliable, machine learning algorithms can be used. Features can be extracted based on the body posture and its transitions.

[0133] As the body posture during sleep is tracked, the signals are processed based on a predetermined window with a certain overlap, for example, about 25%. Because body movements and their transition times vary among different people, the window width should be carefully determined. The width should not be too narrow so that the acceleration signal does not capture the movement, nor should it be too wide so that several significant movements exist within the same window. As a result, the data processing unit 150 can also determine the G-index based on one or more postures of the subject. A reference for the measurement value can be obtained by measuring a control group in one or more predetermined postures.

[0134] The raw signal set from the glymphatic measurement device may include, for example, four fNIRS signals from various wavelengths, EEG, acceleration signals, and other possible modalities. Due to hardware complexity and other reasons, these signals may contain unwanted spikes that must be filtered out or otherwise affect the concentration calculations. These spikes may be either rising or falling spikes.

[0135] 18A to 20 show examples of how signals are preprocessed before pulse analysis. FIG. 18A shows an example of a raw signal for G-index measurement. FIG. 18B shows an example of a preprocessed fNIRS signal in which unwanted spikes in the raw signal are replaced with vertical lines using their derivatives. The slope of the spikes can be either positive or negative.

[0136] Sometimes, there is another unwanted spike within an existing unwanted spike. This phenomenon results in more than two perpendicular lines. This can lead to the problem of identifying a single point to mark the location of the unwanted spike. For this reason, these perpendicular lines can be combined into a single pulse, e.g., a square pulse, convolution with N=20, followed by applying, e.g., an MA filter (N=25). This results in a neat Gaussian-shaped pulse at the location of the unwanted spike. In this case, the Gaussian-shaped pulse can represent the location of the unwanted spike.

[0137] Next, to make the signal curve continuous, the vertical gaps are replaced with a straight line between their endpoints. That is, each unwanted spike is replaced with a linear line based on the linear equation between the two points. Finally, the signal can be smoothed with a moving average filter.

[0138] 19A to 19C show an example of normalizing a baseline-linearized signal. To smooth the raw signal, i.e., to remove unwanted frequencies outside the cardiac band, a bandpass filter, e.g., from about 0.5 Hz to about 3 Hz, can be applied. The upper and lower envelopes of the raw signal are then used to estimate baseline fluctuations. To eliminate or reduce baseline fluctuations, the estimated baseline fluctuation based on the lower envelope is subtracted from the raw signal. The upper envelope can be estimated from a signal without baseline fluctuations. The signal without baseline fluctuations can be divided by the upper envelope and normalized to the range [0, 1], for example. Each pulse of the normalized signal can be isolated by slicing it at zero. The steps of FIGS. 19A to 19C are also illustrated as a flow chart in FIG. 20.

[0139] Figure 21 shows an example of the data processing unit 150 shown in Figure 1. The data processing unit 150 includes one or more processors 2100 and one or more memories 2102 containing computer program code.

[0140] The term "computer" includes a computing device that performs logical and numerical operations. Data processing unit 150 is a computer. For example, a "computer" may include an electronic computing device such as an integrated circuit, a microprocessor, a mobile computing device, a laptop computer, a tablet computer, a personal computer, or a mainframe computer. A "computer" may include a central processing unit, an ALU (arithmetic logic unit), a memory unit, and a control unit that controls the operation of the other components of the computer so that the steps of a computer program are executed in a desired order. Furthermore, a "computer" may include auxiliary memory (such as a disk drive or flash memory) and / or include data processing circuitry.

[0141] 1 refers to input / output devices and / or units. Non-limiting examples of user interfaces include a touch screen, other electronic display screen, a keyboard, a mouse, a microphone, a handheld electronic game controller, a digital stylus, a display screen, a speaker, and / or a projector for projecting visual displays.

[0142] The one or more memories 2102 and computer program code are configured to cause the one or more processors 2100 to perform at least the method steps of the analysis flow chart shown in FIG.

[0143] 22 is a flow diagram of the analysis method. In step 2200, optical measurements of intracranial dynamics, including intracranial neuronal and fluid dynamics, brain tissue pulsation, and / or glymphatic activity, are made through the skull by the optical measurement arrangement 120.

[0144] In step 2202, at least one of the following analyses is applied by the data processing unit 150 to the electrical signals received from the optical measurement arrangement 120 and transmitting information about said intracranial dynamics through the skull in the form of pulses: Decomposing 2202A at least one of the pulses of the intracranial dynamics into a plurality of intrinsic pulses for processing using at least one relationship between the intrinsic pulses; At least one moment of the statistical analysis of the intracranial dynamics pulse is determined in step 2202B. In step 2202C, hydro-hemodynamic coupling is determined from the intracranial dynamic pulse based on the definition that the sum of brain volume, cerebrospinal fluid volume, and intracranial blood volume is constant. A ratio between the power spectral density of the portion of the entire measurement frequency band and the power spectral density of the entire measurement frequency band is determined in step 2202D. An entropy associated with the electrical signal is estimated in step 2202E.

[0145] In step 2204, a G-index is determined that represents the relative intracranial dynamics of the brain relative to the baseline based on the above analysis of the electrical signals representative of the state of brain function and a baseline based on a corresponding analysis of at least one control group with known and / or estimated intracranial dynamics.

[0146] The method shown in Figure 22 can be implemented as a logic circuit solution or a computer program. The computer program can be placed on a computer program distribution means for distribution thereof. The computer program distribution means is readable by a data processing device that encodes the computer program instructions, performs measurements and analysis, and optionally controls processes based on the measurements.

[0147] The computer program may be distributed using a distribution medium, which may be any medium readable by a controller. The medium may be a program storage medium, a memory, a software distribution package, or a compressed software package. In some cases, the distribution may be performed using at least one of a short-range communication signal, a short-range signal, and a long-range communication signal.

[0148] It will be obvious to those skilled in the art that as technology advances, the concept of the present invention can be implemented in various ways. The present invention and its embodiments are not limited to the exemplary embodiments described above, but rather may vary within the scope of the claims. [Explanation of symbols]

[0149] 2100 processor 2102 memory

Claims

1. 1. A method for analyzing intracranial dynamics, comprising: performing (2200) optoelectronic measurements of intracranial dynamics, including intracranial neuronal and fluid dynamics, brain tissue pulsation, and / or glymphatic activity, through the skull by an optoelectronic measurement arrangement (120); The electrical signals received from the optoelectronic measurement arrangement (120) and conveying information about the intracranial dynamics in the form of pulses are analyzed as follows: Decomposing at least one of the pulses of the intracranial dynamics into a plurality of intrinsic pulses for processing with at least one relationship between the intrinsic pulses (2202A); Determining at least one moment of the statistical analysis of the pulse of the intracranial dynamics (2202B); determining cerebral fluid, particularly water-hemodynamic coupling, from said pulse of intracranial dynamics based on the definition that the sum of the volumes of brain tissue, cerebrospinal fluid (CSF), and intracranial blood is constant (2202C); Determining (2202D) a ratio between a power spectral density of a portion of an entire measurement frequency band and a power spectral density of the entire measurement frequency band; estimating an entropy associated with the electrical signal (2202E); applying (2202) by the data processing unit (150) at least one of determining 2204 a G-index representative of the relative dynamics of the brain based on said analysis of said electrical signals and a criterion based on an analysis of at least one control group with known and / or estimated intracranial dynamics; A method characterized by:

2. directing optical radiation through the skull towards the brain with the optical measurement arrangement (120); receiving, by the optical measurement arrangement (120), the optic radiation that has interacted with the brain; converting the optical signal into an electrical signal by the optical measurement arrangement (120); detecting, by a data processing unit (150), pulses of said intracranial dynamics; 2. The method of claim 1, characterized by:

3. 2. The method of claim 1, characterized by: performing the decomposition of at least one of the pulses of the intracranial dynamics into the plurality of intrinsic pulses by converting the at least one pulse into a predetermined number of intrinsic pulses by Gaussian decomposition, wherein the predetermined number is adjustable; and forming the G-index based on at least one temporal distance between the intrinsic pulses.

4. 2. The method of claim 1, characterized in that the determination of the at least one moment of the statistical analysis of the pulse of the intracranial dynamics is performed as a variation of the moment relative to a mean of the moment.

5. 2. The method of claim 1, wherein the determination of the water-hemodynamic coupling from the pulse of intracranial dynamics based on the definition that the sum of the volumes of the brain, cerebrospinal fluid, and intracranial blood is constant is performed by determining oxygenated hemoglobin (HbO), deoxygenated hemoglobin (HbR), and total hemoglobin (HbT), and their differences.

6. determining a fractional amplitude of the physiological variation based on a ratio of at least one of a very low frequency (VLF) band from 0.008 Hz to 0.1 Hz to the total band, a respiratory band, and a cardiac band covering at least 0.008 Hz to the upper end of the cardiac band used for the measurement to the total band; 2. The method of claim 1, wherein the determination of the ratio between the power spectral density of the portion of the entire measurement frequency band and the power spectral density of the entire measurement frequency band is performed by:

7. 2. The method of claim 1, characterized in that the estimation of the entropy is performed in the time domain and / or in the frequency domain of the electrical signal.

8. Measuring the electroencephalogram signals of the brain and performing the following analyses by the data processing unit (150): estimating entropy associated with said electroencephalogram signals; determining a ratio between a power spectral density of a portion of an entire measured frequency band of the electroencephalogram signal and a power spectral density of the entire measured frequency band; and combining information from said electroencephalogram and from the corresponding analysis of said optical measurements to form said G-index; 2. The method of claim 1, characterized by applying at least one of the following to the electroencephalogram signal:

9. An intracranial dynamics analyzer, comprising: one or more processors (2100); and one or more memories (2102) containing computer program code; Including, The one or more memories (2102) and the computer program code are configured to cause an apparatus, using the one or more processors (2100), to perform at least the method steps of claim 1. An apparatus characterized in that

Images