Apparatus for generating a score indicative of a cognitive disorder such as attention deficit hyperactivity disorder - Patent Application 20070122997

Abstract

The present disclosure relates to generating a score indicative of a subject having a cognitive disorder, such as a cognitive impairment, such as attention deficit hyperactivity disorder (ADHD), where the score can be generated based on electroencephalography or magnetoencephalography (EEG / MEG) data by calculating one or more correlation values ​​between pairs of time series determined based on EEG / MEG data. The present disclosure includes receiving a subject's EEG / MEG data and determining a first plurality of time series based on the EEG / MEG data, each of the first plurality of time series corresponding to a respective source location located within the subject's cranial cavity. Next, a first correlation value is calculated for the first pair of time series, and the first pair of time series is included in the determined first plurality of time series. A score is generated based on the first correlation values, and the score is output to a physician who can diagnose the subject and indicate that the subject has a cognitive disorder, such as a cognitive impairment, such as ADHD.

Description

[Background technology]

[0001] Bayesian brain theory establishes that the entire cortex represents a probability distribution, which collapses into an estimate only when a decision is required. Nevertheless, how this transition from a probabilistic brain to a belief or one-off decision occurs is not fully understood. Therefore, deriving a metric that reconciles stimulus-limited surprise (Shannon surprise) and allows for the calculation of the information gain (Bayes surprise) between prior and posterior beliefs remains an open question. Expressing the probability of an event in terms of its information content (e.g., the -log function of the probability) (also known as Shannon surprise), this can mean moving from a probabilistic environment to a deterministic environment, where one is actually surprised by the occurrence of the event and believes that the event has already occurred. Therefore, the generated information bits can provide precision, i.e., counteract uncertainty about the event's occurrence. Therefore, it may be necessary to distinguish between the precision bits of Shannon information and the Shannon entropy bits, which represent uncertainty.

[0002] In attention research, it is hypothesized that a salience map generated by the sensory salience of an object is modulated by top-down action recognition signals to generate a priority map. While much work has been done on neural encoding (e.g., through computational modeling of attention), adequate processing of top-down attentional signals has not been achieved. Instead, much of the prior art modeling work has focused on the bottom-up component of visual attention.

[0003] Furthermore, due to the complexity and heterogeneity of cognitive disorders, such as cognitive disorders like attention deficit hyperactivity disorder (ADHD), the gap between the neurobiology and symptoms of such disorders is traditionally often large, and therefore the underlying neurocognitive mechanisms are often unknown.

[0004] It would therefore be advantageous to provide an improved system, corresponding computer-readable storage medium, and corresponding computer-implemented method for generating a score that indicates that a subject has a cognitive disorder, such as a cognitive disorder like attention deficit hyperactivity disorder (ADHD). Summary of the Invention

[0005] The present disclosure relates to generating a score indicative of a subject having a cognitive disorder, such as a cognitive impairment, such as attention deficit hyperactivity disorder (ADHD). In particular, a score can be generated based on electroencephalography or magnetoencephalography (EEG / MEG) data by calculating one or more correlation values ​​between pairs of time series determined based on the EEG / MEG data. The devices, systems, computer-readable storage media, and computer-implemented methods described herein generate a score indicative of a subject having a cognitive disorder, such as a cognitive impairment, such as attention deficit hyperactivity disorder (ADHD).

[0006] According to the present disclosure, an apparatus is configured to receive first EEG / MEG data of a subject. The apparatus is further configured to determine a first plurality of time series based on the EEG / MEG data, each of the first plurality of time series corresponding to a respective source location located within the subject's cranial cavity. The apparatus is further configured to calculate a first correlation value for a pair of the first time series, the pair of first time series being included in the determined first plurality of time series. The apparatus is further configured to generate a score based on the first correlation value, the score indicating that the subject has a cognitive disorder, such as a cognitive disorder like attention deficit hyperactivity disorder (ADHD). The apparatus is further configured to output the score.

[0007] According to one embodiment, the source location of each of the first time series of the first pair of time series may be located in the frontal lobe, such as the middle frontal gyrus (MFG) and / or the inferior frontal gyrus (IFG), of the subject's cranial cavity, and the source location of each of the second time series of the first pair of time series may be located in the superior parietal lobe (SPL) of the subject's cranial cavity.

[0008] According to a further embodiment, the apparatus may be further configured to calculate a second correlation value for a second pair of time series, the second pair of time series being included in the determined first plurality of time series. A score may be generated based on the first correlation value and the second correlation value. According to this embodiment, a source location of each of the first time series of the second pair of time series may be located in a ventral attention network region, such as the temporoparietal junction (TPJ) of the subject's cranial cavity, and a source location of each of the second time series of the second pair of time series may be located in a dorsal attention network region, such as the superior parietal lobe (SPL) of the subject's cranial cavity.

[0009] According to an embodiment, the device may be further configured to calculate a third correlation value for a third pair of time series, the third pair of time series being included in the determined first plurality of time series. A score may be generated based on the first correlation value, the second correlation value, and the third correlation value. Furthermore, a source location of each first time series of the third pair of time series may be located in a ventral attention network region, such as the temporoparietal junction (TPJ) of the subject's cranial cavity, and a source location of each second time series of the second pair of time series may be located in a ventral visual pathway region, such as the inferior temporal gyrus (ITG) of the subject's cranial cavity.

[0010] In a further embodiment, the device may be further configured to receive second EEG / MEG data of the subject and determine a second plurality of time series based on the second EEG / MEG data, each of the second plurality of time series corresponding to a respective source location located within the subject's cranial cavity. Furthermore, the device may be further configured to calculate a fourth correlation value for a fourth pair of time series, the fourth pair of time series being included in the determined second plurality of time series, the fourth pair of time series corresponding to the same respective source locations as the first pair of time series. Furthermore, the device may be configured to calculate a comparison value between the fourth correlation value and the first correlation value. Furthermore, a score may be generated based on the comparison value.

[0011] In an embodiment, the step of determining the first plurality of time series may include filtering the first EEG / MEG data such that only the first plurality of time series includes a subset of all available time series in the first EEG / MEG data, the subset including the first pair of time series, the second pair of time series, and the third pair of time series.

[0012] In an embodiment, a system may comprise an apparatus according to any of the preceding embodiments, and may further comprise an EEG or MEG device for measuring at least first EEG or MEG data of the subject, and a task presentation device for presenting one or more tasks to the subject while the EEG / MEG device is measuring the at least first EEG / MEG data of the subject.

[0013] According to the present disclosure, a computer-implemented method includes receiving first electroencephalogram or magnetoencephalogram (EEG / MEG) data from a subject. The method further includes determining a first plurality of time series based on the first EEG / MEG data, each of the first plurality of time series corresponding to a respective source location located within the subject's cranial cavity. The method further includes calculating a first correlation value for a pair of the first time series, the pair of first time series being included in the determined first plurality of time series. The method further includes generating a score based on the first correlation value, the score indicating that the subject has a cognitive disorder, such as a cognitive disorder like ADHD, and outputting the generated score.

[0014] According to an embodiment, the source location of each of the first time series of the first pair of time series may be located in the frontal lobe, such as the middle frontal gyrus (MFG) and / or the inferior frontal gyrus (IFG), of the subject's cranial cavity, and the source location of each of the second time series of the first pair of time series may be located in the superior parietal lobe (SPL) (340) of the subject's cranial cavity.

[0015] According to an embodiment, the method may further include calculating a second correlation value for a second pair of time series, the second pair of time series being included in the determined first plurality of time series. A score may be generated based on the first correlation value and the second correlation value. In this embodiment, a source location of each of the first time series of the second pair of time series may be located in a ventral attention network region, such as the temporoparietal junction (TPJ) of the subject's cranial cavity, and a source location of each of the second time series of the second pair of time series may be located in a dorsal attention network region, such as the superior parietal lobe (SPL) of the subject's cranial cavity.

[0016] According to a further embodiment, the method may further include calculating a third correlation value for a third pair of time series, the third pair of time series being included in the determined first plurality of time series. A score may be generated based on the first correlation value, the second correlation value, and the third correlation value. Furthermore, a source location of each first time series of the third pair of time series may be located in a ventral attention network region, such as the temporoparietal junction (TPJ) of the subject's cranial cavity, and a source location of each second time series of the second pair of time series may be located in a ventral visual pathway region, such as the inferior temporal gyrus (ITG) of the subject's cranial cavity.

[0017] According to an embodiment, the method may further include receiving second EEG / MEG data of the subject and determining a second plurality of time series based on the second EEG / MEG data, each of the second plurality of time series corresponding to a respective source location located within the subject's cranial cavity. Furthermore, the method may further include calculating a fourth correlation value for a fourth pair of time series, the fourth pair of time series being included in the determined second plurality of time series and corresponding to the same respective source locations as the first pair of time series. Furthermore, the method may include calculating a comparison value between the fourth correlation value and the first correlation value. A score may be generated based on the comparison value.

[0018] According to an embodiment, determining the first plurality of time series may include filtering the first EEG / MEG data such that only the first plurality of time series includes a subset of all available time series in the first EEG / MEG data, the subset including a first pair of time series, a second pair of time series, and a third pair of time series.

[0019] The embodiments may be combined with each other.The embodiments may also be implemented in a computer-readable storage medium comprising computer-readable instructions that, when executed by a processor, cause the processor to perform the steps.

[0020] This Summary is provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description. This Summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used as an aid in determining the scope of the claimed subject matter.

[0021] These and further aspects and features of the present disclosure will be explained in the following detailed description of embodiments of the present disclosure, with reference to the accompanying drawings. [Brief explanation of the drawings]

[0022] [Figure 1] FIG. 1 illustrates a Shannon brain model according to an embodiment of the present disclosure. [Figure 2] FIG. 1 illustrates a dorsal-ventral attention network according to an embodiment of the present disclosure. [Figure 3] FIG. 10 illustrates conditional probabilities and information from an exemplary task (“Task 2”) performed by a subject while EEG or MEG data is measured from the subject. [Figure 4] FIG. 1 shows experimental results showing links with significantly different PLV values ​​in a comparison between two conditions of the same task ("Task 1") measured from healthy subjects. [Figure 5] FIG. 1 shows experimental results showing links with significantly different PLV values ​​in a comparison between two conditions of two respective tasks ("Task 1" and "Task 2") measured from healthy subjects. [Figure 6] FIG. 1 shows experimental results showing links with significantly different PLV values ​​in a comparison between two conditions of two respective tasks ("Task 1" and "Task 2") measured from subjects with ADHD. [Figure 7] FIG. 1 shows experimental results showing links of the central executive network with significantly different PLV values ​​in comparison between healthy subjects and subjects with ADHD for the conditions of the task ("Task 1"). [Figure 8] FIG. 1 shows experimental results showing links in the salience network with significantly different PLV values ​​in healthy subjects compared to subjects with ADHD for conditions of a task ("Task 1"). [Figure 9] FIG. 1 shows experimental results showing links of the central executive network with significantly different PLV values ​​in comparison between healthy subjects and subjects with ADHD for conditions of a task ("Task 2"). [Figure 10]FIG. 1 shows experimental results showing links of the central executive network with significantly different PLV values ​​in comparison between healthy subjects and subjects with ADHD for conditions of a task ("Task 2"). [Figure 11] FIG. 1 shows experimental results showing links in the salience network with significantly different PLV values ​​in healthy subjects compared to subjects with ADHD for conditions of a task ("Task 1"). [Figure 12] FIG. 1 shows experimental results showing links of the central executive network with significantly different PLV values ​​in comparison between healthy subjects and subjects with ADHD for the conditions of the task ("Task 1"). [Figure 13] FIG. 1 illustrates an exemplary system in accordance with one or more embodiments of the present disclosure. [Figure 14] FIG. 1 illustrates an exemplary data analysis apparatus according to an embodiment of the present disclosure. [Figure 15] FIG. 1 illustrates the brain within the cranial cavity of a subject. [Figure 16] FIG. 1 illustrates a method according to an embodiment of the present disclosure. [Figure 17] FIG. 1 is a block diagram illustrating various exemplary components of an attention detection system for accurately detecting attention deficits, according to some embodiments of the present disclosure. [Figure 18A] 1 is a table of example tasks for attention deficit detection, according to some embodiments of the present disclosure. [Figure 18B] 1 is a table of example tasks for attention deficit detection, according to some embodiments of the present disclosure. [Figure 19] FIG. 1 illustrates a connectivity network between brain regions, according to some embodiments of the present disclosure. [Figure 20-1] 1 is a flowchart illustrating the operation of an exemplary method for attention deficit detection, according to some embodiments of the present disclosure. [Figure 20-2] 1 is a flowchart illustrating the operation of an exemplary method for attention deficit detection, according to some embodiments of the present disclosure. [Figure 20-3]1 is a flowchart illustrating the operation of an exemplary method for attention deficit detection, according to some embodiments of the present disclosure. [Figure 20-4] 1 is a flowchart illustrating the operation of an exemplary method for attention deficit detection, according to some embodiments of the present disclosure. [Figure 21] 1 is a flowchart illustrating the operation of an exemplary method for attention deficit detection, according to some embodiments of the present disclosure. DETAILED DESCRIPTION OF THE INVENTION

[0023] Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings, wherein like reference numerals refer to like elements throughout.

[0024] According to one embodiment, the probability in Bayes' theorem can be converted to Shannon information by applying a -log function to both sides of the equation. Therefore, we can define Transferred Information (TIC), a Shannon metric that can reconcile Shannon surprise and Bayes surprise. Furthermore, within the field of behavioral selection or decision-making, TIC can explain how information flows (e.g., through which pathways or networks) to increase or decrease uncertainty about behavior depending on the behavioral relevance of stimuli. Therefore, TIC can be a link between belief and knowledge in cognitive modeling.

[0025] The decomposition of TIC can show that a single observation of a stimulus can generate two information quantities. One of the information quantities, called the bottom-up component, can be evidence-related information (e.g., a stimulus-limiting surprise that may trigger the occurrence of the event) and can contribute a positive information bit to the information flow. The other, called the top-down component, can be Bayesian likelihood-related information (a negative information bit) and can contribute a negative information bit to the information flow. Therefore, if the absolute value of the latter is smaller than that of the former, this results in a net positive information measure (TIC>0). Otherwise, if the negative bit introduced by the top-down component is greater than the bottom-up positive bit, this results in a net negative information measure (TIC<0).

[0026] Furthermore, we can define conditional TICs that can be derived from conditional Bayes' theorem, thus deriving information flows that are conditional on internally maintained contextual representations (hereafter referred to as "contextual TICs") or temporal contexts (hereafter referred to as "episodic TICs"). TICs can be additive, so by adding up successive sensorimotor TICs, contextual TICs, and episodic TICs generated during evidence integration, we can calculate the total uncertainty reduction. Thus, the uncertainty about action selection can be increased or decreased depending on the signs of these TICs. Only when this uncertainty is converted into certainty can a given action be selected.

[0027] According to embodiments, insights gained from TIC can be used to discover attention networks for cognitive disorders, such as cognitive disorders like attention deficit hyperactivity disorder (ADHD). For example, subjects with cognitive disorders, such as ADHD, may need to increase prefrontal cortex (PFC) activity when dealing with exceptions to certain conditions (e.g., episodic task switching) or when filtering out unimportant stimuli, whereas healthy subjects may perform better by increasing activity of the dorsal-ventral attention network discovered by one or more embodiments of the present disclosure without increasing PFC activity when dealing with exceptions to conditions or filtering out unimportant stimuli. This may explain why subjects with cognitive disorders (e.g., ADHD) may have more difficulty (compared to healthy subjects) staying focused, filtering out unimportant information, or dealing with exceptions. Subjects with cognitive disorders may need to use the PFC to filter out unimportant information or when dealing with exceptions, which can be extremely tiring compared to using a direct dorsal-ventral attention network without the PFC (as healthy subjects do). This network may therefore play a central role in bridging the gap between the neurobiology and symptoms of ADHD.

[0028] As described below, activity in the PFC pathway or at least one of the activity in the dorsal-ventral attention network can be used as an indicator to detect whether a subject has a cognitive disorder, such as a cognitive disorder like ADHD.

[0029] Shannon Brain Model In the prior art, a neuron code with log probability that a feature (e.g., an image) can take a specific value has been proposed. The neuron code with log probability was described in Rao, RPN, 2004, "Bayesian Computation in Recurrent Neural Circuits" (Neural Comput. 16, pp. 1-38) and Pouget, A. et al., 2013, "Probabilistic brains: knowns and unknowns" (Nat. Neurosci. 16, pp. 1170-1178). Nevertheless, how the Bayesian brain transitions from a probabilistic brain to a belief or one-off decision occurs has not been fully understood in the prior art.

[0030] According to an embodiment, event x i The Shannon information (or simply information in the following) of (which can be measured in bits) can be given as:

[0031] I(x i )=-log2[p(x i )](bits)(Formula 1)

[0032] Shannon entropy, also measured in bits, can be expressed as the sum of the information content of each event weighted by its probability of occurrence.

number

number

[0033] According to one embodiment, it has been shown that expressing traditional probability theorems (such as Bayes' theorem) in terms of Shannon information can enable understanding how information is transmitted and how traditional negative information is generated. Averaging, such as the entropy expression in Equation (2), can provide interesting statistical implications, such as uncertainty about the content of a message, which Shannon was interested in for communication purposes, but can result in a loss of precision gained by a given measurement. As events occur, probabilistic reasoning (performed in the brain) can transition to a deterministic environment given by Shannon information. As further shown below, this deterministic feature of Shannon information can serve as a bridge between belief and knowledge in cognitive modeling.

[0034] According to one embodiment, a Shannon metric can be derived that can reconcile stimulus-constrained surprises (such as Shannon surprises) and allow for the calculation of information gains (such as Bayes surprises) between prior and posterior beliefs. According to this embodiment, Bayes' theorem is expressed in terms of Shannon's information by applying a -log function to both sides of Bayes' theorem. The effect of data x on an observer can be a change in the prior distribution P(θ) of model θ to the posterior distribution P(θ|x) according to Bayes' theorem. P(θ|x)=P(x|θ)P(θ) / P(x) (Equation 4) where P(x) is known as the evidence and P(x|θ) is known as the likelihood. Next, taking the -log2 function on both sides of equation (4) can result in -log2P(θ|x)=-log2[P(x|θ)P(θ) / P(x)] (Equation 5)

[0035] By applying the properties of the log function, we can expand the right-hand side of equation (5) to obtain the following equation: -log2P(θ|x)=-log2P(x|θ)-log2P(θ)+log2P(x) (Equation 6)

[0036] Considering equation (1), equation (6) can be expressed in terms of the amount of information corresponding to each of the probabilities that appear in Bayes' theorem. I(θ|x)(bits)=I(θ)(bits)-I(x)(bits)+I(x|θ)(bits) (Equation 7)

[0037] In equation (7), each term can be the information content of a particular event, and can represent the surprise that triggers the occurrence of the event in bits. By interpreting this last equation, it becomes possible to understand how the bits of information introduced into the space of models can be transmitted to each model, and concepts such as negative information can naturally arise. From equation (7), it can be deduced that the effect of data on the observer can be to change the information content of a given model θ in the space of models, i.e., to change (e.g., increase or decrease) the surprise that triggers the occurrence of the event. Therefore, the term I(θ|x) on the left side of equation (7) can be called the "posterior information content of a single model θ," and the first term I(θ) on the right side of equation (7) can be called the "prior information content of a single model θ."

[0038] An aspect of equation (7) is that the change in the information content of a single model θ in the space of models that may be induced by a single observation x may include two information quantities I(x), I(x|θ) generated by the single observation (rather than just one information quantity as would be expected). Thus, in equation (7), -I(x) (bits) + I(x|θ) (bits) may represent the information conveyed to a single model θ by observation x. The transferred information quantity (TIC) may be defined as follows: TIC(x→θ)(bits)=I(x)(bits)-I(x|θ)(bits) (Equation 8) Therefore, equation (7) can be expressed as follows: I(θ|x)(bits)=I(θ)(bits)-TIC(x→θ)(bits) (Equation 9)

[0039] In Equation (8), the first term I(x) can be Shannon information associated with the evidence (e.g., Shannon surprise), and positive bits can contribute to reducing the surprise of the posterior information I(θ|x). The second term I(x|θ) in Equation (8) can also be Shannon information, but because it is related to the Bayesian likelihood, it can also be Bayesian. This second term can contribute to increasing the surprise of the posterior information I(θ|x) due to negative bits. For example, this can be considered a surprise in the evidence I(x), and therefore only bits associated with a single model θ can be transmitted to that model. Therefore, TIC can make it possible to calculate information gain (e.g., Bayesian surprise) between prior beliefs and posterior beliefs. Because Shannon information associated with the Bayesian likelihood can be a link between probabilistic and deterministic frameworks, this component of TIC can be a candidate for modeling Bayesian brain transitions (discussed further below).

[0040] As mentioned above, since TIC can account for information flow, it is not surprising that its sign can be of interest. In particular, if the negative information bits of the Bayesian likelihood I(x|θ) do not cancel out the positive information bits of the evidence I(x), then TIC is positive (i.e., there can be positive net information transfer). Conversely, if the negative bits corresponding to I(x|θ) are greater than the bits of the evidence I(x), then TIC is negative, and there can be negative net information transfer. Thus, TIC can overcome the limitations of other information gain models.

[0041] According to one embodiment, a conditional TIC can be defined as follows:

number

[0042] Using equation (9) in the context of action selection, the information transfer that does or does not cause action a upon the occurrence of stimulus s can be expressed in terms of the so-called sensorimotor TIC(s→a). I(a|s) = I(a) - TIC(s→a) (Equation 11) Here, s can denote a stimulus and a can denote an action. Equation (11) indicates that in behavioral selection, the TIC from stimulus s to action a can reduce (TIC > 0) or increase (TIC < 0) the uncertainty about action a. In the extreme case where TIC > 0 cancels out the prior uncertainty I(a), there can be a certainty that the action will be performed, i.e., I(a|s) = 0. In other words, if the uncertainty about an action is reduced to zero (through TIC), a decision to perform that action can be made. As mentioned above, TIC can be the subtraction of two information quantities. Addition and subtraction of information quantities can be completed on a short timescale (e.g., the timescale of the brain's attention), in contrast to the multiplication and division of probabilities that can occur when the brain (e.g., a Bayesian brain) cannot derive estimates (e.g., Shannon information). Therefore, we propose that TIC can be a (deterministic) neural response that can reflect behavioral selection (or decision-making). This can bridge the gap between belief and knowledge. Since the net information transfer by the TIC can be positive or negative, this gap can be increased or decreased, in either case adding precision to the previously estimated uncertainty.

[0043] The TIC in equation (11) can be decomposed as follows: I(a|s)=I(a)-TIC(s→a)=I(a)-I(s)+I(s|a) (Equation 12)

[0044] This decomposition of the TIC may enable us to identify and separate bottom-up and top-down components in action selection (or decision-making). This decomposition into bottom-up and top-down components may correspond to an attentional model of action selection called the Shannon brain model.

[0045] The first component of the TIC, I(s), can represent the amount of information related to the evidence (e.g., the stimulus-limited Shannon surprise elicited by stimulus s). Therefore, I(s) can be a bottom-up component of the attention model that can generate positive information bits and thus contribute to reducing the information content of behavior (e.g., reducing or even counteracting uncertainty). The second component of the TIC, I(s|a), (e.g., the amount of information related to Bayesian likelihood), can represent negative information generated to calculate Bayesian surprises that can increase the information content of behavior (e.g., increasing uncertainty). Therefore, I(s|a) can be a top-down component of the attention model and thus can appropriately process top-down attentional signals. In other words, a high top-down contribution may result in non-performance of behavior.

[0046] According to a further embodiment, the information transfer that causes or does not cause behavior a when stimulus s occurs in context c can also be expressed in terms of TIC: The amount of information corresponding to the probability of causing behavior a conditional on the occurrence of stimulus s in context c can be given as: I(a|s,c)=-log2[P(a|s,c)] (Equation 13)

[0047] Applying the conditional version of Bayes' theorem to the posterior probability in equation (13), we obtain

number

number

[0048] With respect to the sensorimotor TIC, the components I(c|s) and I(c|a,s) of the contextual TIC in Equation (17) can represent bottom-up and top-down components, respectively, in behavioral selection. It may be possible for the sensorimotor TIC to have a different sign than the contextual TIC; thus, an informative stimulus with a positive sensorimotor TIC, when considered in a given context, can generate a negative contextual TIC that compensates for the positive sensorimotor TIC, making the stimulus uninformative about that context.

[0049] In additional embodiments, a stimulus s occurs in a context c and an episode (e.g., temporal context) e, and the information transfer that causes or does not cause behavior a can be calculated in terms of TIC. In this case, the amount of information can be given by the following formula: I(a|s,c,e)=-log2[P(a|s,c,e)] (Equation 18)

[0050] Applying the conditional version of Bayes' theorem to the posterior probability in equation (18), we obtain

number

number

[0051] The first term of equation (20) may be given by equations (13) and (16). From equation (10), the second term may be the episode-conditional TIC: TIC(c→a|s,c), which can represent the information bit conveyed to behavior when a stimulus occurs in a given context c and episode e. Therefore, the amount of information can be expressed as follows: I(a|s,c,e)=I(a)-TIC(s→a)-TIC(c→a|s)-TIC(e→a|s,c) (Equation 21)

[0052] Furthermore, the sensorimotor TIC, the context-conditioned TIC, and the episode-conditioned TIC in Equation (21) can be decomposed into two components.

[0053] I(a|s,c,e)=I(a)-TIC(s→a)-TIC(c→a|s)-[I(e|s,c)-I(e|a,s,c)]=I(a)-I(s)+I(s|a)-I(c|s)+I(c|a,s)-I(e|s,c)+I(e|a,s,c) (Equation 22)

[0054] The components I(e|s,c) and I(e|a,s,c) of the episode-conditional TIC in (22) can represent the bottom-up and top-down components, respectively, in action selection. Following the reasoning in the context case, if the episode-conditional TIC in (22) is positive and compensates for the negativity of the context-conditional TIC, an uninformative stimulus in a given context can become informative.

[0055] From Equation (22), it can be seen that the uncertainty about the behavior when a stimulus occurs in a given context and episode can increase or decrease depending on the bottom-up and top-down contributions of information (e.g., log-probabilities that can be added and subtracted). This will be further explained below. In Equations (21) and (22), I(a) can represent the bits of uncertainty about the behavior before the stimulus appears, and I(a|s,c,e) is the remaining uncertainty (e.g., which can be higher or lower than I(a)) depending on the sign of each TIC after the stimulus appears and the subject considers context c and episode e (if applicable). The first component of each of the three TICs (I(s), I(c|s), I(e|s,c)) can correspond to the bottom-up contribution of stimulus s and can always contribute to reducing the uncertainty of the behavior (e.g., pressing a button). These bottom-up components appear with negative signs in Equation 22, i.e., they contribute to a reduction in the uncertainty about the behavior (increasing the probability of occurrence). Bottom-up components can always appear with positive signs in the TIC equation because they are considered positive information components (e.g., Shannon surprises). The opposite is true for the second components of each of the three TICs (I(s|a), I(c|a,s), I(e|a,s,c), and the top-down components). Top-down components appear with positive signs in Equation 22, i.e., they can contribute to an increase in the uncertainty about the behavior. Top-down components appear with negative signs in the TIC, so they can be negative information components. Therefore, if their absolute values ​​are greater than the respective Shannon surprises (I(s), I(c|s), I(e|s,c)), the corresponding TIC can be negative, and their net contribution is an increase in the uncertainty about the behavior (decreasing the probability of occurrence). These second components of the TIC can be thought of as top-down components because they can contribute to making stimuli behaviorally relevant (e.g., when low or zero) or non-behaviorally relevant (e.g., when large) (depending on the context or episode).For example, subjects with cognitive disorders, such as cognitive impairments like ADHD, may have difficulty distinguishing between behaviorally relevant and non-behaviorally relevant stimuli, and therefore the top-down component of TIC may be processed differently by these subjects (compared to healthy subjects).

[0056] While the respective roles of the dorsal and ventral attention networks are known in the prior art, how the two networks communicate with each other remains a subject of considerable debate.

[0057] Once the task is explained to the subject, it can be stored in the subject's memory (i.e., brain). The dorsal attention network then sends one or more top-down signals, which may correspond to non-targets, distractors, or other stimuli different from the target, to the ventral attention network through the middle frontal gyrus (MFG), and the top-down signals are then stored in a brain region called the temporoparietal junction (TPJ). The TPJ may include the posterior part of the superior temporal sulcus (STS), the posterior part of the superior temporal gyrus (STG), and the inferior parietal lobule (IPL) (Corbetta et al., 2008, The Reorienting System of the Human Brain: From Environment to Theory of Mind, Neuron 58, 306-324). The inferior parietal lobule includes the angular gyrus (AG) and the supramarginal gyrus (SMG).

[0058] Below, with reference to FIG. 1, we describe the information flow of positive and negative bits that can occur between the dorsal attention network (DAN) and the ventral attention network (VAN) (e.g., within the Shannon brain model). Furthermore, as previously mentioned, the top-down contribution in the Shannon brain model may be given by the amount of information related to Bayesian likelihood, which can generate negative information bits that can cancel out bits generated by evidence (or surprise). Thus, the TPJ (e.g., including the IPL and STG) can function as a memory for negative information bits. This is consistent with the role of the ventral attention network (VAN), which can be activated only when stimuli are behaviorally relevant (e.g., one or more targets or incidental distractors) in a given context or episode (e.g., when the stimuli are informative, such as when TIC>0). Then, the information content associated with the Bayes likelihood may be low or zero, and thus the bottom-up bits (e.g., generated in the brain's visual cortex) that may be transmitted by the stimulus to the dorsal attention network (DAN) cannot be counteracted by the top-down negative bits arising from the ventral attention network, and the stimulus may be prioritized and reinforced in one or more visual cortical areas. On the other hand, if the stimulus is uninformative (TIC≦0), the absolute value of the information content associated with the Bayes likelihood may be quite high, and thus the ventral attention network (VAN) may send a large amount of negative bits to the dorsal attention network (DAN) that may counteract the positive bottom-up bits of the stimulus. As a result, non-behaviorally relevant stimuli (non-targets, distractors, etc.) may lose priority and, as a result, be ignored.

[0059] This may be applicable to the framework of sensorimotor TIC, contextual TIC, and episodic TIC (e.g., due to the linearity of Equations 21 and 22, where the contributions of these TICs are taken into account). Therefore, the TPJ can function as a memory for the negative information bits of sensorimotor, contextual, and episodic information provided by I(s|a), I(c|a,s), and I(e|a,s,c) (e.g., the top-down contributions of TICs). In the Shannon brain model described above, these three information quantities related to Bayesian likelihood can be subtracted by the corresponding bottom-up contributions of TICs, i.e., I(s), I(c|s), and I(e|s,c), respectively, in the priority map brain region. Therefore, sensorimotor TIC, contextual TIC, and episodic TIC can be generated and subtracted from the prior uncertainty of behavior I(a), deriving the posterior uncertainty I(a|s,c,e). As mentioned above, different situations may arise depending on the sign of these TICs. For example, the contextual TIC can be negative, meaning the stimulus is uninformative in the context, but if the episodic TIC becomes positive, the episodic TIC cancels out the contextual TIC, making the stimulus informative in the new temporal context. This is illustrated in Figure 1.

[0060] As shown in Figure 1, the TPJ can function as a memory for sensorimotor, contextual, and episodic negative information bits, given by I(s|a), I(c|a,s), and I(e|a,s,c), respectively. These three components can be subtracted from the bottom-up sensorimotor, contextual, and episodic contributions I(s), I(c|s), and I(e|s,c), respectively, in one or more of two priority-mapped brain regions (e.g., the anterior inferior temporal cortex (AIT) and / or the intraparietal sulcus (IPS) / superior parietal lobule (SPL)). Thus, sensorimotor, contextual, and episodic TICs can be generated and subtracted from the prior uncertainty of behavior I(a), deriving the posterior uncertainty I(a|s,c,e). The dorsal attention network (DAN) and ventral attention network (VAN) can communicate through the MFG through the involvement of the salience network (SN) (not shown in Figure 1).

[0061] The ventral attention network (VAN) not only sends signals (e.g., one or more bottom-up signals for reorientation) to the dorsal attention network (DAN) in an activated state, but also sends signals (negative bits of one or more top-down signals) to the dorsal attention network (DAN) in a deactivated state. Thus, there can be information flow from the ventral attention network (VAN) to the dorsal attention network (DAN) not only in an activated state but also in a deactivated state. This can be counterintuitive because deactivation and information flow can be considered opposing concepts, but on the other hand, it can help understand the equally counterintuitive concept of negative information because it can be the type of information expected to come from a deactivated source.

[0062] Furthermore, the prior art has known about a color-biased network (CBN) (within the visual ventral pathway (VVP)), which may include the bilateral lingual gyrus, fusiform gyrus, right inferior occipital gyrus, and one or more adjacent occipito-temporal-parietal regions. While the posterior regions of this network are passive and capable of computing low-level color information (e.g., color perception), the anterior regions are active and capable of encoding more complex functions (e.g., color knowledge), such as combining shape and color for object recognition purposes, storing typical colors of given objects in memory and associating them, or storing the color and shape of given behaviorally relevant stimuli in memory. The relationship between this network and memory allows this relationship to be linked to the hippocampal brain structure. These associations are known as object-color memory. Shape and color can be processed separately throughout much of the visual ventral pathway (VVP), and indeed, a shape-biased lateral occipital (LO) network is known. Both networks converge on the most anterior color-biased regions (color knowledge) (e.g., the anterior inferior temporal cortex (AIT)), where shape and color can be combined for the purpose of object recognition.

[0063] Similar to other networks such as the dorsal attention network (DAN) and ventral attention network (VAN), the color-bias network (CBN)'s cognitive functions (color perception, color naming, and color knowledge) can be linked through one or more bottom-up (e.g., feedforward) posterior-to-anterior pathways and can also be top-down modulated (e.g., feedback) through one or more anterior-to-posterior pathways. Thus, irrelevant stimuli can be suppressed in early visual cortex by top-down modulation based on the encoding of object-color knowledge. Similar to what has been described above for the dorsal attention network (DAN) and ventral attention network (VAN), top-down modulation of the color-bias network (CBN) and shape-bias network (SBN) can arise from filtering signals generated in the prefrontal cortex (PFC) and stored in the TPJ of the ventral attention network (VAN), whose activation can be restricted to relevant object-color knowledge. The ventral attention network (VAN) can then transmit this top-down signal to the AIT, which can modulate the representation of preferential object-color knowledge by enhancing or suppressing the influence of bottom-up inputs from early visual cortex.

[0064] Again, the framework of sensorimotor TIC, contextual TIC, and episodic TIC finds a natural place in this top-down modulation role of the ventral attention network (VAN), along with the color-biased network (CBN) and shape-biased network (SBN). The TPJ can store negative bits of object-color knowledge, corresponding to the sensorimotor, contextual, and episodic information amounts associated with the Bayesian likelihood in Equation (22), namely, I(s|a), I(c|a,s), and I(e|a,s,c), respectively. We can then adjust the representation of preferential object-color knowledge by subtracting each of these information amounts from the bottom-up components of the AIT, I(s), I(c|s), and I(e|s,c), respectively, and calculating the corresponding sensorimotor, contextual, and episodic TICs in Equation (21), as shown in Figure 1. If stimulus color is considered to be a temporal context or episode that may be more or less behaviorally relevant, then the sign of the episodic TIC contributes to increasing or decreasing uncertainty about behavior. Thus, the ventral attention network (VAN) may top-down modulate (e.g., inhibit or reorient) not only the dorsal attention network (DAN), but also the color-bias network (CBN) and the shape-bias network (SBN).

[0065] Anterior-posterior communication may exist between the dorsal attention network (DAN) and the ventral attention network (VAN). In particular, frontal areas (e.g., the prefrontal cortex (PFC)) may be candidates for this interaction. In one example, the inferior frontal gyrus (IFG) and the middle frontal gyrus (MFG) may be candidates for this interaction. Furthermore, this interaction may be driven by a different network, the salience network (SN, not shown in Figure 1), based on the sustained activation of two of its nodes, the anterior cingulate cortex (ACC) and the anterior insula (AI), during top-down filtering or reorientation. The same interaction may also apply to the shape bias network (SBN) and color bias network (CBN) and the ventral attention network (VAN); that is, their communication may be driven by the SN through the IFG and MFG. During cognitive tasks, the SN can play two roles: "set maintenance" and monitoring decisions made to monitor task goals. Therefore, the SN may be involved in anterior-posterior interaction between the ventral attention network (VAN) and these other networks.

[0066] As shown in Figure 2, in addition to anterior-posterior transmission, a (direct) dorsal-ventral tract (DVAN) can coexist with the anterior-posterior transmission described above in the context of Figure 1. In the dorsal-ventral attention network (DVAN), the TPJ of the ventral attention network (VAN) can top-down modulate (e.g., filter or reorient) any one of the dorsal attention network (DAN), ventral shape network (VVP), and color bias network (CBN) to adjust sensory input without involving PFC sites. This dorsal-ventral attention network (DVAN) is illustrated in Figure 2. In particular, Figure 2 shows the coexisting dorsal-ventral tract (DVAN) in addition to the anterior-posterior transmission described above. This dorsal-ventral attention network (DVAN) may include one or more of the occipital visual cortex, the shape-bias network (SBN) and color-bias network (CBN) of the VVP, two priority maps (IPS / SPL and AIT), and the posterior parts of the dorsal and ventral attention networks. The TPJ of the ventral attention network may also be a node of this network, communicating dorsally and ventrally with the IPS priority map and the AIT priority map, respectively. This network may be a fast track for addressing top-down violations (or filtering irrelevant stimuli) and can compute sensorimotor, contextual, and episodic TICs without the involvement of additional PFC resources. In particular, by using direct links between the IPS / SPL and TPJ and / or between the TPJ and AIT (both shown in Figure 2 as DVAN), information can be exchanged between the dorsal attention network (DAN) and the ventral attention network (VAN) without the involvement of the MFG or IFG (shown by dashed lines in Figure 2). These direct links may therefore ensure that information can be processed by the brain without the involvement of the frontal cortex.

[0067] Experimental results A sample of 28 young children was recruited for a magnetoencephalography (MEG) study: 14 participants with ADHD (age 10 years 5 months, SD = 1 year 11 months, female = 35.7%) and 14 control participants (age 10 years 7 months, SD = 1 year 8 months, female = 35.7%). A neurologist diagnosed and evaluated the ADHD participants and issued a neurological report including a diagnosis of ADHD. The study was approved by the Hospital Clinico San Carlos Review Board in Madrid, and all participants and their legal guardians signed informed consent prior to MEG recording.

[0068] In one example, MEG data, including brain signals, were measured using a 306-channel (102 magnetometers and 204 planar gradiometers) whole-head MEG Elekta Neuromag system. In one example, brain signals were measured using an online anti-aliasing bandpass filter of 0.1 to 330 Hz and a sampling rate of 1000 Hz. In one example, the participant's head shape, along with three reference points (the nasion and the left and right preauricular points) as landmarks, was acquired using a 3D Fastrak digitizer (Polhemus, Colchester, VT). In one example, four head position indicator (HPI) coils were recorded on the participant's scalp (forehead and mastoid). In one example, two pairs of bipolar electrodes were used to record eyeblinks and heart rate, respectively.

[0069] Data preprocessing was performed in several steps. In one example, a temporal extension of spatial signal separation (tSSS) method was applied to remove external noise from the raw data, using a window length of 10 seconds and a correlation threshold of 0.90 as input parameters. In one example, automatic detection of ocular, cardiac, and muscle artifacts was performed and verified by an MEG expert. Finally, in one example, independent component analysis was performed to remove eyeblinks and cardiac activity. In one example, the data was segmented into 1-second trials (300 milliseconds of baseline and 700 milliseconds of task-related data) according to task events, and trials marked as containing artifacts were discarded from subsequent analysis.

[0070] In one example, a Montreal Neurological Institute (MNI) template was used to create a uniform grid with 1 cm spacing between source locations within the participant's cranial cavity, resulting in 2,459 source locations within the participant's cranial cavity. In one example, these source locations were labeled according to the Automated Anatomical Labeling Atlas (AAL), retaining only the 1,188 source locations labeled as one of 76 cortical regions. In one example, this surface, along with a source model in object space, was linearly transformed to match the participant's position within the MEG scanner using head shape as an aid. In one example, the internal skull interface, combined with the previously described source model and the system-provided sensor definitions, was used to solve a forward model and construct a lead field based on a modified spherical solution. In one example, source-level activity was reconstructed using a linearly constrained minimum variance (LCMV) beamformer as the inverse method, averaging the single-trial covariance matrix across segments, and normalized using the Tikhonov method and a lambda factor of 10% of the mean channel power. To calculate a common beamformer filter for the magnetometer and gradiometer, in one example, the data and leadfield matrices were normalized per channel, so that signal amplitudes for different channel types were comparable. In a further example, the three-dimensional time series were projected onto spatial principal components to obtain a single source-level time series for each source location.

[0071] To calculate phase synchronization, source-level activity was calculated separately for each of the following conventional bands: theta (4-8 Hz), alpha (8-12 Hz), low beta (12-20 Hz), high beta (20-30 Hz), and low gamma (30-45 Hz). In one example, band-specific data were filtered using an 1800th-order FIR filter constructed using a Hamming window and applied in a two-pass procedure to remove potential phase distortions. In one example, data were filtered using 1 second of real data as padding on each side to avoid edge effects. It can be seen that a separate time series was obtained for each source location within the participant's cranial cavity.

[0072] In one example, correlation values ​​(e.g., functional connectivity (FC) analysis) were calculated for pairs of time series i, j (each time series corresponding to a respective source location). In one example, FC analysis was calculated using a phase-locking value (PLV, e.g., based on the phase synchronization hypothesis). PLV may be based on the distribution of instantaneous phase differences between two time-dependent signals (e.g., pairs of time series).

number

number

number

[0073] The PLV can take values ​​between 0 and 1, where 0 indicates absolute phase independence (e.g., low or no correlation) and 1 indicates absolute phase dependence (e.g., high correlation). In one example, the instantaneous phase of the signal can be estimated using the Hilbert analytic signal after bandpass filtering. To avoid edge effects on the extraction of the instantaneous phase, in one example, the Hilbert analytic signal can be calculated using 1 second of real data as padding on each side of the trial. In one example, in a time window from 50 to 450 milliseconds after the stimulus, the PLV i,j was calculated for each frequency band (theta 4–8 Hz, alpha 8–12 Hz, low beta 12–20 Hz, high beta 20–30 Hz, and low gamma 30–45 Hz).

[0074] In one example, PLV i,j was calculated independently for each task (e.g., Task 1 and Task 2, described further below) and for each condition (Go, No-Go, and Red E vowel, also described further below). i,j was calculated for each pair of source locations i and j, generating a 1188 × 1188 FC matrix (i.e., i = 1...1188, j = 1...1188). In another example, the region-level PLV i,j We calculated PLV as the average PLV of all cortical sources contained within each of the 76 cortical regions of interest (ROIs), e.g., based on the AAL atlas, to obtain a 76 × 76 whole-brain matrix (i.e., i = 1...76, j = 1...76). i,j It is clear that other sizes of PLV are possible. i,j was calculated for specific cortical networks, such as the salience network (SN) and the global, right, and left central executive networks (ECN). In this case, PLV pairs of ROIs constituting each network were extracted, obtaining a subsequent matrix of 8 × 8 ROIs for the SN and 14 × 14 (right / left 7 × 7 ROIs) for the ECN.

[0075] For the whole-brain analysis example (where i = 1...1188, j = 1...1188 as described above), a permutation-based T-test was employed to compare the PLV values ​​of each pair of ROIs. For comparisons between groups within each functional network and for each condition, a T-test independent-samples analysis can be employed. For within-group comparisons, related-samples analysis can be performed between pairs of conditions of interest, as well as between similar conditions of both tasks within each group and functional network. In the example, the resulting p-values ​​can be corrected for multiple comparisons (number of pairwise comparisons and number of frequency bands) with a false positive rate (FDR) of 0.1 to obtain an FDR-corrected alpha threshold for each analysis. In one example, p-values ​​below the threshold can be reported as significant.

[0076] In one example, in a Go / No-Go task (hereinafter referred to as "Task 1"), one or more stimuli, such as one or more letters, are presented to a participant on a task presentation device, and the participant is instructed to press a button when a vowel appears on the device and not press a button when a consonant appears on the device. In this example, the letters can have different colors (e.g., red, green, blue, white, etc.). Additionally, an exception is added to the task, so that the participant is instructed not to press the button when the letter is red. After completing this task, the participant can begin a new task (hereinafter referred to as "Task 2"), which can add a temporal context or episode in which the participant should press a button when a red vowel "e" appears (e.g., an exception to the exception). This second task can measure cognitive flexibility and the ability to change patterns learned during the first task.

[0077] [Table 1] [Table 2]

[0078] Further in this example, stimuli can be generated in four uniformly distributed groups: GoA = non-red vowels, NoGoA = consonants, GoB = red letter "e," and NoGoB = red vowels other than "e." The probabilities derived from the uniformly distributed stimulus groups can yield the following prior probabilities for the go and no-go conditions of each task: Task 1: Go condition = GoA → P(Go) = 1 / 4, No-Go condition = GoB + NoGoA + NoGoB → P(NoGo) = 3 / 4. Task 2: Go condition = GoA + GoB → P(Go) = 1 / 2, No-Go condition = NoGoA + NoGoB → P(NoGo) = 1 / 2. Figure 3 shows the conditional probabilities of each relevant feature according to the model described above. To avoid infinity when a particular probability tends to zero, 10 is used to calculate the relevant information. -5 A value of can be assumed.

[0079] In the Task 1 condition where consonants are presented to participants, there may be no relevant context, and therefore applying equation (12) gives

number

[0080] Furthermore, in the Task 1 condition where participants are presented with vowels other than red, applying Equation (17) to the context of colors other than red gives us the following equation:

number

[0081] Furthermore, in the Task 1 condition where participants were presented with vowels other than red, applying Equation (17) to the red context yields the following equation:

number

[0082] Furthermore, in the episodic condition of Task 2, when the red vowel "e" is presented to the participant, they should press a button, in which case the same context as the last part (the vowel is red) may occur, but top-down control may not be generated. Applying Equation (22), the values ​​of each previous component may be the same as in the previous case, except for I(a), which may be different in Task 2 due to the increased frequency of pulsations due to the new episodic condition, but two new terms may appear to take into account the episodic component information.

number

[0083] Again, the uncertainty about the action can be reduced to zero, so the outcome is a decision to press the button. Thus, in this episodic case, I(e|s,c) >> I(e|a,s,c) ⇒ TIC(e → a|s,c) > 0, i.e., there may be a positive information transfer from the evidence of the occurrence of the red vowel e to the button press. In this case, the contextual TIC may be negative, but this may be compensated for by the positive episodic TIC, which is greater than the contextual TIC. There are several contributions to Equation 27:

[0084] Sensorimotor TIC: The subject sees a vowel, or Shannon surprise I(s), whose occurrence probability is 3 / 4, or 0.42 bits of positive information, which momentarily reduces uncertainty to 0.58 bits. At this sensorimotor stage, the negative information generated by the other component of the TIC, I(s|a), is null and has no effect. In this case, the subject may have been informed of the vowel occurrence probability in advance, or may have learned it during the task.

[0085] Contextual TIC: Subjects can consider the context, in this case the color of the stimulus, through the bottom-up component I(c|s) of the context-conditional TIC. This again could be the Shannon surprise with a value of 0.58 (probability 2 / 3 of a red vowel). Next, subjects can consider the top-down component (e.g., I(c|a,s)), which can introduce one bit of negative information, increasing the uncertainty about pressing the button (e.g., decreasing its probability).

[0086] Episodic TIC: Up until now, the initial 1-bit uncertainty was reduced by exactly 1 bit (0.42 + 0.58) by the two Shannon surprises I(s) and I(c|s). However, the contextual TIC, introduced by the top-down component I(c|a,s), introduced 1 bit of negative information, increasing the uncertainty back to 1 bit. Thus, when the subject processed that the stimulus was a vowel and that it was red, the subject had 1 bit of uncertainty. If the stimulus was "e," the subject might wait to process it because this could be the episodic condition that should trigger a button press in Task 2. The episodic TIC is 1 bit, and episodic TIC(e = red e → a|s,c) = I(e = red E|s = vowel, c = red) - I(e = red E|a,s = vowel, c = red) = 1 - 0 = 1 bit. Therefore, the episodic TIC can be 1 bit as expected, and subtracting it in Equation 27 reduces the uncertainty to 0 bits, triggering a button pulsation.

[0087] The MEG data measured from participants can include brain signals in the alpha, beta, and theta bands. Below, the experimental results are explained in the context of Figures 4-12. On the left side of each of these figures, views of the posterior, upper, left, and right sides of the brain are shown, along with links showing significantly different PLV values. On the right side of each of these figures, a circle plot shows a schematic of the significant links (corresponding to the plot on the left). The labels are abbreviations of brain regions according to the automated anatomical labeling atlas shown at the end of the description.

[0088] In the first example, the PLV of a healthy control (HC) group (e.g., healthy participants) performing Task 1 in the No-Go condition (consonants or red vowels) can be compared with the Go condition (no red vowels) in the alpha band (8–12 Hz). Figure 4 shows links with significantly different PLV values ​​for this first example. In the No-Go condition, deactivation of the occipito-temporal-parietal cortex and ventral frontal cortex can be observed. In particular, Figure 4 shows lines indicating links with significantly different PLV values ​​in the comparison between the No-Go and Go conditions of the HC group performing Task 1. On the left, diagrams of the posterior, superior, left, and right brain regions are shown. On the right, a circular plot shows a schematic diagram of the significant links. The lines indicate a decrease in PLV values ​​in the No-Go condition compared to the Go condition. It can be seen that there is a link with significantly different PLV values ​​between the left supramarginal gyrus (ISMG, a region of the temporoparietal junction) and the left inferior temporal gyrus (IITG, a region belonging to the VVP). This is consistent with the aforementioned Shannon brain model, in which the ventral attention network (VAN) sends such top-down signals to the VVP network, which can modulate the representation of priority object knowledge by enhancing or suppressing the influence of bottom-up inputs from early visual cortex.

[0089] In this case, in the Go condition (non-red vowel), Equation (25) indicates that the two information quantities related to the Bayesian likelihood, I(s = vowel | a) and I(c = non-red | a, s = vowel), are zero. Conversely, in the No-Go condition, when the stimulus is a consonant, Equation (24) indicates that the information quantity related to the Bayesian likelihood, I(s = consonant | a), is large, and the resulting sensorimotor TIC is negative. When the stimulus is a red vowel, Equation (26) indicates that the contextual TIC is negative due to the non-null value of I(c = red | a, s = vowel). In the No-Go condition, these negative information quantities related to the Bayesian likelihood can deactivate the ventral attention network. Furthermore, when these negative top-down bits are subtracted from the bottom-up bits corresponding to the Shannon surprise of the stimulus, a negative TIC is generated in the anterior inferior temporal lobe, which can deactivate the VVP and occipital visual cortex. This can be seen in Figure 4, where there is a link between the left supramarginal gyrus of the TPJ and the left inferior temporal cortex of the VVP, as well as a link between the temporal and occipital regions, showing the deactivation of all of these regions in the No-Go condition, which can be explained by a negative sensorimotor TIC when the stimulus is a consonant, or a negative contextual TIC when the stimulus is a red vowel. Furthermore, the link between the IFG and the middle temporal gyrus can indicate that the interaction between the ventral attention network and the shape network is driven by the SN.

[0090] In the second example, we compare the PLV in the alpha band (8–12 Hz) of the HC group performing Task 1 in the No-Go condition when the stimulus was a red vowel e with that of Task 2 in the Go episodic condition (red vowel e). Figure 5 shows the links with significantly different PLV values ​​for this second example. In the Task 1 No-Go condition (red vowel e), we observe deactivation of the occipito-temporal-parietal cortex. In particular, Figure 5 shows the links with significantly different PLV values ​​in the Task 1 No-Go condition (red vowel e) compared to the Task 2 episodic Go condition (red vowel e) for the HC group. On the left, diagrams of the posterior, upper, left, and right regions of the brain are shown. On the right, pie charts provide a schematic representation of the significant links. The lines indicate decreased PLV values ​​in the Task 1 No-Go condition (red vowel e) compared to the Task 2 episodic Go condition (red vowel e). For completeness, the links with significantly different PLV values ​​for this second example, as shown in Figure 5, are repeated in the following table:

[0091] [Table 3-1] [Table 3-2] [Table 3-3]

[0092] The PLV results are also consistent with the Shannon brain model described above. For the No-Go condition of Task 1, the amount of information associated with the Bayesian likelihood I(c = red | a,s = vowel) can generate a negative bit that can be stored in the TPJ of the ventral attention network, which can then be sent to the priority map region to counteract the bottom-up bit from the input signal. In the case of red vowels, the Shannon brain model described above predicts that the negative information bit corresponding to the term I(c = red | a,s = vowel) is stored in the TPJ of the ventral attention network, and when a red vowel is presented, this negative information bit is sent to inhibit the color-bias network. This can be observed in Figure 5. There are links with significantly different PLV values ​​between regions of the TPJ (e.g., left supramarginal gyrus, bilateral angular gyrus) and regions of the color-bias network, such as the bilateral lingual gyrus, fusiform gyrus, and right inferior occipital gyrus.

[0093] Furthermore, there is a link between the superior parietal gyrus (a region belonging to the dorsal attention network where a priority map has been identified) and the occipital visual cortex, which also belongs to the dorsal attention network. These links confirm that top-down modulation of the occipital visual cortex by higher-order regions also occurs in the dorsal attention network. In Task 2, as shown in Equation (32), uncertainty about action a can be completely counteracted by episodic TIC, thus strengthening the color-bias network, in contrast to the inactivation of the color-bias network in Task 1. Therefore, the difference in PLV values ​​observed in Figure 5 may be due not only to the inactivation of the color-bias network in Task 1, but also to the strengthening of this network in Task 2. Furthermore, Figure 5 shows that the reconfiguration of mental capacities for task switching implied by the different processing of the red vowel e in Task 1 and Task 2 can be achieved without additional PFC recruitment. This result may support the hypothesized dorsal-ventral attention network shown in Figure 2, in which the ventral attention network top-down TPJ can modulate (e.g., filter or reorient) not only the dorsal attention network but also the ventral shape and color networks to adjust sensory inputs without the involvement of PFC sites.

[0094] In the third example shown in Figure 6, the PLV in the alpha band (8–12 Hz) of the ADHD group performing Task 1 in the No-Go condition when the stimulus was a red vowel e is compared to Task 2 in the Go episodic condition (red vowel e). Figure 6 shows the central executive network (CEN) of the ADHD group (e.g., participants diagnosed with ADHD). Links with significantly different PLV values ​​in the Task 1 No-Go condition (red vowel e) compared to the Task 2 episodic Go condition (red vowel e) are indicated by lines. On the left, diagrams of the posterior, superior, left, and right sides of the brain are shown. On the right, pie charts provide a schematic representation of the significant links. The lines indicate decreased PLV values ​​in the Task 1 No-Go condition (red vowel e) compared to the Task 2 episodic Go condition (red vowel e). Links with significantly different FC values ​​are visible in Figure 6. In contrast to the HC group, in the ADHD group, the reorganization of mental abilities for task switching implied by the different processing of the red vowel "e" in Task 1 and Task 2 can be realized by additional PFC recruitment. For example, the PFC regions involved are the IFG (e.g., the left inferior frontal triangular gyrus and opercular gyrus) and the right middle frontal gyrus, indicating communication between the dorsal and ventral attention networks through the SN. There are three links between these three PFC regions and the superior parietal gyrus (e.g., the left superior parietal gyrus (ISPG)), a node of the dorsal attention network. Furthermore, direct dorsal-ventral communication can be observed in the link between the left superior parietal gyrus and the right angular gyrus of the ventral attention network.

[0095] In the dorsal-ventral attention network (DVAN) described above with reference to Figure 2, the ventral attention network (VAN) can serve to transmit dorsally and ventrally to the dorsal attention network (DAN) and VVP, respectively, which may represent a direct pathway for filtering irrelevant stimuli or redirecting behaviorally relevant stimuli. This pathway may be much more efficient than an anterior-posterior pathway (such as the pathway through the IFG and MFG described with reference to Figure 1) that involves the PFC and may be driven by the SN through the PFC. Therefore, subjects performing Task 1 and Task 2 may exhibit the inefficiency of subjects with cognitive disorders (e.g., ADHD). This is because such subjects may need to increase activity in the PFC pathway when dealing with task switching (e.g., a change in temporal context or episode, such as an exception to a condition). HCs can perform task switching by increasing activity in the dorsal-ventral network without increasing activity in the PFC pathway. Figure 1 can be said to represent the brain connectivity of ADHD, and Figure 2 can be said to represent the brain connectivity of HC, which are involved in the reconfiguration of mental capacity through task switching.

[0096] The fourth example compares PLV values ​​in the Go condition for both Task 1 and Task 2. In the Go condition of Task 1 (non-red vowels), connectivity results show increased PLV values ​​in the alpha band (8–12 Hz) for the HC group compared to the ADHD group. Therefore, behaviorally relevant stimuli, non-red vowels, can be top-down modulated, while irrelevant stimuli (consonants and red vowels) can be suppressed. Figure 7 shows the links of the central executive network (CEN) where the HC group had significantly different PLV values ​​in the alpha band (8–12 Hz) for the Go condition of Task 1 (non-red vowels) compared to the ADHD group. On the left, diagrams of the posterior, superior, left, and right brain regions are shown. On the right, circle plots show a schematic diagram of the significant links. Lines indicate increased PLV values ​​for the HC group compared to the ADHD group in the Go condition of Task 1 (non-red vowels). In particular, Figure 7 shows links belonging to the central executive network (CEN), which includes the dorsal and ventral attention networks. We found that PLV values ​​were increased in HC subjects compared with those in the ADHD group in links between regions of the dorsal attention network (e.g., superior parietal gyrus), regions of the ventral attention network (e.g., bilateral angular gyrus), and regions of the IFG (e.g., opercular gyrus, inferior frontotriangular gyrus), and the middle frontal gyrus. These links indicate interactions between the dorsal and ventral attention networks through the SN, which is the middle frontal gyrus, a link between the two. This may mean that the ventral attention network in HC subjects sends stronger bottom-up reorientation signals to the dorsal attention network than in the ADHD group.

[0097] Furthermore, Figure 8 shows links with significantly different (i.e., increased) PLV values ​​in the alpha band (8–12 Hz) between the HC and ADHD groups in the Go condition (vowels other than red). The illustrated network is the salience network (SN), also known as the cingulo-opercular network. Increased PLV values ​​in the HC compared with the ADHD group can be observed in links between most of the nodes in the cingulo-opercular network (e.g., bilateral insula, bilateral inferior frontal triangular gyrus, bilateral infraorbital frontal gyrus, left inferior opercular frontal gyrus, and right anterior cingulate cortex). The superior cognitive performance of the HC group in Task 1 compared with the ADHD group can also be explained by this increased PLV value in the SN, which may provide stable "set maintenance" throughout the task and serve to monitor the achievement of desired goals. In particular, Figure 8 shows links in the salience network with significantly different PLV values ​​in the HC group compared with the ADHD group in the Go condition (vowels other than red) in Task 1 in the alpha band (8–12 Hz). On the left, diagrams of the posterior, superior, left, and right brain regions are shown. On the right, the circle plots provide a schematic representation of the significant links. The lines indicate increased PLV values ​​in the HC group compared to the ADHD group in the Go condition of Task 1 (non-red vowels).

[0098] In the fifth example, we compare the HC group's significantly reduced PLV values ​​in the beta band (20-30 Hz) during the Go condition of Task 2 (non-red vowels and the red vowel "e") with the ADHD group. In the Go condition of Task 2, the "e" vowel may represent an episodic top-down violation, requiring behavioral flexibility afforded by the reduced beta band. Therefore, the ADHD group's higher beta band synchronization may explain their performance deficits in Task 2, as evidenced by their lower accuracy rates (86.23% for ADHD vs. 92.63% for HC, p=0.009) for both non-red vowels (89.93% for ADHD vs. 95.00% for HC, p=0.017) and red vowel "e" (82.53% for ADHD vs. 90.27% for HC, p=0.022). Figure 9 shows the central executive network (CEN) links in the beta band (20–30 Hz) where the HC group had significantly different PLV values ​​compared to the ADHD group during the Go condition of Task 2. On the left, the posterior, superior, left, and right views of the brain are shown. On the right, the circular plots show a schematic diagram of the significant links. The lines indicate reduced PLV values ​​in the HC group compared to the ADHD group during the Go condition of Task 2. From the PLV links in Figure 9, we can deduce that the HC group exhibits even greater cognitive flexibility (less beta-band synchronization) due to the dorsal-ventral connectivity between the right superior marginal gyrus of the ventral attention network and the right superior parietal gyrus of the dorsal attention network. Nevertheless, there may be other links connecting sites in the dorsal attention network (e.g., bilateral superior parietal gyrus) with the IFG and middle frontal gyrus, indicating the involvement of the SN.

[0099] In the sixth example, we compare the links of the HC group with significantly different (i.e., increased) PLV values ​​in the theta band (4-8 Hz) in the Go condition of Task 2 with those of the ADHD group. Increased PLV values ​​can be observed between regions of the dorsal attention network (e.g., bilateral superior parietal lobule) and regions of the ventral attention network (e.g., left supramarginal gyrus, right angular gyrus, and inferior parietal gyrus). Two links are shown between the inferior parietal gyrus and bilateral middle frontal gyrus in the ventral attention network. Figure 10 shows the links of the central executive network with significantly different PLV values ​​in the theta band (4-8 Hz) in the Go condition of Task 2 for the HC group compared to the ADHD group. On the left, diagrams of the posterior, superior, left, and right regions of the brain are shown. On the right, pie charts show a schematic diagram of the significant links. Lines indicate increased PLV values ​​in the HC group compared to the ADHD group in the Go condition of Task 2.

[0100] In the seventh example, we consider the PLV values ​​in the No-Go conditions of both Task 1 and Task 2. Figure 11 shows links with significantly different (i.e., increased) PLV values ​​in the alpha band between the HC and ADHD groups in the No-Go condition (consonant or red vowel). As in the previous case, the illustrated network is the SN. Increased PLV values ​​in the HC compared with the ADHD group can be observed in links between nodes in the SN (e.g., left insula, bilateral infraorbital frontal gyrus, bilateral anterior cingulate cortex). In this No-Go condition, negative information bits stored in the TPJ can be transmitted to the dorsal attention network through the SN. Therefore, the increased PLV values ​​in the HC group in the alpha band can indicate better inhibition of irrelevant stimuli (e.g., better management of negative information). This can be observed as a higher accuracy rate in the No-Go condition for the ADHD group (6.04% for ADHD vs. 4.04% for HC, p-value = 0.031). Figure 11 shows the links of the salience network in the alpha band (8-12 Hz) where the HC group had significantly different PLV values ​​compared to the ADHD group in the No-Go condition of Task 1 (consonants or red vowels). On the left, diagrams of the posterior, superior, left, and right sides of the brain are shown. On the right, circle plots show a schematic diagram of the significant links. The lines indicate increased PLV values ​​in the HC group compared to the ADHD group in the No-Go condition of Task 1 (consonants or red vowels).

[0101] Furthermore, Figure 12 shows a link between significantly reduced PLV values ​​in the beta band (20-30 Hz) for the HC group compared with the ADHD group in the No-Go condition of Task 1 (consonants and red vowels). In the No-Go condition of Task 1, the red vowel may represent a top-down violation, for which a reduction in the beta band may be desirable to provide cognitive flexibility. This may explain the superior performance of the HC group in Task 1 compared with the ADHD group. Again, these links may indicate that the HC group addresses greater cognitive flexibility (reduced beta band synchronization) through a direct dorsal-ventral pathway, as well as through the SN, which involves the involvement of sites in the IFG and middle frontal gyrus, and through interactions between sites in the ventral attention network (right angular gyrus, right supramarginal gyrus) and the dorsal attention network (right superior parietal gyrus). In particular, Figure 12 shows the links of the central executive network in which the HC group had significantly different PLV values ​​compared to the ADHD group in the beta band (20-30 Hz) during the No-Go condition of Task 1 (consonants or red vowels). On the left, diagrams of the posterior, superior, left, and right sides of the brain are shown. On the right, circle plots provide a schematic representation of the significant links. The lines indicate decreased PLV values ​​for the HC group compared to the ADHD group during the No-Go condition of Task 1 (consonants or red vowels).

[0102] As previously mentioned, Bayes' theorem can be expressed in terms of Shannon information. By applying a -log function to both sides of Bayes' theorem, a Shannon metric called transfer information (TIC) can be defined, which measures the information conveyed by a stimulus and can reduce (for informative stimuli or TIC>0) or increase (for uninformative stimuli or TIC<0) the uncertainty about action selection. Whether a feature of a stimulus is informative or uninformative can depend on its behavioral relevance, for example, in sensorimotor cognitive tasks such as those described above. TIC can be a function of Shannon information and, at the same time, can enable updating of the Bayesian model by measuring the increase or decrease in uncertainty about action uncertainty (e.g., reconciling Shannon surprises with Bayesian surprises).

[0103] The significant decrease in PLV values ​​in the alpha band (8–12 Hz) for the HC group in the No-Go condition (consonant or red vowel) of Task 1 compared with the Go condition (non-red vowel) suggests that the ventral attention network is deactivated by the negative information associated with Bayesian likelihood. Furthermore, when these negative top-down bits are subtracted from the bottom-up bits corresponding to the Shannon surprise of the stimuli, a negative TIC is generated in the anterior inferior temporal lobe, deactivating the VVP and occipital visual cortices. Interestingly, one of the regions with significantly different FC values ​​between the No-Go and Go conditions is the left fusiform gyrus, a region that responds more strongly to letters than to digits. This is consistent with the nature of Task 1, in which the stimuli are letters.

[0104] Furthermore, in the alpha band (8-12 Hz), we also compared the PLV values ​​of the HC group performing Task 1 in the No-Go condition when the stimulus was the red vowel e with those in Task 2 in the Go episodic condition (red vowel e). In the No-Go condition (red vowel e) of Task 1, compared with the same stimulus in Task 2, where uncertainty about action a is completely counteracted by episodic TIC, we observed that the color-bias network was deactivated, and thus the aforementioned color-bias network may have been strengthened. We observed that not only the color-bias network but also the entire dorsal-ventral attention network described above was activated by the HC group, enabling them to manage the top-down violation implied by the red vowel e in Task 2.

[0105] These results may indicate inefficiency in ADHD subjects because they need to increase PFC activity when dealing with task switches (such as exceptions to conditions), which in this case involve changes in temporal context or episodes. In contrast, the HC group can execute task switches by increasing activity in a new temporoparietal-dorsal-ventral attention network without increasing activity in the PFC pathway. In the alpha band, decreased PLV values ​​in the ADHD group compared with the HC group in both the Go and No-Go conditions are observed, a link that may indicate an interaction between the dorsal and ventral attention networks through the SN. Therefore, the ventral attention network in the HC group can send stronger reorienting (Go condition) or filtering (No-Go condition) signals to the dorsal attention network than the ADHD group. Furthermore, again indicating an interaction link between the dorsal and ventral attention networks through the SN, we observe significantly increased PLV values ​​in the beta band (20-30 Hz) in the ADHD group compared to the HC group in the No-Go condition (red vowel) of Task 1 and the Go condition (red vowel e) of Task 2. These two conditions could represent top-down violations requiring a certain cognitive flexibility (e.g., low beta band synchronization).

[0106] System, computer-readable storage medium, and method for generating a score indicative of cognitive disorder

[0107] In consideration of the above discussion and experiments, it has been found that a system, a computer-readable storage medium, and a method can be provided that may be suitable for generating a score indicating that a subject (e.g., a person) has a cognitive disorder, such as a cognitive impairment like ADHD. In particular, the above-mentioned results indicate that when a subject has a cognitive disorder (e.g., a cognitive impairment like ADHD), brain information may be transmitted through a different pathway (or network) compared to a healthy subject. A pathway may include two locations (or nodes) in the brain through which information is transmitted. Therefore, to generate a score indicative of a subject with such a cognitive disorder, it is proposed to search for one or more pathways that distinguish healthy subjects from subjects with a cognitive disorder (e.g., ADHD) (e.g., a pathway that is activated in healthy subjects and inactivated or less activated in subjects with a cognitive disorder, or vice versa). According to one embodiment, the information transmitted through a particular pathway may correspond to a correlation value (e.g., the PLV described above) between two time series (e.g., time series obtained from an EEG or MEG device) measured at two respective source locations located within the subject's cranial cavity. These two source locations may be the two locations of the pathway that distinguish healthy subjects from subjects with the aforementioned disorders. Furthermore, as previously mentioned, subjects with cognitive disorders, such as cognitive disorders like ADHD, may lack the ability to increase activity in the dorsal-ventral attention network (DVAN) without increasing activity in the PFC pathway (e.g., healthy subjects may have this ability due to the aforementioned dorsal-ventral attention network (DVAN) as shown in Figures 1, 2, 5, and 6).

[0108] The inability to increase activity in the dorsal-ventral network (DVAN) may not only be present in subjects with ADHD, but also in subjects with other cognitive disorders (e.g., disorders that may be associated with patients having difficulty maintaining concentration or switching between mental tasks). Thus, scores generated by the present disclosure may be indicative of such cognitive disorders and may assist physicians in diagnosing subjects with such disorders. Furthermore, scores generated by the present disclosure may be used to assess whether a subject is capable of working in certain occupations where maintaining concentration or switching between mental tasks is important (e.g., pilot, truck driver, air traffic controller, etc.).

[0109] According to a first embodiment, a correlation value between a pair of time series of source locations within a subject's cranial cavity (i.e., the subject's brain) included in a PFC pathway can be used to generate a score indicating that the subject has a cognitive disorder, such as a cognitive impairment like ADHD. As previously described, subjects with such cognitive disorders may have increased activity in the PFC pathway (e.g., compared to subjects without the cognitive disorder). According to one embodiment, the PFC pathway may include the frontal lobe (e.g., the first source location of the pathway) and the superior parietal lobule (e.g., the second source location of the pathway) of the brain. Thus, a correlation value (corresponding to information transfer) may be calculated based on a pair of time series including a first time series measured in the subject's frontal lobe and a second time series measured in the subject's superior parietal lobule (SPL). In one embodiment, the frontal lobe (i.e., the first source location) may include at least one of the middle frontal gyrus (MFG) and the inferior frontal gyrus (IFG).

[0110] In a second embodiment, a correlation value between a time series of a source location (within the subject's cranial cavity) that is part of a ventral attention network and another time series of a source location (within the subject's cranial cavity) that is part of a dorsal attention network can be used to generate a score indicative of the subject having a cognitive disorder, such as a cognitive disorder like ADHD. As shown and described above, a healthy subject may have the ability to (directly) transfer information between these two source locations (i.e., a healthy subject can use the dorsal-ventral network (DVAN) described above), whereas a subject having a cognitive disorder, such as a cognitive disorder like ADHD, may lack this ability. Therefore, one or more correlation values ​​between a pair of time series between these two source locations may be appropriate for generating a score indicative of a cognitive disorder (e.g., a cognitive disorder like ADHD). In one embodiment, the location that is part of the ventral attention network may be the temporoparietal junction (TPJ), and the location that is part of the dorsal attention network may be the superior parietal lobule (SPL). As previously mentioned, the TPJ may include the posterior portion of the superior temporal sulcus (STS), the posterior portion of the superior temporal gyrus (STG), and the inferior parietal lobule (IPL), which includes the angular gyrus (AG) and supramarginal gyrus (SMG).

[0111] In a third embodiment, a correlation value between a time series of a source location (within the subject's cranial cavity) that is part of the ventral attention network and a location that is part of the ventral visual pathway (VVP) can be used to generate a score indicative of the subject having a cognitive disorder, such as a cognitive disorder like ADHD. As illustrated and described above, a healthy subject may have the ability to transfer information between these two locations, while a subject with a cognitive disorder, such as a cognitive disorder like ADHD, may lack this ability. Therefore, a correlation value between a pair of time series of these two locations may be appropriate for generating a score indicative of a cognitive disorder (such as a cognitive disorder like ADHD). In one embodiment, the location that is part of the ventral attention network may be the TPJ, and the location that is part of the ventral visual pathway may be the inferior temporal gyrus.

[0112] It should be apparent that not only a single correlation value can be used to generate a score. Instead, multiple correlation values ​​(such as multiple correlation values ​​of the first, second, or third embodiment described above) can be used. Furthermore, the first through third embodiments described above can be combined, using one or more correlation values ​​obtained by one of these embodiments together with one or more correlation values ​​obtained from another one or both of these embodiments to generate a score indicating that a subject has a cognitive disorder, such as a cognitive disorder like ADHD.

[0113] 13 illustrates an exemplary system 100 that can be used to generate a score 150 indicating that a subject 110 has a cognitive disorder, such as a cognitive impairment like ADHD. The system 100 can include an electroencephalography (EEG) or magnetoencephalography (MEG) device 120 for measuring EEG or MEG data 130 and a data analysis device 140 for generating the score 150. In one embodiment, the system 100 can further include a task presentation device 160 for presenting one or more tasks to the subject 110 and an input device 170 for receiving input from the subject 110 as the subject's response to the presented tasks. In yet an additional embodiment, the system 100 can further include a score presentation device 180 for presenting the score 150. In one embodiment, the score 150 is presented to a physician 190, who can diagnose whether the subject 110 has a cognitive disorder, such as a cognitive impairment like ADHD.

[0114] The EEG / MEG device 120 may be a standard EEG or MEG device known in the art. The EEG / MEG device 120 may measure multiple signals from respective source locations located within the cranial cavity of the subject 110. Each source location may correspond to a known brain region, such as a lobe, a lobule, a sulcus, or a gyrus. In some examples, the multiple signals may be measured while the subject 110 performs one or more tasks that may be presented to the subject 110 by the task presentation device 160. The signals measured by the EEG / MEG device 120 may be divided into one or more bands based on the frequency range of the signals. For example, the signals may include (among others) an alpha band (e.g., a frequency band of 8-12 Hz), a beta band (e.g., a frequency band of 12-30 Hz), and / or a theta band (e.g., a frequency band of 4-8 Hz). In one or more embodiments, the signals may correspond to measurements from electrodes of an EEG device or measurements from sensor coils of an MEG device, respectively. In some embodiments, the signal may correspond to an aggregate signal (such as an average signal) across measurements of multiple electrodes (in the case of EEG) or multiple sensor coils (in the case of MEG). Such an aggregate signal may correspond to a region of the brain that includes the location of each of the signals aggregated in the aggregate signal.

[0115] Data analyzer 140 is configured to receive EEG / MEG data 130 and generate a score 150 that indicates that subject 110 has a cognitive disorder, such as a cognitive impairment like ADHD. Data analyzer 140 can be located in the same location (e.g., on-premise) as EEG / MEG device 120 (e.g., data analyzer may be a computing system located at the hospital where EEG / MEG device 120 is located), or data analyzer 140 can be located elsewhere (e.g., off-premise, such as in the cloud) and communicate with EEG / MEG device 140, input device 170, and / or score display device 180 via a network (e.g., the internet). Further details about data analyzer 140 are further described below with respect to FIG. 2.

[0116] The task presentation device 160 can present one or more tasks to the subject 110 while the EEG / MEG device 120 is measuring the brain signals of the subject 110. In one embodiment, the task presentation device 160 can present a first task during a first period when the MEG / EEG device 120 is measuring the brain signals of the subject 110. The first task can include three conditions: a first condition, a second condition, and an exception to the second condition (exception condition). Each condition can include one or more stimuli s and a corresponding action a to be performed by the subject 110 when one of the one or more stimuli s is presented to the subject 110. In one example, subject 110 may be instructed not to interact with input device 170 (e.g., not press a button) when a first condition is presented during a first task, to interact with input device 170 (e.g., press a button) when a second condition is presented during the first task, and to not interact with input device 170 (e.g., press a button) when an exception to the second condition is presented. In one embodiment, the first task may be an example of Task 1, described further above (in the context of FIG. 3) and in more detail below in the context of FIGS. 18A, 18B, and 20-1 through 20-4.

[0117] Additionally, task presentation device 160 may present a second task during a second period (the second period being after the first period) while MEG / EEG device 120 is (still or again) measuring brain signals of subject 110. In one embodiment, the second task may include three conditions like the first task described above and an additional, fourth condition that is an exception to the exception of the second condition, such that subject 110 may be instructed to interact with the input device when the fourth condition occurs during the second task. In one example, the second task may include a first condition in which subject 110 is instructed not to interact with input device 170 and a second condition in which subject 110 is instructed to interact with input device 170. Additionally, the second task may include a third condition that defines an exception to the second condition, such that for the third condition, subject 110 is instructed not to interact with input device 170. Additionally, the second task may include a fourth condition that defines an exception to the exception of the second condition, for which the subject 110 is instructed to interact with the input device 170. In one embodiment, the second task may be a further example of Task 2 described above. The task presentation device 160 may be any device suitable for presenting one or more tasks, such as a screen (e.g., the screen of a tablet or computing device).

[0118] The different conditions of the first task and the second task may occur multiple times, for example, periodically (e.g., every 5 seconds) during each of the first and second periods. Furthermore, the different conditions may occur at a predefined frequency for each different condition (e.g., each condition may occur with the same frequency, 1 / 3 for the first task described above and 1 / 4 for the second task described above). The multiple times the different conditions are presented by the task presentation device 160 can be matched with brain signals measured by the MEG / EEG device 120 to identify brain responses to each different condition. Each of the different conditions (i.e., the three conditions for the first task described above and the four conditions for the second task described above) may correspond to a different stimulus for the brain of the subject 110.

[0119] Input device 170 may be any device capable of receiving input from subject 110 via subject 110's interaction with input device 170. In one embodiment, input device 170 may include a button that subject 110 can press. In another embodiment, the input device may detect gestures or facial expressions of subject 110 that may correspond to subject 110's interaction with input device 170. The subject's 110 input received by input device 170 may be used to determine whether subject 110's interaction with input device 170 was a correct (or incorrect) action in response to the stimulus of the presented condition of the task presented by task presentation device 160. As previously described, subject 110 may be instructed to interact with input device 170 in some conditions (e.g., the second condition and the exception to the second condition) and not to interact with the input device in other conditions (e.g., the first condition and the exception to the second condition). The input of the subject 110 can be used to determine whether the subject 110 interacted with the input device 170 according to these instructions (a correct response) or did not interact (an incorrect response).

[0120] The score presenting device 180 may be any device capable of presenting one or more scores 150 to a physician 190, allowing the physician 190 to diagnose whether the subject 110 has a cognitive disorder, such as a cognitive disorder like ADHD. In particular, the one or more scores 150 presented by the score presenting device 180 may aid the physician 190 in diagnosing whether the subject 110 has a cognitive disorder, such as a cognitive disorder like ADHD.

[0121] The EEG / MEG data 130 may be transmitted to the data analysis device 140. In one embodiment, the EEG / MEG data 130 may include brain signals measured by the EEG / MEG device 120. In one embodiment, the EEG / MEG data 130 may include first EEG / MEG data that may be measured while the subject 110 is performing one task (such as the first task described above) and second EEG / MEG data that may be measured while the subject 110 is performing another task (such as the second task described above). In one embodiment, the brain signals correspond to raw brain signals (such as brain signals measured by the EEG / MEG device 120 without preprocessing). In another embodiment, the brain signals may be preprocessed (e.g., by removing external noise or ocular, cardiac, and / or muscular artifacts) before transmitting the EEG / MEG data 130 to the data analysis device 140. In one embodiment, the EEG / MEG data 130 may include all brain signals measured by the EEG / MEG device 120. In another embodiment, the EEG / MEG data 130 may include only a subset of the brain signals measured by the EEG / MEG device 120. The subset of brain signals may be brain signals from locations within the cranial cavity of the subject 110 as described above for the first, second, and third embodiments, such as brain signals from the frontal lobe (e.g., the IFG and / or MFG) and the superior parietal lobe (see the first embodiment described above), from locations that are part of a ventral attention network (e.g., the TPJ) and locations that are part of a dorsal attention network (e.g., the superior parietal lobe, see the second embodiment described above), and / or from locations that are part of a ventral attention network (e.g., the TPJ) and locations that are part of the ventral visual pathway (e.g., the inferior temporal gyrus, see the third embodiment described above).

[0122] In one embodiment, EEG / MEG data 130 may further include information (e.g., time information) that allows matching the time at which task conditions were presented to subject 110 by task presentation device 160 to brain signals measured by EEG / MEG device 120. Additionally, EEG / MEG data 130 may include data from input device 170 that indicates an interaction between subject 110 and input device 170.

[0123] 2 illustrates an example of a data analysis device 140, 200. As previously described, the data analysis device 140, 200 may be on-premise (e.g., located within the local area network of the EEG / MEG device 120) or off-premise (e.g., located in the cloud). For example, the data analysis device may be a computing system (e.g., a server). The data analysis device 140, 200 may include a receiving component 210, a time series determining component 220, a correlation value calculating component 230, a score generating component 240, and a score output component 250.

[0124] The receiving component 210 may be configured to receive EEG / MEG data 130, such as the first EEG / MEG data and / or the second EEG / MEG data described above. In one embodiment, the receiving component 210 may receive the EEG / MEG data 130 from the EEG / MEG device 120.

[0125] The time series determination component 220 may be configured to determine a first plurality of time series based on the received EEG / MEG data 130 (e.g., based on first EEG / MEG data measured while the subject 110 is performing a task). Each of the first plurality of time series may correspond to a respective source location located within the cranial cavity of the subject 110. Further, the time series determination component 220 may be configured to determine a second plurality of time series based on the received EEG / MEG data 130 (e.g., based on second EEG / MEG data measured while the subject 110 is performing another task). Each of the second plurality of time series may correspond to a respective source location located within the cranial cavity of the subject. For example, the first plurality of time series and the second plurality of time series may correspond to the same respective source location.

[0126] In one embodiment, determining the first (and / or second) plurality of time series may include filtering the first (and / or second) series of EEG / MEG data 130 so that the first (and / or second) plurality of time series may include only a subset of all available time series (e.g., a subset of all available time series that may be obtained from the EEG / MEG device 120). The subset may include only time series between the source locations (e.g., the frontal lobe, SPL, TPJ, and / or ITG) described above with respect to the first, second, and third embodiments. Additionally, filtering may include filtering certain frequencies of the EEG / MEG data (e.g., restricting the time series to the alpha, beta, and theta bands, restricting to only the alpha band, restricting to only the beta band, and / or restricting to only the theta band). Filtering the EEG / MEG data 130 may reduce the amount of data that needs to be processed by the data analyzer 140. For example, filtering the EEG / MEG data 130 can help obtain the data that needs to be processed for analysis by the data analyzer 140 .

[0127] Further, according to one embodiment, determining the first (and / or second) plurality of time series may include preprocessing the first (and / or second) EEG / MEG data 130, such as by removing external noise from the EEG / MEG data (e.g., brain signals measured by the EEG / MEG device 120), removing ocular, cardiac, and muscle artifacts, segmenting the EEG / MEG data 130 so that the time series have the same length (e.g., 1 second), etc. In one embodiment, the time series determination component 220 may be configured to match the time at which the task conditions were presented to the subject 110 by the task presentation device 160 with the EEG / MEG data 130 (e.g., brain signals measured by the EEG / MEG device 120), such that each of the determined plurality of time series includes the start time of each of the task conditions. This can ensure that the subject's brain responses to the conditions are included in the time series. In a particular embodiment, the preprocessing may be performed according to the preprocessing example further described above with respect to the experimental results.

[0128] It should be apparent that the time series determination component 220 can determine not only one time series, but multiple time series. In particular, as previously described, the time series determination component 220 can determine a respective time series for each source location, for each condition of a particular task, and for each subject being tested. By considering only a subset of all available time series (as previously described), the amount of time series to be determined and / or processed can be reduced.

[0129] The correlation value calculation component 230 can calculate a correlation value based on pairs of time series included in the multiple time series (e.g., the first multiple time series or the second multiple time series) determined by the time series determination component 220. In particular, in one embodiment, the correlation value calculation component 230 can calculate a first correlation value for a first pair of time series, the first pair of time series being included in the determined first multiple time series. Further, the correlation value calculation component 230 can calculate a second correlation value for a second pair of time series, the second pair of time series being included in the determined first multiple time series. Further, the correlation value calculation component 230 can calculate a third correlation value for a third pair of time series, the third pair of time series being included in the determined first multiple time series. Further, the correlation value calculation component 230 can calculate a fourth correlation value for a fourth pair of time series, the fourth pair of time series being included in the determined second multiple time series. In one embodiment, the correlation value may be related to the phase of the pair of time series (e.g., the correlation value may be given by the PLV described above or by a circular correlation coefficient), while in another embodiment, the correlation value may be related to the amplitude of the pair of time series (e.g., the correlation value may be given by a Pearson correlation).

[0130] The score generation component 240 can generate one or more scores 150 based on the first correlation value, where the one or more scores indicate that the subject 110 has a cognitive disorder, such as a cognitive disorder like attention deficit hyperactivity disorder (ADHD). In one embodiment, each of the one or more scores 150 can be provided by a respective correlation value calculated by the correlation value calculation component 230. In one embodiment, a score can be the correlation value for a time series corresponding to the frontal lobe and a time series corresponding to the SPL (first pair of time series). Another score can be the correlation value for a time series corresponding to the SPL and a time series corresponding to the TPJ (second pair of time series). Yet another score can be the correlation value for a time series corresponding to the TPJ and a time series corresponding to the ITG (third pair of time series). In one embodiment, these scores can be combined into a single score, and in another embodiment, all three scores can be output.

[0131] According to the above-mentioned considerations (e.g., TIC, its top-down contribution, and / or dorsal-ventral attention network (DVAN)) and experimental results, a high correlation value between a time series corresponding to the frontal lobe and a time series corresponding to the SPL can indicate that the subject has a cognitive disorder, such as ADHD (because the PFC pathway may be increased due to a lack of dorsal-ventral communication). Similarly, a high correlation value between a time series corresponding to the TPJ and a time series corresponding to the SPL can indicate that the subject does not have a cognitive disorder (because dorsal-ventral communication may be activated). Furthermore, a high correlation value between a time series corresponding to the TPJ and a time series corresponding to the ITG can indicate that the subject does not have a cognitive disorder (because dorsal-ventral communication may be activated). A physician who is presented with one or more scores (e.g., one or more of the three scores described above) can use the scores to diagnose whether the subject has a cognitive disorder, such as ADHD.

[0132] Additionally or alternatively, one or more scores 150 may be generated based on comparing the correlation value calculated by the correlation calculation component 230 to one or more other values ​​(e.g., values ​​obtained from the second EEG / MEG data). In particular, a comparison value between the correlation value and the other value may be calculated, and a score may be generated based on the comparison value.

[0133] In one embodiment, the correlation value can be compared to another correlation value, where the other correlation value is obtained for another time series pair from the same source location, the same subject, and the same task, but where another condition of the task may be presented to the subject 110 (e.g., via the task presentation device 160). For example, the condition for the correlation value may be a second condition of the first task described above, and the other condition for the other correlation value may be an exception to the second condition of the first task described above. In another example, the condition for the correlation value may be an exception to the second condition of the second task described above, and the other condition for the other correlation value may be an exception to the exception of the second task described above. The score generation component 240 may be configured to calculate a comparison value between these two correlation values. In one example, the score 150 may be the comparison value. In this case, the score may indicate the subject's ability to deal with the exception. In the same manner as calculating the correlation value, the other correlation value can be calculated by the correlation value calculation component 230 (e.g., based on a time series from the second EEG / MEG data).

[0134] In another embodiment, the correlation value can be compared to another correlation value obtained for another pair of time series from the same source location, the same subject, and the same condition, but when the subject 110 is performing a different task (e.g., presented via the task presentation device 160). For example, the task for the correlation value may be the first task described above (without the exception of the exception of the second condition), and the other task may be the second task described above (with the exception of the exception of the second condition). As described above, the subject may perform the second task after the first task, and thus the subject may need to switch, for example, from not interacting with the input device 170 in the first task to interacting with the input device 170 in the second task (although the task presentation device 160 may present the same stimuli). The score generation component 240 may be configured to calculate a comparison value between these two correlation values. In one example, the score 150 may be the comparison value. In this case, the score may indicate the subject's ability to switch (i.e., reorient) from one task to another. In the same manner as the correlation value is calculated, another correlation value can be calculated by the correlation value calculation component 230 (eg, based on a time series from the second EEG / MEG data).

[0135] In yet another embodiment, the correlation value can be compared to another correlation value obtained for another pair of time series from the same source locations, same conditions, and same task, but for at least one other subject known to have a cognitive disorder, such as a cognitive disorder like ADHD. This other correlation value can be stored in a data store of the data analysis device 140, 200 (not shown in FIG. 2 ). Similarly, the correlation value can be compared to another correlation value obtained for another pair of time series from the same source locations, same conditions, and same task, but for at least one other subject who does not have a cognitive disorder, such as a cognitive disorder like ADHD. Again, this other correlation value can be stored in a data store of the data analysis device 140, 200.

[0136] Additionally, a score can be generated based on information from the input device 170 that indicates whether the subject 110 interacted with the input device correctly according to the requirements of the task. In one example, a high number of incorrect responses can indicate that the subject has a cognitive disorder.

[0137] Each of the aforementioned scores may be output separately as one or more scores 150 to, for example, assist a physician 190 in diagnosing the subject 110. Alternatively, a combined score may be generated (e.g., based on a weighted average of the aforementioned scores) and the combined score may be output to the physician 190.

[0138] The score output component 250 is configured to output one or more scores 150. In one example, the score output component 250 transmits the one or more scores 150 to a score presenting device 180, which can present the one or more scores to a physician 190.

[0139] 15 illustrates a brain within the cranial cavity 300 of a subject 110. The brain may include a frontal lobe 310, a parietal lobe 370, an occipital lobe 380, and a temporal lobe 390. As previously discussed, subjects with cognitive disorders, such as ADHD, may lack the ability to activate the dorsal-ventral network (DVAN), which can link ventral and dorsal attention network regions and / or ventral visual pathway regions. Instead, subjects with such cognitive disorders may exhibit increased activity in PFC pathways. Thus, according to embodiments, one or more scores indicating that a subject has such a cognitive disorder can be generated based on at least one of a correlation value between a time series from the subject's frontal lobe 310 (e.g., from the MFG 320 and / or IFG 330) and a time series from the SPL 340, a correlation value between a time series from the subject's SPL 340 and a time series from the TPJ 350, and a correlation value between a time series from the subject's TPJ 350 and a time series from the ITG 360. Furthermore, for any of these three correlation values ​​(which can be further combined as described above), a correlation value for a pair of time series can be used, where the source location of the first time series (of the pair of time series) is the location where a link (i.e., a link shown in any one of FIGS. 4-12) begins, and the source location of the second time series (of the pair of time series) is the location where that same link ends. One example is the correlation value between time series from the left superior marginal gyrus (ISMG) and the left inferior temporal cortex (IITG), shown as links in FIG. 4. Another example is the correlation value between time series from the left superior parietal gyrus (ISPG) and the left inferior frontal gyrus orbitalis (IIFGo), shown as links in Figure 6. It is clear that the other links shown in Figures 4-12 can be used to calculate correlation values ​​that are used to generate a score indicating that a subject has a cognitive disorder, as previously described.

[0140] FIG. 16 illustrates a computer-implemented method 400 according to one embodiment of the present disclosure. The steps of the method illustrated in FIG. 16 and described below may also be implemented as a computer-readable storage medium that causes a computing system (such as the data analysis device 140) to perform these steps. In a first step 410, EEG / MEG data 130 (such as first EEG / MEG data or second EEG / MEG data) of a subject is received. In a next step 420, a plurality of time series are determined based on the EEG / MEG data. For example, a first plurality of time series is determined based on the first EEG / MEG data, and a second plurality of time series is determined based on the second EEG / MEG data. Each time series of the plurality of time series may correspond to a respective source location within the cranial cavity 300 of the subject 110. Next, in step 430, a correlation value for a pair of time series included in the plurality of time series may be calculated. In step 440, a score 150 is generated based on the correlation value. The score 150 may indicate that the subject has a cognitive disorder, such as a cognitive disorder like attention deficit hyperactivity disorder (ADHD). Next, in step 450, the generated score is output.

[0141] Further examples that may be helpful in understanding the invention Systems and methods consistent with the present disclosure are directed to efficiently and accurately detecting ADHD in a person. The systems and methods described below include techniques for presenting simple tasks to accurately detect the presence of ADHD and a certainty probability of ADHD. The invention further identifies metrics for modeling the tasks and calculating the certainty probability of ADHD. In some embodiments, the disclosed techniques include multiple tasks for presenting content with different conditions that trigger a response to the content. As described below, the specialized tasks with response conditions can result in various technical improvements in ADHD detection accuracy using underlying systems, hardware, and software, as well as other applications running on the underlying hardware and software.

[0142] FIG. 17 is a block diagram illustrating various exemplary components of an Attention Deficit Detection System (ADDS) 1700 for accurately detecting ADHD, according to some embodiments of the present disclosure. In one embodiment, the ADDS 1700 may be the system 100 shown in FIG. 13 and described above. Detecting ADHD may include evaluating a user of the ADDS 1700 to detect the presence of ADHD and a certainty rate of ADHD. The measure of the ADHD indicator may be based on the amount of agreement between the user's response and the expected outcome of the task. In some embodiments, the ADDS 1700 may measure the amount of task response deviation from a task response defined by the user of the ADDS 1700 as part of the initialization of the ADDS 1700. The user of the ADDS 1700 may define a measure of a healthy individual's attention level when configuring the ADDS 1700. For example, a user 1760 may configure the ADDS 1700 by providing a text configuration file that is parsed by the ADDS 1700.

[0143] As shown in FIG. 17 , ADDS 1700 can include a measurement module 1710 that assesses a person's attention level and a database 1730 that stores the assessed attention level in performance metrics 1734. Measurement module 1710 can assist in determining the attention level using data from database 1730. Database 1730 can assist in assessing the attention level based on a control group 1731 associated with a user 1732 through a task definition 1733. Reporting module 1720 can assess the attention level to assist in determining an attention deficit of user 1732 associated with control group 1731. ADDS 1700 can determine the attention levels of control group 1731 and user 1732 using responses 1740 provided using user device 150. User device 1750 can be a processor or a complete computing device such as a laptop, desktop computer, mobile device, smart home appliance, IoT device, etc.

[0144] The measurement module 1710 can measure the attention level of a user (e.g., user 1760) by measuring vibration signals in different regions of the user's brain. The measurement module 1710 can measure the attention level of the user 1760 by determining the level of connectivity between different regions of the user's 1760's brain. The measurement module 1710 can measure the vibration signals upon receiving a response 1740 of the user 1760 from the user device 1750. The ADDS 1700 can be configured to enable the measurement module 1710 to collect signals from different brain regions based on the response. The configuration can include whether to collect brain signals, when to collect brain signals, and from which regions to collect brain signals. A user of the ADDS 1700 can configure the measurement module 1710 using configurations provided using the user device 1750 over the network 1770.

[0145] The reporting module 1720 can evaluate the measured signals to determine the presence of ADHD and calculate a certainty percentage of ADHD. The reporting module 1720 can evaluate the signals by measuring the connectivity of the signals. In some embodiments, the reporting module 1720 can evaluate the signals by comparing the determined connectivity levels. A detailed description of various evaluation techniques is provided below in the description of Figures 19 and 20-1 through 20-4.

[0146] In various embodiments, database 1730 can take several different forms. For example, database 1730 can be a SQL database or a NoSQL database, such as databases developed by MICROSOFT®, REDIS, ORACLE®, CASSANDRA, or MYSQL, or various types of databases including data returned by web service calls, data returned by computational function calls, sensor data, IoT devices, or various other data sources. Database 1730 can store data used or generated during operation of the application, such as data generated by measurement module 1710. For example, if measurement module 1710 is configured to evaluate responses 1740, measurement module 1710 can access task definition 1733 to share stimuli 1780 and, upon receiving responses 1740, measure the signals using connectivity levels and store them as performance metrics 1734. Similarly, if reporting module 1720 is configured to determine ADHD, reporting module 1720 may access performance metrics 1734 to obtain an assessed signal connectivity level associated with a user (e.g., user 1760) for a stimulus compared to a signal connectivity level for the same stimulus provided to a control group 1731. In some embodiments, database 1730 may be supplied with data from external sources (e.g., servers, databases, sensors, IoT devices, etc.). In some embodiments, database 1730 may provide data storage for a distributed data processing system (e.g., Hadoop Distributed File System, Google File System, ClusterFS, and / or OneFS).

[0147] Network 1770 can take various forms. For example, network 1770 can include or utilize the Internet, a wired wide area network (WAN), a wired local area network (LAN), a wireless WAN (e.g., WiMAX), a wireless LAN (e.g., IEEE 802.11), a mesh network, a mobile / cellular network, a corporate or private data network, a storage area network, a virtual private network using a public network, or other types of network communication. In some embodiments, network 1770 can include an on-premises (e.g., LAN) network, while in other embodiments, network 1770 can include a virtualized (e.g., AWS®, Azure®, IBM Cloud®, etc.) network. Furthermore, in some embodiments, network 1770 can be a hybrid on-premises and virtualized network that includes components of both on-premises and virtualized types of network architectures.

[0148] The ADDS 1700 can provide stimuli 1780 to a user 1760 of the user device 1750 by presenting content that matches the conditions of the task definition 1733. The user 1760 can provide a response 1740 to the stimuli 1780 using a device attached to the user device 1750. For example, the user 1760 of the user device 1750 can respond to a task presented to the user 1760 by clicking a button or touching a screen to respond to the stimuli 1780. The response 1740 to the stimuli associated with a task can include sharing a response to the task. For example, a task can include a question that provides a list of response options from which the user 1760 can select in response to the question. In some embodiments, the response 1740 can include a user's non-response to the task. Non-response can be recorded based on the lack of a response to the stimuli after a certain period of time. The period for determining non-response to the stimuli can be configurable by a user (e.g., user 1760) of the ADDS 1700.

[0149] The task definition 1733 includes response rules that can be pre-presented to the user 1760 of the user device 150 to respond to stimuli 1780 based on the task rules of the task definition 1733. The rules can include when to respond and how to respond. In some embodiments, the rules can include multiple conditional statements that divide and subdivide response types and response times based on content that satisfies the task definition. In some embodiments, a task can include rules that extend the rules of another task.

[0150] 18A and 18B are tables of an exemplary task for detecting ADHD and its probability, according to some embodiments of the present disclosure. The table includes various response conditions, shown as row and column headers 1810-1840 and 1860-1890. As shown in FIG. 18A, row headers 1810 and 1820 divide possible content that may be presented to a user (e.g., user 1760 in FIG. 17 ) at a user device (e.g., user device 1750 in FIG. 17 ). The divided content indicates when the user is expected to respond to the content presented as stimuli 1780 at user device 1750. For example, task 1800 divides the content into consonants and vowels, with the rule that no response is expected when a consonant is presented as a stimulus. The task may include additional conditions to constrain the response to the stimuli (e.g., stimulus 1780 in FIG. 17 ) presented to the user. In task 1800, additional conditions may be included as column headers (e.g., column headers 1830, 1840). When a first condition for a response is met, additional conditions may impose restrictions. For example, as shown in Figure 18A, task 1800 may include a second condition for the color of the letter, restricting a response to a vowel when the vowel is red.

[0151] A task can include multiple sub-conditions that add different restrictions or permissions. For example, Figure 18B shows task 1850 in tabular form, where one condition can include a condition that determines whether the vowel is the letter "E" or some other vowel, and a second condition can include a condition that determines whether the color of the character is "red" or "some other color."

[0152] A response is a user (e.g., user 1760) action on a user device (e.g., user device 1750). Responses can include positive responses (e.g., clicking a button) and negative responses (e.g., clicking another button). In some embodiments, a negative response may be no response. For example, in task 1800, when a consonant is displayed (rule 1810), user 1760 is expected to have a negative response by not responding to the displayed character. Task conditions can divide possible regions of content into active and silent states. The active and silent states of a condition indicate when user 1760 is expected to respond or not respond to an instance from a region of content presented as stimulus 1780. For example, as shown in FIG. 18A , task 1800 includes a region of characters as content, divided into consonants associated with negative responses and vowels associated with positive responses. In some embodiments, a task can include restrictions on the conditions to further restrict the expected active states. For example, in task 1800, condition 1820 may have a restriction that a displayed vowel is not to react when displayed in red. In some embodiments, a second condition may impose a restriction on content belonging to both the active and silent states.

[0153] FIG. 18B illustrates a table for task 1850 with an additional condition for removing the restriction on the active state content. As shown in FIG. 18B, rules 1860 and 1870, similar to rules 1810 and 1820 of the first condition of task 1800, include the active and silent states of the first condition. Similar to task 1800, task 1850 includes a second condition with rules 1880 and 1890 for adding a restriction when the first condition matches the active state. A third condition, in addition to the second condition, can split the active state content. The third condition splits the active state content to remove the restriction added by rules 1880 and 1890 of the second condition. For example, as shown in FIG. 18B, task 1850 includes a third condition that splits the active state vowel content into the letter "E" and other vowels. When the third condition is affirmative, the task can remove the restriction on the response added to the active state content by the second condition. For example, in task 1850, second conditional rule 1880 imposes the restriction of not responding when the vowel is "red," but third conditional rule 1875 removes this restriction for the letter "E" by allowing a response.

[0154] A user of ADDS 1700 (e.g., user 1760) can be presented with content belonging to a domain of content that satisfies the first condition, second condition, or third condition defined in response tasks 1800, 1850, along with rules for expected responses. ADDS 1700 can then collect the response (e.g., response 1740 in FIG. 17 ) and signals generated in the brain of user 1760 upon presentation of response 1740. The brain signals collected upon receiving user 1760's response 1740 to the display of stimuli 1780 are evaluated to determine a level of connectivity between the regions of user 1760's brain from which the brain signals were collected. A detailed description of different brain regions and the determination of connectivity levels is provided in the description of FIG. 19 below.

[0155] Figure 3 shows the conditional probability of each output based on the cascade model according to some embodiments of the present disclosure. The conditional probability and information transmission are calculated using the transmission information content (TIC) metric shown below. For example, using the following TIC metric, in task 1800, when consonant data is observed by a user (e.g., user 1760 in FIG. 1), the TIC metric calculates the amount of information transmission as follows using the TIC metric defined below.

Number

[0156] In the scenario, I(s) << I(s|a) ⇒ TIC(s→a) < 0, that is, there is negative information transmission from stimulus 1780, including the display of consonants as part of task 1800 that causes a reaction. The negative information bit is an indicator that the user 1760 is suppressing or restricting himself from showing a positive reaction (e.g., clicking a button). As a result, since I(a|s) is much larger than I(a), the button is not pressed.

[0157] When vowel data is presented to user 1760 as part of task 1800, user 1760 needs to check whether the color of the character is red (e.g., a red vowel). For vowels other than red, the information transmission is calculated as follows using the conditional TIC metric.

Number

[0158] When red vowel data is presented to user 1760 as part of task 1800, user 1760 needs to check whether the character is red. Applying the conditional TIC to the red context gives the following equation.

Number

[0159] In this case, component I(c|a,s) generates negative information bits that counteract the information generated by the evidence, which represents the subject's top-down control. To derive the value of the probability P(c|a,s), we can infer the probability that a vowel appears and a user (e.g., user 1760 in Figure 17) presses a button. The answer is that this probability is very low, and therefore the associated information content is very high: a large number of negative information bits are being generated to counteract the positive bits generated by evidence I(c|s).

[0160] Thus, in this context, there is a negative transfer of information from the evidence of the appearance of the red vowel to the action a of pressing the button.

[0161] When the red vowel data is presented to the user 1760 as part of task 1850, the user 1760 must identify whether the letter is "E." Information transfer is calculated using the episodic TIC metric as follows:

number

[0162] The aforementioned amount of information bits can be used as default values ​​for comparing different users to determine their attention levels. The ADDS1700 starts with these default values ​​as expected signal values ​​and can modify them as signals from more people considered part of the control group are added.

[0163] 19 illustrates a connectivity network between brain regions, according to some embodiments of the present disclosure. The ADDS 1700 can identify the presence of signals in various brain regions representing different connectivity networks when a user (e.g., user 1760 of FIG. 17 ) exhibits response 1740. The ADDS 1700 can use the signals identified in the connectivity networks to determine a level of connectivity between different brain regions. The ADDS 1700 evaluates the level of connectivity between various brain regions to determine an attention level and, based on the determined attention level, determines whether the user is indicative of having ADHD.

[0164] The ADDS 1700 can determine and use signals of various frequencies to determine attention levels and deficits. When evaluating signals and connectivity using the signals, the ADDS 1700 can group the signals by frequency spectrum to determine attention levels and deficits. The ADDS 1700 groups signals in the 8-12 Hz frequency range as an alpha group, signals in the 20-30 Hz frequency range as a beta group, and signals in the 4-8 Hz frequency range as a theta group. In some embodiments, the ADDS 1700 can collect signals of different frequencies from different regions of the brain at different times.

[0165] Each signal group indicates a particular aspect of the task stimulus presentation and response behavior. For example, oscillation amplitude in the alpha band can indicate suppression of activity. In some embodiments, the characteristics (e.g., amplitude) of the signal groups can have multiple meanings. For example, the same oscillation amplitude in the alpha band can also indicate active processing, which is considered task-relevant information. Changes in the signals or characteristics, such as the amplitude of the signals, can be considered differences in the cognitive activity of the user's 1760 brain that generated the signals.

[0166] In some embodiments, decreased connectivity between brain regions indicates increased attention. The increased connectivity is based on the tasks used to activate different regions of the user's 1760 brain.

[0167] As shown in FIG. 19 , disclosed embodiments of the ADDS 1700 assess communication between various known regions of the brain used to assess attention (e.g., regions within the dorsal attention network (DAN) 1910 or the ventral attention network (VAN) 1920). Using the tasks defined in FIGS. 18A and 18B , the disclosed embodiments of the ADDS 1700 can generate positive and negative information flows occurring between the DAN 1910 and the VAN 1920 and identify the pathways involved in this information flow. The ADDS 1700 can also identify alternative pathways and connectivity networks that may include signals when different individuals respond to the same stimuli. For example, the ADDS 1700 can detect signals in the dorsal-ventral attention network (DVAN) 1930.

[0168] The ADDS 1700 determines whether a user (e.g., user 1760 in FIG. 17 ) has ADHD by determining connectivity levels between brain regions generated in the user's brain and comparing them to connectivity levels in a group of known healthy individuals. Disclosed embodiments of the ADDS 1700 provide stimuli using tasks as a way to trigger activity in brain networks, generate signals through the transfer of information (positive and negative bits), and use the connectivity of the generated signals to screen for ADHD in a user (e.g., user 1760 in FIG. 17 ). For example, the ADDS 1700 can activate and screen for the presence of signals in the DVAN 1930 by using tasks 1800, 1850 and presenting instances of the content regions defined by the tasks 1800, 1850 for the user's response. Use of the DVAN 1930 is screened by comparing different types of measured signals assessed in the user 1760 with a previously identified set of healthy users without ADHD (e.g., control group 1731). A detailed description of the comparative analysis of signal groups between different users is provided in the following description of Figures 20-1 to 20-4.

[0169] In some embodiments, attentional deficits can also be determined by comparing connectivity fluctuations to a switched task with a different rule. In such a scenario, instances of content regions with opposite responses defined in the task are selected. For example, a red "E" presented as part of task 1800 results in a no-go condition with its active state constrained by a negative second condition. The same red "E" presented as part of task 1850 results in a constrained second condition, which is lifted by a third condition. A detailed description of attentional network analysis in a task-switching scenario is provided in detail in the description of Figure 21 below.

[0170] The disclosed embodiments of the ADDS 1700 can help model the probability of ADHD level based on the presented behavioral tasks. The probability of ADHD level can help accurately predict the presence of ADHD and the probability of ADHD. The model for determining the probability of ADHD level can be implemented using the example tasks 1800, 1850 shown in the description of FIG. 3 above.

[0171] Bayesian brain theory establishes that the entire cortex represents a probability distribution, and this probability distribution is transformed into an estimate only when a decision is required. Under Bayes' theorem, to transform a prior distribution P(θ) into a posterior distribution P(θ|x), the effect of data x as stimulus 1780 on user 1760 can be expressed as follows: P(θ|x)=P(x|θ)P(θ) / P(x) where P(x) is known as the evidence in Bayes' theorem formula and P(x|θ) is known as the likelihood.

[0172] The same equation expressed using the -log function, used to represent the amount of information transmitted in the brain when data x is presented as a stimulus 1780, is: -log2[P(θ|x)]=-log2[P(x|θ)P(θ) / P(x)] -log2[P(θ|x)]=-log2P(x|θ)-log2P(θ)+log2P(x)

[0173] The information content of an event x_j (measured in bits), also known as its self-information, is given by: I(x i )=-log2[p(x i )]bits

[0174] Thus, the logarithmic equation expressed in terms of information bits becomes: I(θ|x)(bits)=I(θ)(bits)-I(x)(bits)+I(x|θ)(bits) An aspect of the above equation is that the aforementioned change in the amount of information of a single model θ in the space of models that induce a single observation x occurring upon presentation of data x does not result in a complication, but rather in two amounts of information I(x) and I(x|θ) generated by the single observation. Thus, in the above equation, -I(x) (bits) + I(x|θ) (bits) represents the information conveyed by observation x to a single model θ. ADDS1700 calculates this information transfer using a new metric called Transferred Information (TIC). TIC is defined as follows: TIC(x→θ)(bits)=I(x)-I(x|θ) We can also interpret I(θ) as the prior uncertainty about the model θ, and I(θ|x) as the posterior uncertainty about the model θ after a single observation x.

[0175] When the TIC metric is applied to the information content formula, it is converted as follows: I(θ|x)(bits)=I(θ)(bits)-TIC(x→θ)(bits)

[0176] The ADDS 1700 uses the TIC metric to account for information flow. If the negative bits of the Bayesian likelihood I(θ|x) do not cancel out the positive bits of the evidence I(x), the TIC metric is positive, i.e., there is a positive net information transfer. Conversely, if the negative bits corresponding to I(θ|x) are greater than the bits of the evidence I(x), the TIC is negative, i.e., there is a negative net information transfer. A detailed explanation of using this bit gain or loss in the form of signaling between two regions to detect ADHD is provided in detail in the description of Figures 20-1 through 20-4 below.

[0177] Using the TIC metric described above, the amount of information represented using stimuli (e.g., stimulus 1780 in FIG. 17) and actions (e.g., response 1740 in FIG. 17) translates to: I(a|s)=I(a)-TIC(s→a)(bits) I(a) can also be interpreted as the prior uncertainty about action a, and I(a|s) can also be interpreted as the posterior uncertainty about action a that a user (e.g., user 1760 in FIG. 17) will take after the occurrence of stimulus s.

[0178] The amount of information conveyed that will or will not trigger an action a when a stimulus s occurs in a context c defined in tasks 1800 and 1850 using different conditions is as follows:

number

[0179] The ADDS 1700 can use conditional TICs to explain the information bits to be conveyed when additional conditions are presented to the user 1760, such as the second condition of the task 1800 defined using rules 1830, 1840.

[0180] In some embodiments, ADDS 1700 can calculate a TIC metric of the information transfer that can cause behavior a, given the occurrence of stimulus s in context c and episode (temporal context) e. I(a|s,c,e)=-log2[P(a|s,c,e)]

[0181] As in the context case, ADDS1700 can then apply a conditional version of Bayes' theorem to the posterior probabilities as follows:

number

number

[0182] Next, the sensorimotor TIC, the context-conditioned TIC, and the episode-conditioned TIC can be decomposed into two components. I(a|s,c,e)=I(a)-TIC(s→a)-TIC(c→a|s)-[I(e|s,c)-I(e|a,s,c)]=I(a)-I(s)+I(s|a)-I(c|s)+I(c|a,s)-I(e|s,c)+I(e|a,s,c)

[0183] The components I(e|s,c) and I(e|a,s,c) of the episode-conditional TIC represent bottom-up and top-down contributions, respectively, in action selection. Following the reasoning in the context case, it is possible that an uninformative stimulus in a given context can become informative if the episode-conditional TIC is positive and compensates for the negativity of the context-conditional TIC. ADDS 1700 can also interpret I(a) as the prior uncertainty about action a, and I(a|s,c,e) as the posterior uncertainty about action a after stimulus s occurs in context c and episode e.

[0184] ADDS 1700 can use conditional TICs to describe the information bits conveyed when additional conditions are presented to user 1760, such as the second condition of task 1850 defined using rules 1880, 1890. A detailed description of the conditional probabilities and information bits conveyed when presenting various stimuli to a user (e.g., user 1760 of FIG. 1) is provided in the description of FIG. 3 above.

[0185] 20-1 through 20-4 are flowcharts illustrating the operation of an exemplary method for attention deficit detection, according to some embodiments of the present disclosure. For illustrative purposes, the steps of method 2000 may be performed by ADDS 1700. It will be understood that the illustrated method 2000 may be modified to modify the order of steps and to include additional steps.

[0186] In step 2010, ADDS 1700 can provide a stimulus to a user (e.g., user 1760 of FIG. 17 ) that elicits a response. ADDS 1700 can provide the stimulus by presenting a task (e.g., tasks 1800, 1850 of FIGS. 18A, 18B ) to a user device (e.g., user device 1750 of FIG. 17 ). ADDS 1700 can present the task and elicit a response by displaying the rules of the task and then displaying content to consider. User 1760 of user device 1750 can respond to the provided stimulus, which is an instance of possible content based on the rules of the task. ADDS 1700 can continuously provide the stimulus by presenting content that matches the task. In some embodiments, ADDS 1700 can loop through different tasks to provide stimuli to user 1760 of ADDS 1700.

[0187] In step 2020, the ADDS 1700 can acquire signals about the presented stimuli from sensors. The sensors can be electroencephalography (EEG) or magnetoencephalography (MEG) devices, or probes for other functional neuroimaging techniques for mapping brain activity, such as functional magnetic resonance imaging (fMRI) and functional near-infrared spectroscopy (fNIRS). The sensors are placed on the head to correspond to different known regions (e.g., brain regions 1910-1930 in FIG. 19 ). The ADDS 1700 can acquire signals based on a specific trigger event. In some embodiments, the trigger event can include a time period. The time period can include a waiting time after the stimulus is provided in step 2010. The ADDS 1700 can enable configuration of the trigger event by selecting the trigger event and acquiring signals and parameters for the trigger event. For example, the ADDS 1700 can enable a waiting time after the stimulus is provided before acquiring signals from different brain regions. The EEG / MEG device can be a standard EEG device or an MEG device. The EEG / MEG device can measure multiple signals for respective source locations located within the cranial cavity of the user 1760. Each source location can correspond to a known brain region, such as a lobe, a lobule, a sulcus, or a gyrus. In some examples, the multiple signals can be measured while the subject performs one or more tasks, which can be presented to the user 1760 by the ADDS 1700. In one or more embodiments, the signals can correspond to measurements of electrodes of an EEG device or measurements of sensor coils of an MEG device. In some embodiments, the signals can correspond to an aggregate signal (such as an average signal) across measurements of multiple electrodes (when using EEG) or multiple sensor coils (when using MEG). Such an aggregate signal can correspond to a brain region that includes the respective locations of the signals aggregated in the aggregate signal.

[0188] MEG signals can be measured using a 306-channel (102 magnetometers and 204 planar gradiometers) whole-head MEG Elekta Neuro imaging system. In some embodiments, brain signals can be measured using an online anti-aliasing bandpass filter of 0.1 to 330 Hz and a sampling rate of 1000 Hz. In some embodiments, in addition to three reference points (the nasion and the left and right preauricular points) as landmarks, the head shape of the user (e.g., user 1760 in FIG. 17 ) can be acquired using a 3D Fastrak digitizer (Polhemus, Colchester, Vermont). In another embodiment, four head position indicator (HPI) coils can be recorded on the scalp (forehead and mastoid) of user 1760. In some embodiments, two pairs of bipolar electrodes can be used to record eyeblinks and heart rate, respectively.

[0189] In step 2030, ADDS 1700 can evaluate the signal for each brain region from the brain signals acquired in step 2020. ADDS 1700 can evaluate signal quality, remove artifacts from the signals, and reconstruct brain activity using signals acquired from each sensor in the user's head that approximately represent each brain region. ADDS 1700 can refine the measured EEG / MEG data by removing external noise, such as ocular, cardiac, and muscle artifacts, and by segmenting the EEG / MEG data so that time series have the same length (e.g., 1 second).

[0190] In step 2031, ADDS 1700 may filter the signals extracted in step 2030 into multiple frequency bands, for example, using a pass filter. The frequency bands may be functionally related signal bands. As noted above in the description of FIG. 19 , the signals may be filtered into different frequencies to indicate different attentional activities of the user of ADDS 1700. For example, ADDS 1700 may extract signals into theta (4-8 Hz), alpha (8-12 Hz), low beta (12-20 Hz), high beta (20-30 Hz), and low gamma (30-45 Hz) frequency bands. In some embodiments, the frequency bands representing theta, alpha, low beta, high beta, and low gamma bands may be different. In some embodiments, additional signal groups may exist between the exemplary frequency bands. ADDS 1700 may allow configuration of the frequency bands that are part of the signal groups. ADDS 1700 extracts signals for each frequency band by filtering out signals that do not fall into one of the selected frequency bands.

[0191] In step 2032, the ADDS 1700 can determine the level of connectivity between two brain regions based on a measure of dependency between the signals of the two brain regions for each frequency band. For example, the ADDS 1700 can measure phase synchronization between signals grouped by frequency band to calculate a connectivity value. Measures of phase synchronization include the phase lag index or phase lock value (PLV) and its derivative, the imaginary part of the PLV (iPLV), and the corrected imaginary part of the PLV (ciPLV). The ADDS 1700 can use other different connectivity measures, including information-theoretic measures such as mutual information or transfer entropy, measures of Granger causality, and traditional measures such as correlation, cross-correlation, coherence, or phase gradient index. The ADDS 1700 can measure the level of connectivity between two brain regions based on a metric derived from the TIC. In some embodiments, the dependency measure can be based on connectivity between brain regions across different frequency bands.

[0192] In step 2033, ADDS 1700 may filter the theta band signal to determine if the signal type matches the theta band signal frequency. If the answer is no, proceed to step 2038.

[0193] In step 2034, ADDS 1700 may check whether the first condition of the second task is met. If the answer is yes, proceed to step 2035. If the answer is no, proceed to step 2056.

[0194] In step 2035, ADDS 1700 may check whether the second condition of the second task is met. If the answer is "no," it proceeds to step 2037.

[0195] In step 2036, ADDS 1700 may check whether the third condition of the second task is met. If the answer is "yes," it proceeds to step 2037. If the answer is "no," it proceeds to step 2056.

[0196] In step 2037, ADDS 1700 may determine whether the user's connectivity level calculated in step 2032 is lower than the connectivity level of a healthy control group. If the answer to the question is "YES," the process proceeds to step 2070. ADDS 1700 may obtain the connectivity level of a healthy control group stored in population database 1730. ADDS 1700 receives the control group type of user as input and stores the result in control group 1731. ADDS 1700 may compare the connectivity level by determining whether the calculated connectivity level is lower than the range of possible connectivity levels for a healthy control group. ADDS 1700 may determine a range of values ​​for the connectivity level of a healthy control group based on the connectivity level of the person identified as having ADHD as assessed by method 2000.

[0197] In step 2038, ADDS 1700 may filter the beta band signal to determine if the signal type matches the beta band signal frequency. If the answer is "NO," proceed to step 2049.

[0198] In step 2039, ADDS 1700 may check to see if the stimulus is from the first task. If the answer is "no," it proceeds to step 2044.

[0199] In step 2040, ADDS 1700 may check whether a first condition of a first task is satisfied. If the answer is "no," it proceeds to step 2048.

[0200] In step 2041, ADDS 1700 may check whether the second condition of the first task is met. If the answer is "yes," it proceeds to step 2048. If the answer is "no," it proceeds to step 2056.

[0201] In step 2044, ADDS 1700 may check whether the first condition of the second task is satisfied. If the answer is "yes," it proceeds to step 2045. If the answer is "no," it proceeds to step 2056.

[0202] In step 2045, ADDS 1700 may check whether the second condition of the second task is satisfied. If the answer is "yes," it proceeds to step 2046. If the answer is "no," it proceeds to step 2047.

[0203] In step 2046, ADDS 1700 may check whether the third condition of the second task is met. If the answer is "yes," it proceeds to step 2048. If the answer is "no," it proceeds to step 2056.

[0204] In step 2047, ADDS 1700 may check whether the third condition of the second task is met. If the answer is "no," it proceeds to step 2048. If the answer is "yes," it proceeds to step 2056.

[0205] In step 2048, ADDS 1700 may determine whether the user's connectivity level calculated in step 2032 is higher than the connectivity level of a healthy control group. If the answer to the question is "YES," the process proceeds to step 2070. ADDS 1700 may obtain the connectivity level of a healthy control group stored in population database 1730. ADDS 1700 may receive as input a control group type of user and store the result in control group 1731. ADDS 1700 may compare the connectivity levels by determining whether the calculated connectivity level is lower than the range of possible connectivity levels for a healthy control group. ADDS 1700 may determine a range of values ​​for the connectivity level of a healthy control group based on the connectivity level of the person identified as having ADHD as assessed by method 2000.

[0206] In step 2049, ADDS 1700 can filter the alpha band signal to determine if the signal type matches the alpha band signal frequency. If the answer is "YES," proceed to step 2050. If the answer is "NO," proceed to step 2056.

[0207] In step 2050, ADDS 1700 may check to see if the stimulus is from the first task. If the answer is "no," it proceeds to step 2054.

[0208] In step 2051, ADDS 1700 may check whether a first condition of a first task is met. If the answer is "no," it proceeds to step 2051.2. Otherwise, it proceeds to step 2052.

[0209] In step 2051.2, if the connectivity level of user 1760 calculated in step 2032 has been saved in container A1, proceed to step 2055. Container A1 may be a data structure such as a list or an array. In some embodiments, container A1 may be persisted to disk in a flat file or database. In step 2052, ADDS 1700 may check whether the second condition of the first task is met. If the answer is "NO," proceed to step 2052.2. Otherwise, proceed to step 2052.3.

[0210] In step 2052.2, if the user's connectivity level calculated in step 2032 is saved in container A2, proceed to step 2055. Container A2 may be a data structure such as a list or array. In some embodiments, container A2 may be persisted to disk in a flat file or database. Container A2 contains values ​​of alpha band connectivity between brain regions when the first condition of the first task is met and the second condition is not met.

[0211] In step 2052.3, the connectivity level of the user 1760 calculated in step 2032 is saved in container A1.

[0212] In step 2053, ADDS 1700 may check whether the third condition of the first task is met, i.e., whether the vowel "E" is red. If the answer to the question is "YES," proceed to step 2053.2. Otherwise, proceed to step 2055.

[0213] In step 2053.2, if the user's connectivity level calculated in step 2032 is saved in container A3, proceed to step 2055. Container A3 may be a data structure such as a list or array. In some embodiments, container A3 may be persisted to disk in a flat file or database. Container A3 contains the alpha band connectivity values ​​between brain regions when the first condition, second condition, and third condition of the first task are met, i.e., the red vowel "E."

[0214] In step 2054, ADDS 1700 may check whether the first condition, second condition, and third condition of the second task are met. If the answer is "yes," proceed to step 2054.2. Otherwise, proceed to step 2056.

[0215] In step 2054.2, if the user's connectivity level calculated in step 2032 is saved in container A4, proceed to step 2055. Container A4 may be a data structure such as a list or array. In some embodiments, container A4 may be persisted to disk in a flat file or database. Container A4 contains the alpha band connectivity values ​​between brain regions when the first, second, and third conditions of the second task are met, i.e., the red vowel "E."

[0216] In step 2055, ADDS 1700 may determine whether the user's connectivity level calculated in step 2032 is lower than the connectivity level of a healthy control group. If the answer to the question is "YES," the process proceeds to step 2070. ADDS 1700 may obtain the connectivity level of a healthy control group stored in population database 1730. ADDS 1700 may receive as input a control group type of user and store the result in control group 1731. ADDS 1700 may compare the connectivity levels by determining whether the calculated connectivity level is lower than the range of possible connectivity levels for a healthy control group. The range of values ​​to include in the connectivity level of a healthy control group may be determined based on the connectivity levels of individuals identified as having ADHD.

[0217] In step 2056, ADDS 1700 may check whether all connectivity values ​​calculated in step 2032 have been verified for ADHD identification by performing steps 2033 through 2055. If the answer is "YES," it proceeds to step 2057. If the answer is "NO," it proceeds to step 2033 to verify the next connectivity value.

[0218] In step 2057, ADDS 1700 may obtain the connectivity level values ​​for containers A1 and A2 from steps 2051.2, 2052.2, and 2052.3. ADDS 1700 may compare the change in connectivity levels. If the decrease in connectivity level is less than the decrease in connectivity level of the healthy control group, proceed to step 2070.

[0219] In step 2058, ADDS 1700 may obtain connectivity level values ​​for containers A3 and A4 from steps 2053.2 and 2054.2, respectively. ADDS 1700 may compare changes in connectivity levels. If the fluctuations in connectivity levels between the user's brain regions are different from the decreased connectivity levels between the same regions in healthy controls, ADDS 1700 proceeds to step 2070.

[0220] In some embodiments, ADDS 1700 can compare the variability in connectivity levels between two regions based on the signal-based connectivity levels determined in steps 2033, 2038, and 2049 of method 2000. ADDS 1700 can calculate the probability of ADHD based on the difference in the variability in connectivity levels between a user of ADDS 1700 and a healthy control group. ADDS 1700 can collect multiple connectivity levels for the same user or a set of users and store those connectivity levels as data for healthy controls 1731, or store users identified as having ADHD as users 1732. ADDS 1700 can use the stored values ​​of the connectivity levels to determine the variability in connectivity levels. In some embodiments, the variability in connectivity levels can include only connectivity levels observed for the same data under different conditions. For example, a user of ADDS 1700 can be presented with the same data and then presented with task rules to determine the connectivity levels for the same data across multiple tasks. In step 2060, ADDS 1700 may mark users (e.g., users 1760) whose brain signals are assessed as having ADHD symptoms for connectivity levels. Upon completion of step 2060, ADDS 1700 completes execution of method 2000 (step 2099).

[0221] 21 is a flowchart illustrating the operation of an exemplary method for ADHD detection, according to some embodiments of the present disclosure. For illustrative purposes, the steps of method 2100 may be performed by ADDS 1700. It will be understood that the illustrated method 2100 may be modified to modify the order of steps and to include additional steps.

[0222] In step 2110, ADDS 1700 can provide stimuli to a user (e.g., user 1760 of FIG. 17) that elicit a response consistent with a first task (e.g., task 1800 of FIG. 18A). ADDS 1700 can select a set of tasks for testing a user (e.g., user 1760 of FIG. 17) for ADHD that can satisfy the rules (e.g., rules 1820, 1830 of task 1800, rules 1875, 1880 of task 1850). ADDS 1700 can provide the stimuli by displaying content on a user device (e.g., user device 1750 of FIG. 17) that satisfies the rules of task 1800.

[0223] In step 2120, the ADDS 1700 may acquire a first set of signals from brain regions. The ADDS 1700 may acquire signals from different brain regions by collecting electrical charge from probes attached to the head of a user 1760 being tested for ADHD. The ADDS 1700 may acquire signals when the user responds to the stimuli provided in step 2110 by performing an action. Performing an action may include clicking a button or using a pointing device to select something displayed on a screen.

[0224] In step 2130, ADDS 1700 can present the same stimulus as part of a second task (e.g., task 1850 of FIG. 18B). ADDS 1700 can select a task if the rules are satisfied by the first task and the second task. ADDS 1700 can select a task from task definition 1733 only if the same content satisfies different rules that require opposite user responses. For example, when a red letter "E" is presented as a stimulus in task 1800, rules 1820 and 1830 are satisfied and the user is expected to not respond, but in task 1850, rules 1875 and 1880 are satisfied and the user is expected to respond.

[0225] In step 2140, ADDS 1700 may acquire a second set of signals from the brain region. Similar to step 2120, ADDS 1700 may acquire the signals by collecting brain images or electrical charges from the brain of user 1760 upon displaying the stimuli in step 2130. The signals may be generated as part of a response performed by user 1760.

[0226] In step 2150, the ADDS 1700 can assess the level of connectivity using the first set of signals and the second set of signals. The ADDS 1700 can assess the level of connectivity by measuring signals in certain regions of the brain (e.g., the dorsal attention network (DAN) 1910 in FIG. 19 ) and measuring signals in other regions of the brain (e.g., the ventral attention network (VAN) 1920 in FIG. 19 ) to determine the amount of transmitted signal indicative of the level of connectivity.

[0227] In step 2160, the ADDS 1700 can determine a variance in the connectivity level of the first set of signals and the second set of signals. The ADDS 1700 can determine the variance in the connectivity level by calculating the connectivity level by providing multiple stimuli as part of the first task and the second task in steps 2110, 2130, and obtaining signals of the user's 1760 response in steps 2120, 2140.

[0228] At step 2170, ADDS 1700 can compare the variance in connectivity levels to known variances in connectivity levels to identify ADHD. ADDS 1700 can compare the variance in connectivity levels calculated at step 2160 to the variance in connectivity levels of a control group (e.g., control group 1731). ADDS 1700 can obtain the variance in connectivity levels of the control group 1731 from previously generated performance metrics 1734. In some embodiments, ADDS 1700 may need to evaluate the variance in connectivity levels from previously calculated connectivity levels stored in performance metrics 1734. If the variance in connectivity levels of user 1760 differs from the variance in connectivity levels of the control group 1731, ADDS 1700 can indicate that user 1760 has ADHD. ADDS 1700 can consider the variance in connectivity levels to be different if the ranges of values ​​for the connectivity levels are different. In some embodiments, the variance in connectivity levels is considered to be different if the ranges of values ​​do not overlap. ADDS 1700 may store the attention deficit results in database 1730 under user 1732. ADDS 1700 completes execution of method 2100 upon completion of step 2170 (step 2199).

[0229] Example 1 1. A non-transitory computer-readable storage medium containing instructions that, when executed by at least one processor, cause the at least one processor to perform operations for automated detection of attention deficit hyperactivity disorder (ADHD), the operations including: providing one or more stimuli to a user for activating a plurality of brain regions representing an attention network, the one or more stimuli including displaying information having a first condition that elicits a response and a second condition as an exception to the first condition; acquiring a plurality of signals from the plurality of brain regions for each of the one or more stimuli, each signal of the plurality of signals being accessed over a period of time, the period beginning when the information is displayed and ending when a user response is captured; and evaluating the plurality of signals based on a level of connectivity within the attention network when each signal of the plurality of signals corresponds to the one or more stimuli to detect ADHD.

[0230] Example 2 10. The non-transitory computer-readable storage medium of Example 1, wherein displaying information having a first condition that triggers a response and a second condition as an exception to the first condition further includes presenting instances of data that match a first task, the first task including a first condition related to prediction of an active state and a silent state of the first response, and the second condition limiting the prediction of the active state of the first response.

[0231] Example 3 10. The non-transitory computer-readable storage medium of Example 2, further comprising: displaying on a screen information that the first reaction satisfies a first condition regarding a prediction of an active state and a silent state of the first reaction; and waiting a threshold period for any reaction shared by the user.

[0232] Example 4 3. The non-transitory computer-readable storage medium of example 2, wherein the first response is at least one of clicking a pointing device, pressing a button, or taking no action for a time threshold.

[0233] Example 5 and, when a first condition of the first task satisfies an active state and a second condition is negative, comparing the connectivity level between the two regions of the user's brain with the connectivity level of a control group of healthy users, wherein a connectivity level of the user is lower than the connectivity level of the control group of healthy users, indicating that the user has ADHD.

[0234] Example 6 and, when a first condition of the first task satisfies a silent state and a second condition is positive, comparing the connectivity level between the two regions of the user's brain with the connectivity level of a control group of healthy users; wherein a connectivity level of the user is lower than the connectivity level of the control group of healthy users, indicating that the user has ADHD.

[0235] Example 7 10. The non-transitory computer-readable storage medium of Example 2, wherein the operations further include filtering the beta band oscillation signals of the two or more brain regions when the first response is received, and using the beta band oscillation signals to determine connectivity between the two brain regions; and comparing the connectivity level between the two brain regions of the user with the connectivity level of a control group of healthy users when a first condition of the first task satisfies a silent state and a second condition is positive, wherein a higher connectivity level of the user than the connectivity level of the beta band oscillation signals of the control group of healthy users indicates that the user has ADHD.

[0236] Example 8 10. The non-transitory computer-readable storage medium of example 2, wherein the operations further include filtering the alpha band oscillatory signals of the two or more brain regions when the first response is received and using the alpha band oscillatory signals to determine connectivity between the two regions of the brain; determining a connectivity level between the two regions of the user's brain when a first condition of the first task satisfies a silent state and when the first condition of the first task satisfies an active state; and comparing a decrease in the user's connectivity level between when the first task satisfies the silent state and when the first task satisfies the active state with a decrease in the connectivity level of a control group of healthy users between when the first task satisfies the silent state and when the first task satisfies the active state, wherein a decrease in the user's connectivity level less than the decrease in the connectivity level of the control group of healthy users indicates the user has ADHD.

[0237] Example 9 2. The non-transitory computer-readable storage medium of Example 1, wherein evaluating the plurality of signals based on the connectivity level within the attention network to detect ADHD further comprises combining signal data for each of the plurality of signals, wherein combining the signal data comprises determining an increase in connectivity level compared to an existing connectivity level between regions of the plurality of regions or a decrease in connectivity level from an existing connectivity level between regions of the plurality of regions.

[0238] Example 10 10. The non-transitory computer-readable storage medium of example 1, wherein displaying the information having the first condition includes displaying a text or graphic category.

[0239] Example 11 11. The non-transitory computer-readable storage medium of example 10, wherein the second condition includes displaying a category of text or graphics in a particular color.

[0240] Example 12 10. The non-transitory computer-readable storage medium of Example 1, further comprising: displaying second information having a first condition under which the one or more stimuli cause the response, a second condition as an exception to the first condition, and a third condition as an exception to the second condition.

[0241] Example 13 13. The non-transitory computer-readable storage medium of Example 12, wherein displaying the second information having a first condition that causes a response, a second condition as an exception to the first condition, and a third condition as an exception to the second condition further comprises presenting an instance of data that matches a second task, wherein the second task includes a first condition related to predicting an active state and a silent state of the second response, the second condition restricting the prediction of an active state of the second response, and the third condition is an exception to the second condition related to predicting a silent state of the second response.

[0242] Example 14 14. The non-transitory computer-readable storage medium of Example 13, further comprising: displaying on a screen information that the second response satisfies a first condition regarding a prediction of an active state and a silent state of the second response; and waiting for a threshold period of time for any input shared by the user.

[0243] Example 15 13. The non-transitory computer-readable storage medium of Example 13, wherein the third condition includes selecting a subcategory of the text or graphic category that satisfies the first condition and displaying the text or graphic subcategory.

[0244] Example 16 14. The non-transitory computer-readable storage medium of Example 13, wherein the second response is at least one of clicking a pointing device, pressing a button, or taking no action for a threshold time.

[0245] Example 17 14. The non-transitory computer-readable storage medium of Example 13, wherein the operations further include filtering the beta band oscillation signals of two or more of the user's multiple brain regions when the second response is received, and using the beta band oscillation signals to determine connectivity between the two brain regions; and comparing the connectivity level between the two brain regions of the user with the connectivity level of a control group of healthy users when the first condition of the second task satisfies an active state, the second condition is negative, and the third condition is negative, or when the first condition of the second task satisfies an active state, the second condition is positive, and the third condition is positive, wherein a higher connectivity level of the user than the connectivity level of the control group of healthy users indicates that the user has ADHD.

[0246] Example 18 14. The non-transitory computer-readable storage medium of Example 13, wherein the operations further include, when receiving a second response from the user, filtering the theta band oscillation signals of two or more of the user's multiple brain regions and using the theta band oscillation signals to determine connectivity between two regions of the brain; and comparing the connectivity level between the two regions of the user's brain with the connectivity level of a control group of healthy users when a first condition of the second task satisfies an active state, the second condition is negative, and the third condition is negative, or when the first condition of the second task satisfies an active state, the second condition is positive, and the third condition is positive, wherein a signal data of the user's connectivity level indicates that the user has ADHD if it is lower than the connectivity level of the control group of healthy users.

[0247] Example 19 14. The non-transitory computer-readable storage medium of Example 13, wherein the operations further include, when receiving a second response from the user, filtering the alpha band oscillatory signals of two or more of the user's multiple brain regions and using the alpha band oscillatory signals to determine connectivity between two regions of the brain; determining a variation in the level of connectivity between the two regions of the user's brain when presenting an instance of data of a second task where a first condition of the second task satisfies an active state, a second condition of the second task is positive, and a third condition of the second task is positive relative to a level of connectivity between the two regions of the user's brain when presenting an instance of data of the first task where the first condition of the first task satisfies an active state and the second condition of the first task is positive; and comparing the variation in the user's connectivity level with the variation in the connectivity level of a control group of healthy users, wherein a difference between the variation in the user's connectivity level and the variation in the connectivity level of the control group of healthy users indicates that the user has ADHD.

[0248] Example 20 14. The non-transitory computer-readable storage medium of Example 13, wherein upon receiving the first response and the second response, filtering alpha band oscillatory signals, beta band oscillatory signals, and theta band oscillatory signals of two or more of the user's multiple brain regions; determining connectivity between the brain regions using the alpha band oscillatory signals, beta band oscillatory signals, and theta band oscillatory signals; and assessing the probability of detecting ADHD based on a comparison of the connectivity levels, a decrease in the connectivity levels, and a variation in the connectivity levels between the user's brain regions relative to the connectivity levels of a control group of healthy users when a first condition, a second condition of the first task, and a first condition, a second condition, and a third condition of the second task are changed.

[0249] Example 21 1. A computer-implemented method for automated detection of attention deficit hyperactivity disorder (ADHD), comprising: providing one or more stimuli to a user to activate a plurality of brain regions representing an attention network, the one or more stimuli including displaying information having a first condition that elicits a response and a second condition as an exception to the first condition; acquiring a plurality of signals from a plurality of brain regions for each of the one or more stimuli, each signal of the plurality of signals being accessed over a period of time, the period beginning when the information is displayed and ending when a user response is captured; and evaluating the plurality of signals based on a connectivity level of the attention network when each signal of the plurality of signals corresponds to the one or more stimuli to detect ADHD.

[0250] Example 22 1. An attention deficit detection system comprising: one or more memory devices storing processor-executable instructions; and one or more processors configured to execute the instructions to cause the attention deficit detection system to perform operations, the operations including: providing one or more stimuli to activate a plurality of brain regions representing an attention network, the one or more stimuli including displaying information having a first condition that elicits a response and a second condition as an exception to the first condition; acquiring a plurality of signals from the plurality of brain regions for each of the one or more stimuli, each signal of the plurality of signals being accessed over a period of time, the period beginning when the information is displayed and ending when a user response is captured; and evaluating the plurality of signals based on a connectivity level of the attention network when one signal of the plurality of signals corresponds to the one or more stimuli to detect ADHD.

[0251] Example 23 1. An attention deficit detection system comprising: one or more memory devices storing processor-executable instructions; and one or more processors configured to execute the instructions to cause the attention deficit detection system to perform operations, the operations including: receiving first signal data from a user; determining a first plurality of time series based on the first signal data, each of the first plurality of time series corresponding to a respective source location located within the user's cranial cavity; calculating first correlation values ​​for pairs of the first time series, each pair of the first time series being included in the determined first plurality of time series; generating a score based on the first correlation values, the score indicating that the patient has attention deficit or a cognitive disorder such as ADHD; and outputting the generated score.

[0252] [Table 4-1]

Table 4-2

Table 4-3

Claims

1. An apparatus (140, 200) comprising: the device (140, 200) is configured to receive (410) first electroencephalogram or magnetoencephalogram data (first EEG / MEG data) (130) of a subject (110); the device (140, 200) is configured to determine (420) a first plurality of time series based on the first EEG / MEG data, each of the first plurality of time series corresponding to a respective source location located within the cranial cavity (300) of the subject; the apparatus (140, 200) is configured to calculate (430) a first correlation value for a first pair of time series, the first pair of time series being included in the determined first plurality of time series; the device (140, 200) is configured to generate (440) one or more scores (150) based on the first correlation value, the one or more scores indicating that the subject has a cognitive disorder, such as a cognitive disorder like attention deficit hyperactivity disorder (ADHD); the device (140, 200) is configured to output (450) the generated score or scores. Device (140, 200).

2. the respective source locations of the first time series of the first pair of time series are located in a frontal lobe (310), such as the middle frontal gyrus (MFG) (320) and / or the inferior frontal gyrus (IFG) (330) of the cranial cavity (300) of the subject (110); 2. The apparatus (140, 200) of claim 1, wherein the respective source locations of a second time series of the first pair of time series are located in a superior parietal lobe (SPL) (340) of the cranial cavity of the subject.

3. the apparatus (140, 200) is further configured to calculate a second correlation value for a second pair of time series, the second pair of time series being included in the determined first plurality of time series; the one or more scores are generated based on the first correlation value and the second correlation value; and / or the respective source locations of a first time series of the second pair of time series are located in a ventral attention network region, such as the temporoparietal junction (TPJ) (350) of the cranial cavity (300) of the subject (110), and the respective source locations of a second time series of the second pair of time series are located in a dorsal attention network region, such as the superior parietal lobe (SPL) (340) of the cranial cavity of the subject; 3. The apparatus (140, 200) of claim 1 or 2.

4. the apparatus (140, 200) is further configured to calculate a third correlation value for a third pair of time series, the third pair of time series being included in the determined first plurality of time series; the one or more scores are generated based on the first correlation value, the second correlation value, and the third correlation value; and / or the respective source locations of a first time series of the third pair of time series are located in the ventral attention network region, such as the temporoparietal junction (TPJ) (350) of the cranial cavity (300) of the subject (110), and the respective source locations of a second time series of the second pair of time series are located in the ventral visual pathway region, such as the inferior temporal gyrus (ITG) (360) of the cranial cavity of the subject; 4. The apparatus (140, 200) of claim 3.

5. the device (140, 200) is further configured to receive second EEG / MEG data of the subject (110); the device (140, 200) is further configured to determine a second plurality of time series based on the second EEG / MEG data, each of the second plurality of time series corresponding to a respective source location located within the cranial cavity (300) of the subject; the apparatus (140, 200) is further configured to calculate a fourth correlation value for a fourth pair of time series, the fourth pair of time series being included in the determined second plurality of time series, the fourth pair of time series corresponding to the same respective source locations as the first pair of time series; the apparatus (140, 200) is further configured to calculate a comparison value between the fourth correlation value and the first correlation value; The apparatus (140, 200) of any one of claims 1 to 4, wherein the one or more scores (150) are generated based on the comparison values.

6. 6. The apparatus of claim 1, wherein determining the first plurality of time series comprises filtering the first EEG / MEG data such that only the first plurality of time series comprises a subset of all available time series in the first EEG / MEG data, the subset comprising the first pair of time series, a second pair of time series, and a third pair of time series.

7. An apparatus (140) according to any one of claims 1 to 6, an electroencephalogram or magnetoencephalogram device (EEG / MEG device) (120) for measuring at least the first EEG / MEG data (130) of the subject (110); a task presentation device (160) for presenting one or more tasks to the subject while the EEG / MEG device is measuring at least the first EEG / MEG data (130) of the subject; A system (100) comprising:

8. receiving (410) first electroencephalogram or magnetoencephalogram data (first EEG / MEG data) (130) of a subject (110); determining (420) a first plurality of time series based on the first EEG / MEG data, each of the first plurality of time series corresponding to a respective source location located within the cranial cavity (300) of the subject; calculating (430) a first correlation value for a first pair of time series, the first pair of time series being included in the determined first plurality of time series; generating (440) one or more scores (150) based on the first correlation values, the one or more scores indicating that the subject has a cognitive disorder, such as a cognitive disorder like attention deficit hyperactivity disorder (ADHD); outputting (450) the generated score or scores; A computer-implemented method (400) comprising:

9. the respective source locations of the first time series of the first pair of time series are located in a frontal lobe (310), such as the middle frontal gyrus (MFG) (320) and / or the inferior frontal gyrus (IFG) (330) of the cranial cavity (300) of the subject (110); 9. The method (400) of claim 8, wherein the respective source locations of a second time series of the first pair of time series are located in a superior parietal lobe (SPL) (340) of the cranial cavity of the subject.

10. The computer-implemented method (400) further comprises: calculating a second correlation value for a second pair of time series, the second pair of time series being included in the determined first plurality of time series; Including, the one or more scores are generated based on the first correlation value and the second correlation value; and / or the respective source locations of a first time series of the second pair of time series are located in a ventral attention network region, such as the temporoparietal junction (TPJ) (350) of the cranial cavity (300) of the subject (110), and the respective source locations of a second time series of the second pair of time series are located in a dorsal attention network region, such as the superior parietal lobe (SPL) (340) of the cranial cavity of the subject; 10. The method (400) of claim 8 or 9.

11. The computer-implemented method (400) further comprises: calculating a third correlation value for a third pair of time series, the third pair of time series being included in the determined first plurality of time series; Including, the one or more scores are generated based on the first correlation value, the second correlation value, and the third correlation value; and / or the respective source locations of a first time series of the third pair of time series are located in the ventral attention network region, such as the temporoparietal junction (TPJ) (350) of the cranial cavity (300) of the subject (110), and the respective source locations of a second time series of the third pair of time series are located in the ventral visual pathway region, such as the inferior temporal gyrus (ITG) (360) of the cranial cavity of the subject; The method (400) of claim 10.

12. The computer-implemented method (400) further comprises: receiving second EEG / MEG data of the subject (110); determining a second plurality of time series based on the second EEG / MEG data, each of the second plurality of time series corresponding to a respective source location located within the cranial cavity (300) of the subject (110); calculating a fourth correlation value for a fourth pair of time series, the fourth pair of time series being included in the determined second plurality of time series, the fourth pair of time series corresponding to the same respective source locations as the first pair of time series; calculating a comparison value between the fourth correlation value and the first correlation value; Including, the one or more scores (150) are generated based on the comparison values; The method (400) according to any one of claims 8 to 11.

13. The step of determining (420) the first plurality of time series comprises: filtering the first EEG / MEG data so that only the first plurality of time series comprises a subset of all available time series in the first EEG / MEG data (130); wherein the subset includes the first pair of time series, the second pair of time series, and the third pair of time series. The method (400) according to any one of claims 8 to 12.

14. Instructions which, when executed by a processor, perform the method (400) of any one of claims 8 to 13; A computer-readable storage medium storing the above.

Images