WO2018202273A1

WO2018202273A1 - Method for determining brain performance of a subject

Info

Publication number: WO2018202273A1
Application number: PCT/EP2017/060318
Authority: WO
Inventors: Anastasios SMEROS
Original assignee: Individual
Current assignee: Individual
Priority date: 2017-05-01
Filing date: 2017-05-01
Publication date: 2018-11-08
Anticipated expiration: 2019-11-01

Abstract

A method for determining brain performance of a subject from a set of EEG signals utilizes a device comprising a digital processing unit formed of at least four mutually interconnected brain core modules (Mi) in conjunction with an EEG headset equipped with means for processing and transmitting measured EEG signals to the input ports of the device. Each device module comprises an input port for receiving a digitized EEG signal Si(t) obtained from the subject. The method provides an instant system state Ψ(t) defined by the following parameters: • link strength L(i,j)(t) between every pair of modules i and j; • self-link strength L(i, i)(t) of every module i; • total sublink strength L sub i tot ( t ) of each module i.

Description

Method for determining brain performance of a subject

Field of the Invention

The present invention generally relates to the determination of brain performance of a subject.

Background of the Invention

It is well known that brain performance of every individual is influenced by the most diverse external and subjective factors and thus shows considerable variation.

US 2010/324440 A1 discloses an approach for applying a real time stimulus triggered by a certain brain state. It includes receiving data that indicates onset of an instance of the brain state, whereupon a stimulus is initiated before the instance ends. The approach is intended to enhance performance, enhance learning or enhance the probing of impact of that state on perception, action or cognition.

WO 2006/073915 A2 describes a biological interface system comprising a sensor having a plurality of electrodes for detecting multicellular signals emanating from one or more living cells of a patient and a processing unit configured to receive the multicellular signals from the sensor to produce a processed signal. The system is configured to perform an integrated patient training routine to generate one or more system configuration parameters that are used by the processing unit to produce the processed signal.

Horstmann et al. ("State dependent properties of epileptic brain networks: comparative graph-theoretical analyses of simultaneously recorded EEG and MEG", Clinical Neurophysiology 121 (2010) 172-183) discloses a device for determining a brain performance status of an epilepsy patient.

However, from many points of view it would be desirable to have a system that is not primarily intended to detect certain brain conditions in order to apply an appropriate stimulus or to provide a diagnostic tool for a specific circle of patients, but rather is configured to rapidly and repeatedly provide very high quality information on brain performance status in a multitude of diagnostic and non-diagnostic settings. Summary of the Invention

In view of all the above, it is an object of the invention to provide new and improved approaches for determining the brain performance of a subject. These and further tasks are solved by the present invention.

According to one aspect of the invention (claim 1 ), there is provided a method for determining brain performance status P of a subject from a set of EEG signals by means of an acquisition system, the acquisition system comprising a modular device comprising a digital processing unit formed of at least four mutually interconnected brain core modules (Mi) wherein i = 1 to / is a running index associated with each module and wherein J ≥ 4 is the total number of modules, each module further comprising an input port (/_¾) for receiving a digitized EEG signal 5_¾(t) obtained from said subject, the processing unit further comprising:

means for establishing a link strength (L(i,y^')(£)) at time t between any pair of modules, wherein each link strength is a positive real number;

means for establishing a self-link strength L(i, i)(t) at time t of every module i; means for establishing a total sublink strength L^l _s^(l) at time t of each module i, wherein each total sublink strength is a positive real number;

means for determining an instant system state (t) defined by the following parameters:

link strength L(i,j)(t) between every pair of modules i and j;

self-link strength L(i, i)(t) of every module i;

total sublink strength L^l _s^ t) of each module i;

means for updating said instant system state in response to said EEG signals Sj(t); means for determining an instant modular performance p (t) for each module i; means for determining said brain performance status P from previously determined instant modular performances P (t) ;

the device being configured to be operated in a digitally clocked manner defining a sequence of discrete clock instants t = T_k with T_k = T₀ + k AT wherein k is a non-negative integer running from k = 0 to k = K and ΔΓ is a preselected clock interval, whereby said EEG signals, link strengths, total sublink strengths, instant system state and performance status have instant values at each clock instant T_k;

the acquisition system further comprising: an EEG headset comprising a plurality of at least four EEG electrodes (E_t with i = 1 to J and / > 4) configured for obtaining respective EEG raw signals S[^aw(t) from at least four standardized cranial locations {X_t) of the subject;

signal processing means for converting said EEG raw signals S[^aw(t) to corresponding digitized EEG signals Sj (t), and

transmission means for transmitting each one of said digitized EEG signals S_j(t) to one of said input ports associated with one of said EEG electrodes;

the method comprising the following steps:

a) setting an instant system state (k^~) , wherein k is the above defined integer valued running index, to be equal to a predefined initial system state (k = 0) = Ψ₀; b) increasing the running index by 1 to obtain an instant running index k, and acquiring said EEG signals for said predefined digital clock interval ΔΓ to obtain a set of instant digitized EEG signals Sj(/c) ;

c) performing an updating process whereby said instant system state (k - 1) is transformed to an updated system state (k) and whereby an instant modular performance Pi(k) is obtained for each module i;

d) until a predefined termination criterion is reached, repeating steps b) to d), e) determining said brain performance status P from the instant modular performances Px(k) obtained so far;

wherein each system state (k^~) at a given time is uniquely defined by the following properties:

link strength L(i,j)(k) between every pair of modules i and j;

self-link strength L(i, i)(/c) of every module i;

total sublink strength L^l _s^(k of each module i; and wherein said updating process comprises the following steps carried out for each module M . either a first step sequence comprising:

A. increase total sublink strength, starting from the last available value L^l _s^(k - 1), to obtain an updated total sublink strength L^l _s^(k for each module i according to: J

= L^l *£(Jc - 1) + β Si k) + a > L{i,j){k - 1)

followed by steps B and C, in any order: increase link strength between every pair of modules (i,j) and self-link strength of every module (i) to obtain updated values L(i,j)(k) according to:

L(i,y)(/c) = L(i,y)(/c - 1) + γ (S^/c) + S, (/c)) for V(i,y) e (1, K) obtain an instant modular performance Pi(k) for each module according to

or a second step sequence comprising:

B. increase link strength between every pair of modules (i,j) and self-link strength of every module (i) to obtain updated values L(i,j)(k) according to:

L(i,y)(/c) = L(i,y)(fc - 1) + γ (S_t(k) + Sj (k» for V(i,y) 6 (1, K) followed by:

A'. Increase total sublink strength, starting from the last available value L^l _s^(k - 1), to obtain an updated total sublink strength L^l _s^(k for each module i according to:

and

C. obtain an instant modular performance Ρχ (Κ) for each module according to

wherein, in the above sequences, a, β, γ and δ are positive valued scaling factors, the above steps, according to either the first step sequence or the second step sequence, being followed by:

D. optionally, if a predefined refining criterion is fulfilled, carry out a refining step.

According to another aspect of the invention (claim 6), there is provided a modular device for carrying out the method of the invention, the device comprising a digital processing unit formed of at least four mutually interconnected brain core modules ( j), each module further comprising an input port (/;) for receiving a digitized EEG signal Sj(t) obtained from said subject, the processing unit further comprising:

means for establishing a self-link strength L(i, i)(t) at time t of every module i; means for establishing a total sublink strength L^l _s^{l) at time t of each module i, wherein each total sublink strength is a positive real number;

link strength L(i,j)(t) between every pair of modules i and j;

self-link strength L(i, i)(t) of every module i;

total sublink strength L^l _s^ t) of each module i;

the device being configured to be operated in a digitally clocked manner defining a sequence of discrete clock instants t = T_k with T_k = T₀ + k AT wherein k is a non-negative integer running from k = 0 to k = K, whereby said EEG signals, link strengths, total sub- link strengths, instant system state and performance status have instant values at each clock instant T_k;

the the digital processing unit being programmed to execute steps a) to e) of the meth- According to a further aspect of the invention (claim 7), there is provided an acquisition system for carrying out the method of the invention, the system comprising:

a modular device according to the invention,

an EEG headset comprising a plurality of at least four EEG electrodes {E_t with i = 1 to J and / > 4) configured for obtaining respective EEG raw signals S[^aw(t) from at least four standardized cranial locations {X ) of the subject;

signal processing means for converting said EEG raw signals S[^aw(t) to corresponding digitized EEG signals Sj(t), and

transmission means for transmitting each one of said digitized EEG signals S_j(t) to one of said input ports (/j) associated with one of said EEG electrodes.

In one embodiment, the device is configured to operate with digitized signals normalized and integerized, e.g. to be an integer in the range between 0 and 10 or in the range between 0 and 100.

The above defined digitally clocked operation will typically be based on a time step AT = 1 sec and running, for example, for an acquisition time of 1 min, meaning that k is between k = 0 (initial state) and k = 60. However, it should be pointed out that any other convenient definitions of the running index k could be used without departing from the scope of the present invention.

As will be explained in more detail, the link strength between two modules can be understood as a coupling strength which determines how strongly the activity of two modules is correlated.

The total sublink strength of a module plays the role of a state function which characterizes the instant situation of the module and reflects the concept of each module having a plurality of internal nodes that are mutually coupled by respective sublinks. The total internal coupling strength within a given module can be characterized by one scalar property called the total sublink strength.

Devices and procedures for measuring EEG signals are generally known. Basically they comprise some kind of headset or other fixture for holding a plurality of EEG electrodes in contact with the subject's skull at predefined cranial positions. Each EEG electrode primarily generates a time dependent output voltage comprising a physiologically relevant signal component which is usually superimposed on a non-relevant background signal. Various signal processing means for subtracting unwanted background consisting e.g. of inevitable noise and/or a constant or quasi-constant baseline signal, are known in the art. Moreover, such signal processing means allow for transformation of analog primary EEG signals into digitized and optionally normalized EEG signals. Moreover, suitable transmission means serve for transmitting the various digitized and optionally normalized EEG signals to the brain performance device mentioned at the outset.

Although it is generally preferable to carry out an initial signal processing comprising noise filtering, background subtraction, analog-to-digital conversion and normalization in close proximity of each EEG electrode before transmitting the pre-processed signal to the brain performance device, this is not a mandatory processing sequence. In other words, it is contemplated that one could first transmit raw signals to a spatially displaced central processing unit.

It should also be pointed out that the above defined architecture comprising modules and connecting links does not need to be configured as distinct hardware modules. In certain embodiments, such architecture is implemented in a single physical unit with appropriate modular software processing.

It is also contemplated that instead of using just one EEG electrode for each module one could use a group of EEG electrodes for each module, and an input port could also serve multiple modules. Such more complex can be implemented within the signal processing sequence. The practical relevance of such arrangements will depend on the achievable signal quality, i.e. reliability, S/N ratio etc.

In certain embodiments the architecture will comprise an offsite "cloud" server.

Typically, the predefined termination criterion is reached when k reaches its maximum value K. As already mentioned above, for certain applications K is selected as 60, which in conjunction with a time step selected as ΔΓ = 1 sec implies an acquisition time of 1 min. Clearly, however, one may use other conditions, e.g. a substantially extended ac- quisition time with larger time steps and/or a larger number of acquisition steps K. On the other hand, it is also contemplated that additional termination criteria could be introduced leading to an earlier termination in certain predefined situations.

The above listed method steps will now be discussed in more detail.

It is assumed that each system state (k) at a given time t = T_k with T_k = T₀ + k AT is uniquely defined by the following properties: link strength L(i,j)(k) between every pair of modules i and j;

self-link strength L(i, i)(/c) of every module i;

total sublink strength L^l _s^(k) of each module i.

In the following, if not mentioned otherwise either explicitly or by the context of a mathematical expression, the term L(i,y) shall be understood to include both the link strength values L(i,y^')and the self-link strength values L(i, i), that is, it applies for V(i,y) e (1, K).

The compact notation W(k) will be used for brevity whenever there is no need to explicitly mention the values of L(i,j)(k) and L^l _s^(k) defining the system state. Accordingly, the basic method steps can be commented as follows:

a) setting an instant system state (k^~) , wherein k is the above defined integer valued running index, to be equal to a predefined initial system state (k = 0) = Ψ₀, meaning that one starts with initial values L(i,j)(k = 0) (abbreviated as L(i,j)₀ ) and L^l *£{k = 0) (abbreviated as L¾ );

b) increasing the running index by 1 to obtain an instant running index k, and acquiring said EEG signals for said predefined digital clock interval ΔΓ to obtain a set of instant digitized EEG signals Sj(/c), wherein "instant" refers to the integrated or averaged signal that was acquired in the k-th time interval;

c) performing an updating process whereby said instant system state (k - 1) is transformed to an updated system state (k) and whereby an instant modular performance Pi(k) is obtained for each module i; this step firstly involves determining updated values L^l _s^(k and L(i,j)(k) as a function of measured signals and other properties as explained further below, and secondly it involves evaluating the instant modular performance P (_.k), as also explained further below;

d) until a predefined termination criterion is reached, repeating steps b) to d), whereby the updated system state (k) just obtained is used as starting point for the next sequence of steps;

e) determining said brain performance status P from the instant modular performances Px(k) obtained so far, which shall be understood to encompass various possibilities of combining the available values P (_.k) to obtain P.

The loop of updating operations that is carried out after each acquisition time interval comprises the following steps carried out for each module M noting again that a, β, γ and δ are non-negative scaling factors.

In the following, the two possible sequences of the method steps A and B will be discussed in some more detail.

According to the first of two possible step sequences, step A is conducted before step B.

A. Increase total sublink strength, starting from the last available value L^l _s^(k - 1), to obtain an updated total sublink strength L^l _s^(k) for each module i according to:

As seen from the above, the increase of total sublink strength with respect to the last available value L^l _s^(k - 1) is made up from the following contributions:

β Sj(/c), which is directly proportional to the signal Sj(/c) received by the module under consideration, i.e. the i-th module;

a∑j=_i L(i,j)(k - 1) , which is proportional to the sum of the last available link j≠i

strengths L(i,j)(k - 1) between the i-th module under consideration and all the other modules j;

a L(i, i)(k - 1) , which is proportional to the last available self-link strength L(i, i)(k - 1) of the i-th module under consideration. In the above, β and a are are positive valued scaling factors that depend on the units chosen for the various quantities. For example, if the signals S_t are expressed in "mV", the are expressed in "sublinks" and the L(i,j) are expressed in "links", the dimension of β is "sublinks^■ mV^"1 " and the dimension of is "sublinks^■ links^"1 ".

B. Increase link strength between every pair of modules (i,j) to obtain updated values L(i,j)(k) according to:

L(i,y)(fc) = L .jXk - 1) + γ (Sj(/c) + S,(/c)) for V(i,y) with i≠ k and, analogously for self-link strengths:

L(i, i)(/c) = L(i, i)(k - 1) + γ (Sj(/c) + Sj(/c)) wherein γ is a positive valued scaling factor with units of "links^■ mV^"1 ".

The above implies that the fractional increase of link strength between any pair of modules (i,j) in relation to the starting value L(i,j)(k - 1) is proportional to the sum of the signals received by the two modules under consideration. It also implies that the fractional increase of self-link strength of each module (i) in relation to the starting value L(i, i)(k - 1) is proportional to the signal received by the module under consideration.

According to the second of two possible step sequences, step A is conducted after step B and will be called A' due to different indexing:

LQ.jXk) = L .jXk - 1) + γ (Sj(/c) + S,(/c)) for V(i,y) with i≠ k and, analogously for self-link strengths: )(/c) = L(i, (k - 1) + γ (5j (fc) + Si(fc)).

A'. Increase total sublink strength, starting from the last available value L^ik - 1), to obtain an updated total sublink strength L^l _s^(k for each module i according to:

Note that in contrast to step A of the first sequence, step A' of the second sequence uses link strengths and self-link strengths of the instant running index k, not k-1 .

The following step C is carried out after step A or A'.

According to step C, an instant modular performance P (k) for a given module, which is defined as

reflects the increase in total sublink strength most recently experienced by the module under consideration. It can be interpreted as a measure of a module's activity. If the are expressed in "performance units", then δ will be in: "performance units^■ links^"1 ".

D. Optionally, if a predefined refining criterion is fulfilled, carrying out a refining step.

By analogy with known neuronal mechanisms, some kind of partial "resetting" can be periodically performed to deactivate a certain fraction of total sublink strength of each module and/or of link strength between each pair of modules. In order to have a significant effect, the refining criterion should be fulfilled several times before the termination criterion is fulfilled. By way of example, if the termination criterion is chosen as K = 60 and AT = 1 sec, corresponding to a 1 minute EEG acquisition, it is advantageous to select e.g. K=10 for the refining criterion, meaning a refining operation every 10 seconds. It should be noted that the adoption of either the first step sequence or the second step sequence as mentioned above will not have a significant influence on the outcome unless the number of acquisition steps K is quite small. In practical settings where K is at least 10 or even substantially larger, the two step sequences yield essentially identical outcomes.

The advantages provided by the present invention are particularly evident if one bears in mind that EEG measurements are notoriously plagued by signal to noise problems, which prevent one from doing single shot high resolution measurements. By virtue of the above described operations, the system of the present invention progressively adopts the activity patterns of the brain regions contributing to the signals of the various electrodes. This activity pattern is encoded in terms of total sublink strengths, which are a measure of neuronal performance of a given module, and in terms of link strengths, which are a measure of synesthesia type coupling between any module pairs. With such a neuronally pre-trained system, it is possible to use EEG signals with a comparatively poor signal to noise ratio and nonetheless extract significant trends about a subject's brain performance.

It should be noted that appropriate selection of scaling factors a, β, γ and δ may require some initial numerical testing, for which purpose one may use simulations runs. As will be understood, such simulation runs can be carried out in silico, i.e. by inputting realistic model EEG signals S_j (/c) and computing therefrom the modular performances of interest. As a general guidance one may note the following principal effects of the various scaling factors:

Factors and β scale the importance of signal contribution and link strength, respectively, for the updating of sublink strength of every module i and consequent updating of modular performance.

Factor γ affects the importance of updating of link strength between every pair of modules i and j. A large value of γ leads to a gradual rise of modular performance and thereby acts as an integrator of signal inputs from all channels whereas a small value of γ emphasizes the effect of a specific channel input.

Factor δ merely scales the linear relationship between instant modular performance and the underlying increase in total sublink strength. Accordingly, one can select δ in order to display the calcualted performance on a convenient scale. Advantageous embodiments of the invention are defined in the dependent claims and described in the examples.

In principle, the number of EEG electrodes should be at least 4. However, it is preferable to have a somewhat larger number of EEG electrodes, for example 10 electrodes with 5 EEG electrodes each disposed on the left side and on the right side of the skull. According to an advantageous embodiment (claim 8), the EEG electrodes of said EEG headset are arranged according to the International 10-20-System.

It is also contemplated that the transmission means can be configured as conventional electrical or electro-optic signal transmission lines. According to one embodiment (claim 9), the transmission means comprise wireless transmission means.

In principle, the predefined initial system state (k = 0) = Ψ₀ needed to start the method of the present invention could be an arbitrarily selected system state. According to a possible embodiment, the predefined initial system state is a final system state obtained in a previous execution of the method. In principle, one could take the final system state obtained for another subject, preferably of a reference subject with similar characteristics such as age, gender, professional status, health status etc. In particular and if available, one could take a final system state obtained previously for the same subject.

Therefore, according to an advantageous embodiment (claim 2), the initial system state Ψ₀ is selected in such manner that the initial link strengths L(i,j)(k = 0) between every pair of modules i and j and also the initial self-link strengths L(i, i)(k = 0) of all the modules i are all equal to L₀ and wherein the initial total sublinks strength L^l _s^(k = 0) of all the modules i are all equal to L_sub o.

According to the invention, the brain performance status P(k) at any time during execution of the method is determined from the instant modular performances P_x (k) obtained so far. This includes the basic case where P(k = k_x) in a particular time slot k_x only depends on the instant modular performances P^ik = k_x). In another embodiment, an expanded definition of P(k) is adopted in which the instant modular performances at earlier times, i.e. values of Pi(k≤ k_x) are also taken into account. This could involve just a few of the most recent values. It is also contemplated to use some kind of weighting function, for example a weighting function that progressively reduces the impact of increasingly earlier values.

According to a favorable embodiment (claim 3), the brain performance status is obtained according to

which means averaging over the instant modular performances of all the modules.

According to one embodiment (claim 4), the partial resetting step comprises re-scaling each one of said total sublink strengths by multiplication with a scaling factor p_sub selected to be a real number between 0 and 1 . This operation can be understood to be analogous to the process of apoptosis (programmed cell death) of neuronal cells.

According to another embodiment (claim 5), the partial resetting step comprises re- scaling each one of said total link strengths by multiplication with a scaling factor p_Unk selected to be a real number between 0 and 1 . This operation can be understood as a process of reducing an excessive degree of synesthetic behavior that may build up during the brain mapping. As will be understood, using scaling factor p_link near 1 will have only a marginal effect, whereas low scaling factors around 0.1 or smaller will substantially erase any long term patterns accumulated during the measurement. Accordingly, scaling factors of about 0.5 will be useful for many settings.

Further Remarks

A subject's brain performance can be determined within various environments in parallel with real life actions as solo or within a group. Billions of hours per day are being spent by persons in conjunction with machines, mainly driving from a point a to b; or playing games for health benefits, sports, education, fun; or carrying, wearing devices; or using robots to serve, help in small or large tasks. Therefore, the determined brain performance of a subject obtained according to the present invention provides a feedback mechanism between a subject's inner brain functionality and machines; promoting machine intelligence and improving brain health at the same time.

The above process runs on a suitable software platform which acts to control the modular device and the acquisition system and to display, filter, store, compare and otherwise process the results.

Brief description of the drawings

The above mentioned and other features and objects of this invention and the manner of achieving them will become more apparent and this invention itself will be better understood by reference to the following description of various embodiments of this invention taken in conjunction with the accompanying drawings, in which are shown:

Fig. 1 a device for determining a brain performance status of a subject, in a schematic view;

Fig. 2 a first example of a 4-channel EEG, with a first pass single peak in one EEG channel, with:

a signals S1 , S2, S3 and S4 versus time,

b link strengths L(1 ,2), L(1 ,3), L(14), L(2,3), L(2,4) and L(3,4) versus time, c total sublink strengths M1 , M2, M3 and M4 and signal S1 versus time, d performance P1 , P2, P3 and P4 versus time,

Fig. 3 a second example of a 4-channel EEG, with a second pass single peak in one EEG channel, with:

a total sublink strengths M1 , M2, M3 and M4 and signal S1 versus time, b performance P1 , P2, P3, P4 and PTOT versus time,

Fig. 4 a third example of a 4-channel EEG, with a complex signal pattern in all

EEG channels, with:

a signals S1 , S2, S3 and S4 versus time,

b link strengths L(1 ,2), L(1 ,3), L(14), L(2,3), L(2,4) and L(3,4) versus time, c total sublink strengths M1 , M2, M3 and M4 versus time,

d performance P1 , P2, P3, P4 and PTOT versus time,

Fig. 5 a fourth example of a 4-channel EEG, with a complex signal pattern in all

EEG channels as in the fourth example, but with a reduced scaling factor γ = 0.01, with:

a link strengths L(1 ,2), L(1 ,3), L(14), L(2,3), L(2,4) and L(3,4) versus time, b performance P1 , P2, P3, P4 and PTOT versus time;

Fig. 6 a fifth example of a 4-channel EEG, with a repetitive peak in one EEG channel and comparatively low noise level in all channels, with:

a signals S1 , S2, S3 and S4 versus time,

b performance P1 , P2, P3 and P4 versus time; and

Fig. 7 a sixth example of a 4-channel EEG, with a repetitive peak in one EEG

channel and comparatively high noise level in all channels, with - unless otherwise specified - a default scaling factor γ = 1 and all initial link strengths L(i,y^')₀ = 1-0:

a signals S1 , S2, S3 and S4 versus time,

b performance P1 , P2, P3 and P4 versus time;

c performance P1 , P2, P3 and P4 versus time, with an enhanced scaling factor 7 = 100;

d performance P1 , P2, P3 and P4 versus time, with an enhanced scaling factor γ = 100 and an enhanced initial link strength L(l,2)₀ = l'OOO; e performance P1 , P2, P3 and P4 versus time, with an enhanced scaling factor γ = 100 and an enhanced initial link strength L(3,4)₀ = 20Ό00; f performance P1 , P2, P3 and P4 versus time, with a default scaling factor γ = 1 and an enhanced initial link strength L(3,4)₀ = 20Ό00.

Detailed description of the invention

In the following figures, if some components are present multiply, some of the reference numerals are not shown for every instance. In Fig. 1 , there is shown a device for determining a brain performance status of a subject H. In the example shown, the device comprises a digital processing unit PL) formed of four mutually interconnected brain core modules M_{l t} M₂ , M₃ and ₄ . Each module _¾ has an input port I_t for receiving a digitized EEG signal Sj(t) obtained from the subject.

The processing unit PL) further comprises:

means for establishing a total sublink strength L^l _s^{l) at time t of each module i, wherein each total sublink strength is a positive real number;

link strength L(i,j)(t) between every pair of modules i and j;

self-link strength L(i, i)(k) of every module i;

total sublink strength L^l _s^ t of each module i;

means for updating the instant system state in response to said EEG signals 5_¾(t); means for determining an instant modular performance P (t) for each module i; means for determining the brain performance status P from previously determined instant modular performances P (t).

As also shown schematically in Fig. 1 , the device comprises an analog-to-digital converter A/D serving to convert the raw EEG signals S[^aw(t) acquired by means of a headset HS into corresponding digitized signals Sj(t). The device also comprises suitable filters and/or discriminators, which can be configured as analog or digital components acting on the analog or digital version of the signal, respectively.

Moreover, in the example shown, the device is configured to be operated in a digitally clocked manner defining a sequence of discrete clock instants t = T_k with T_k = T₀ + k AT wherein k is a non-negative integer running from k = 0 to k = K, whereby the EEG signals, link strengths, total sublink strengths, instant system state and performance status have instant values at each clock instant T_k. In this manner, the device can operate with digitized signals normalized and optionally integerized, e.g. to be an integer in the range between 0 and 10 or in the range between 0 and 100, and also digitally clocked.

Examples

The following examples are simulations based on a system with 4 EEG electrodes with assumed exemplary EEG signals S1 , S2, S3 and S4 that were fed into the respective signal channels. A signal processing based on the so-called "first step sequence" defined further above, that is (A) followed by (B), was used in all examples.

In all the graphs of Figs. 2 to 5 the horizontal axes shows digital time steps k=0 to 19; in practice, this could typically represent time steps of 1 s each. In the graphs of Figs. 6 and 7, the horizontal axes shows digital time steps k=0 to 60. All the vertical axes are in arbitrary units (a.u.).

In the following the total sublink strength L^l _s^(k of a module will be denoted by the more compact notation Mi(k) with i=1 , 2, 3 or 4.

Unless otherwise noted, the simulations were run assuming a "first pass", i.e. using a default initial state with

L(i,y)₀ = 1.0 for V(i,y) e (1,4) and with scaling factors

α = β = γ = l.o

No refining step (D) was conducted. Example 1 (Fig. 2)

The simulated signal pattern comprises a single peak of S1 at k=6 with a height of 100 a.u. in channel 1 , whereas the signal in the three other channels S2, S3 and S4 has merely some noisy background (Fig. 2a). As shown in Fig. 2b, this signal input leads to a step-like increase of the link strengths L(1 ,2), L(1 ,3) and L(1 ,4) induced by the peak of S1 . These are the link strengths coupled to the module excited by S1 . The other link strengths, i.e. L(2,3), L(2,4) and L(3,4), but also the link strengths L(1 ,j) in the regions displaced from the signal peak S1 , merely show a weak gradual increase. As shown in Fig. 2c, the above signal input also leads to a step-like increase of the total sublink strength M1 of module 1 , again coincident with the peak of S1 , whereas M2, M3 and M4 again show a weak gradual increase. Finally, Fig. 2d shows that the performance value P1 essentially matches the shape of the signal peak of S1 whereas the performance values P2, P3 and P4 in the other channels fluctuate around a very low value.

Example 2 (Fig. 3)

In this example the same signal sequence as that used in Example 1 was fed to a system in which the final system state of Example 1 was used as the initial system state. As seen from Fig. 3a, the total sublink strength M1 starts from a level of about 124 a.u. and undergoes a step at k=6 followed by a rather steeply increasing trend thereafter. Somewhat similarly, M2, M3 and M4 show a markedly increasing behavior after k=6. As seen from Fig. 3b, the performance P1 has a peak coincident with the signal peak of S1 , but in contrast to the first pass case P1 remains at a significant level of about 25 a.u. after the peak. The performances P2, P3 and P4 undergo a step at k=6, also in contrast to the first pass behavior.

Example 3 (Fig. 4)

This example illustrates the behavior using a more complex EEG pattern with a noisy background signal in channel 1 , a single 100 a.u. peak in channel 2 at k=3 and a 50 a.u. sequential double peak at k=7 and k=10 in channel 3 and a 60 a.u. peak in channel 4 at k=15 (see Fig. 4a). Such signals lead to a correspondingly complex behavior of the link strengths (Fig. 4b), total sublink strengths (Fig. 4c) and performance (Fig. 4d).

Example 4 (Fig. 5)

This example was based on using the same complex signal input as in Example 3 but with a substantially reduced signal impact scaling factor of γ = 0.01. As seen by comparing Figs. 4a and 5b, this change of scaling factor had no influence on the performance values. In contrast, the reduction of scaling factor led to a massive reduction of link strengths (cf. Figs. 4b and 5a). Example 5 (Fig. 6)

This example was based on using a threefold repetition of the single peak of S1 according to Example 1 . The modular performance values shown in Fig. 6b are seen to match closely the behavior of the signal inputs shown in Fig. 6a.

Example 6 (Fig. 7)

This example was based using the signal sequence of Example 5, but with a tenfold reduced intensity of the three dominant peaks in channel 1 , see Fig. 7a. Accordingly, the example shows the behavior when the background noise is tenfold higher than in Example 5. The resulting modular performance patterns are shown in Figs. 7b to 7f for different settings. Using the standard "first pass" settings with all initial link strengths L(i,))₀ = 1.0 and a scaling factor γ = 1 again leads to a modular performance pattern closely matching the signal input (Fig. 7b). If the scaling factor is increased to 7 = 100, a substantially monotonous increase of modular performance is found in all channels (Fig. 7c). By increasing the initial link strength L(l,2)₀ = l'OOO and keeping 7 = 100, one introduces a substantial coupling between signal exposed channel 1 and non-exposed channel 2, with a concomitant increase in modular performance of channels 1 and 2 relative to the uncoupled and unaffected channels 3 and 4 (Fig. 7d). By comparison, if one selectively increases the initial link strength L(3,4)₀ = 20Ό00 and thus introduces a substantial coupling between two non-exposed channels (while keeping γ = 100 ), there is an increase in modular performance of non-exposed channels 3 and 4 (see Fig. 7e). Finally, if one reduces the scaling factor back to its original value of γ = 1 while keeping L(3,4)₀ = 20Ό00, one effectively removes the monotonous increase of modular performance (see Fig. 7f).

total number of EEG electrodes (and modules)

EEG electrode (i = l to J)

standardized cranial locations

EEG signal (i = l to J)

digital clock interval

digital clock instant

number of acquisition steps

module (i = 1 to J)

input port of module i

link strength between modules j and M_j self-link strength of module _¾

total sublink strength of module M_t

running index for digitized time slots instant modular performance of module _¾ brain performance status in time slot k instant system state

all pairs (i,j) wherein i and j each are from 1 to K scaling factor

scaling factor

Claims

1 . A method for determining brain performance status P of a subject from a set of EEG signals by means of an acquisition system,

the acquisition system comprising a modular device comprising a digital processing unit formed of at least four mutually interconnected brain core modules (Mi) wherein i = 1 to / is a running index associated with each module and wherein / > 4 is the total number of modules, each module further comprising an input port (/_¾) for receiving a digitized EEG signal 5_¾(t) obtained from said subject, the processing unit further comprising:

means for establishing a link strength (L(i, )(t)) at time t between any pair of modules, wherein each link strength is a positive real number;

link strength L(i,j)(t) between every pair of modules i and j; self-link strength L(i, i)(t) of every module i;

total sublink strength L^l _s^ t) of each module i;

means for updating said instant system state in response to said EEG signals means for determining an instant modular performance p (t) for each module i;

means for determining said brain performance status P from previously determined instant modular performances P (t);

the device being configured to be operated in a digitally clocked manner defining a sequence of discrete clock instants t = T_k with T_k = T₀ + k AT wherein k is a non- negative integer running from k = 0 to k = K and ΔΓ is a preselected clock interval, wherein K is a predefined positive integer number of acquisition steps, whereby said EEG signals, link strengths, total sublink strengths, instant system state and performance status have instant values at each clock instant T_k ;

the acquisition system further comprising: an EEG headset comprising a plurality of at least four EEG electrodes with i = 1 to J and / > 4) configured for obtaining respective EEG raw signals S™^w(t) from at least four standardized cranial locations {X_t) of the subject; signal processing means for converting said EEG raw signals S[^aw(t) to corresponding digitized EEG signals Sj (t), and

transmission means for transmitting each one of said digitized EEG signals Sj(t) to one of said input ports associated with one of said EEG electrodes.

the method comprising the following steps:

a) setting an instant system state (k) , wherein k is the above defined integer valued running index, to be equal to a predefined initial system state

W(k = 0) = Ψ₀, the initial system state being defined by initial values

L (i,j) (k = 0) and = 0) ;

b) increasing the running index by 1 to obtain an instant running index k, and acquiring said EEG signals for said predefined digital clock interval ΔΓ to obtain a set of instant digitized EEG signals Sj (/c) ;

c) performing an updating process whereby said instant system state (k - 1) is transformed to an updated system state (k~) and whereby an instant modular performance Pi (k) is obtained for each module i;

d) until a predefined termination criterion is reached, repeating steps b) to d), e) determining said brain performance status P from the instant modular performances Px (k obtained so far; wherein said updating process comprises the following steps carried out for each module M . either a first step sequence comprising:

A. increase total sublink strength, starting from the last available value

L^ub ik - 1), to obtain an updated total sublink strength L^l _s^ (k) for each module i according to:

j followed by steps B and C, in any order:

B. increase link strength between every pair of modules (i,j) and self-link

strength of every module (i) to obtain updated values L(i,j)(k) according to:

L(i,y)(fc) = γ (Sidk) + S_j(k)) for v y) e (i, K)

C. obtain an instant modular performance P (k) for each module according to

or a second step sequence comprising:

B. increase link strength between every pair of modules (i,j) and self-link

strength of every module (i) to obtain updated values L(i,j)(k) according to:

L(i,y)(fc) = γ (Sidk) + S_j(k)) for v y) e (i, K) followed by:

A', increase total sublink strength, starting from the last available value

^Lisub (_.k - l)_> ^t0 obtain an updated total sublink strength L^l _s^(k) for each module i according to:

j

(/c - 1) + β Si(k) + a Ld.j k)

7 = 1

and

C. obtain an instant modular performance Pi(k) for each module according to

Pi(k) = S [L^{i t}(k) - L^{i t}(k - l)] wherein, in the above sequences, , β, γ and δ are positive valued scaling factors, the above steps being followed by: D. optionally, if a predefined refining criterion is fulfilled, carry out a step of partial resetting.

2. The method according to claim 1 , wherein said predefined initial system state Ψ₀ is selected as follows:

L(i,;) (fc = 0) = L₀ for v y) e (1, K)

and (fe = 0) = Lsub o for v(i) e (l, /0-

3. The method according to claim 1 or 2, wherein said brain performance status P(t) is obtained according to

j

p^{(/c) =} 7∑ ^p*^{i (/c)}

i = l

4. The method according to one of claims 1 to 3, wherein said partial resetting step comprises re-scaling each one of said total sublink strengths by multiplication with a scaling factor p_sub selected to be a real number between 0 and 1 .

5. The method according to one of claims 1 to 4, wherein said partial resetting step comprises re-scaling each one of said total link strengths by multiplication with a scaling factor p_link selected to be a real number between 0 and 1 .

6. A modular device for carrying out the method according to one of claims 1 to 5, comprising a digital processing unit formed of at least four mutually interconnected brain core modules (Mi) , each module further comprising an input port (/¾) for receiving a digitized EEG signal 5_¾(t) obtained from said subject, the processing unit further comprising:

means for establishing a self-link strength L(i, i)(t) at time t of every module i; means for establishing a total sublink strength L^it) at time t of each module i, wherein each total sublink strength is a positive real number;

total sublink strength L^l _s^ t of each module i;

means for updating said instant system state in response to said EEG signals s_t(ty,

means for determining an instant modular performance P (t) for each module i;

the device being configured to be operated in a digitally clocked manner defining a sequence of discrete clock instants t = T_k with T_k = T₀ + k AT wherein k is a non- negative integer running from k = 0 to k = K and ΔΓ is a preselected clock interval, whereby said EEG signals, link strengths, total sublink strengths, instant system state and performance status have instant values at each clock instant T_k ; characterized in that the digital processing unit is programmed to execute steps a) to e) of the method.

7. An acquisition system for carrying out the method according to one of claims 1 to 5, comprising:

a modular device according to claim 6,

an EEG headset comprising a plurality of at least four EEG electrodes {E_t with i = 1 to J and / > 4) configured for obtaining respective EEG raw signals Sf^aw(t) from at least four standardized cranial locations {X ) of the subject; signal processing means for converting said EEG raw signals S[^aw(t) to corresponding digitized EEG signals Sj(t), and

transmission means for transmitting each one of said digitized EEG signals Sj(t) to one of said input ports associated with one of said EEG electrodes. The acquisition system according to claim 7, wherein the EEG electrodes of said EEG headset are arranged according to the International 10-20-System.

The acquisition system according to claim 7 or 8, wherein said transmission means comprise wireless transmission means.