Automated thresholding for unsupervised neurofeedback sessions

1. A system for biofeedback training of a subject, said system comprising:

at least one sensor for obtaining a bio-signal of the subject;

a computing unit comprising a memory, the computing unit in communication with the at least one sensor, the computing unit configured to:

receive the bio-signal of the subject obtained by the at least one sensor;

for each first time window of a plurality of first time windows applied to the received bio-signal at a predefined first pace, calculate one biomarker value based on a portion of the received bio-signal in each first time window of said first time windows, so as to generate a time-series of biomarker values;

for each second time window of a plurality of second time windows applied to the time-series of biomarker values at a predefined second pace, the second time windows corresponding to a computation time t:

calculate an intermediate threshold Thr intermediate(t) based on the biomarker values comprised in said

second time window, said intermediate threshold Thr intermedediate(t) being a reward ratio representative of a stored expected reward ratio,

compute a threshold Thr (t) as a weighted sum of the intermediate threshold Thr intermediate(t) and a threshold of Thr (t−1) , where Thr (t−1) is obtained for a previous time window corresponding to a computation time t−1, and

calculate a difference between each biomarker value of the time-series of biomarker values and the threshold Thr (t) ;

and at least one of a display and an auditory feedback element, for reporting a reward to the subject, where said reward is based on said difference between each value of the time-series of biomarker values and the threshold Thr (t) .

2. The system for biofeedback training of a subject according to claim 1 , wherein a sum of weighting factors of said weighted sum is equal to 1.

3. The system for biofeedback training of a subject according to claim 2 ,

wherein the computing unit computes the threshold Thr (t) as follows:

Thr (t) =α* Thr (t−1) +(1−α) * Thr intermediate(t) ,

and wherein α is a coefficient that is one of constant and variable between 0 and 1.

4. The system for biofeedback training of a subject according to claim 3 , wherein the coefficient α is computed according to a logistic model.

5. The system for biofeedback training of a subject according to claim 4 ,

wherein the computing unit computes the coefficient a corresponding to a computation time t as:

α

t

=

α

t

-

1

+

r

.

α

t

-

1

*

(

k

-

α

t

-

1

k

)

;

and

wherein the memory has stored therein a learning coefficient k, a growth rate r and an initial value α 0 of the coefficient α.

6. The system according to claim 5 ,

wherein the memory also stores a reward ratio tolerance and a third time window, and

wherein the computing unit resets the coefficient α to the initial value α 0 if a reward ratio computed during the third time window deviates from the stored expected reward ratio by more than the reward ratio tolerance.

7. The system for biofeedback training of a subject according to claim 1 ,

wherein the memory has stored therein a time gating parameter, and

wherein a first reward is reported to the subject by the at least one of a display and an auditory feedback element whenever the subject maintains the biomarker values above the computed threshold Thr (t) for a time longer than the time gating parameter for an up training protocol, or whenever the subject maintains the biomarker values below the computed threshold Thr (t) for a time longer than the time gating parameter for a down training protocol.

8. The system for biofeedback training of a subject according to claim 7 , wherein the memory further stores comprises a time boosting parameter and a second reward is reported to the subject by the at least one of a display and an auditory feedback element whenever the subject maintains the biomarker values above the computed threshold Thr (t) during a time longer than the time boosting parameter for an up training protocol, or whenever the subject maintains the biomarker values below the computed threshold Thr (t) for a time longer than said time boosting parameter for a down training protocol.

9. The system for biofeedback training of a subject according to claim 1 , wherein the computing unit is further configured to remove artefacts from the bio-signal of the subject before computing the time-series of biomarker values.

10. A method for biofeedback training of a subject, said method comprising:

receiving a bio-signal of the subject obtained by at least one sensor;

for each first time window of a plurality of first time windows applied to the received bio- signal at a predefined first pace, calculating one biomarker value based on a portion of bio-signal comprised in said first time window, so as to generate a time-series of biomarker values;

for each second time window of a plurality of successive second time windows, each second time window of said second time windows corresponding to a computation time t:

computing an intermediate threshold Thr intermediate(t) based on the series of biomarker values comprised in said second time window, said intermediate threshold Thr intermediate(t) being a reward ratio representative of a stored expected reward ratio,

computing a threshold Thr (t) as a weighted sum of the intermediate threshold Thr intermediate(t) and a threshold Thr (t−1) obtained for a previous time window corresponding to a computation time t−1, and

calculating a difference between each biomarker value of said time-series of biomarker values and the computed threshold Thr (t) ; and

reporting a reward to the subject via at least one of a display and an auditory feedback element, where said reward is based on said difference between each value of the time-series of biomarker values and the threshold Thr (t) .

11. The method for biofeedback training of a subject according to claim 10 , wherein a sum of weighting factors of said weighted sum is equal to 1.

12. The method for biofeedback training of a subject according to claim 11 ,

wherein the threshold Thr (t) is computed as follows:

Thr (t) =α* Thr (t−1) +(1−α) * Thr intermediate(t) ,

wherein a is a coefficient that is one of constant and variable between 0 and 1.

13. The method for biofeedback training of a subject according to claim 12 , wherein the coefficient a follows is computed according to a logistic model.

14. The method for biofeedback training of a subject according to claim 13 ,

wherein α corresponding to a computation time t is computed as

α

t

=

α

t

-

1

+

r

.

α

t

-

1

*

(

k

-

α

t

-

1

k

)

,

and

wherein k is a learning coefficient, and r is a growth rate.

15. The method for biofeedback training of a subject according to claim 12 , wherein the coefficient a is reset to a predefined initial value a 0 if a reward ratio computed during a third time window departs from the stored expected reward ratio by more than a stored reward ratio tolerance.

16. The method for biofeedback training of a subject according to claim 10 , wherein a first reward is reported to the subject by the at least one of a display and an auditory feedback element whenever the subject maintains the biomarker values above the computed threshold Thr (t) for a time longer than a time gating parameter for an up training protocol, or whenever the subject maintains the biomarker values below the computed threshold Thr (t) for a time longer than the time gating parameter for a down training protocol.

17. The method for biofeedback training of a subject according to claim 16 , wherein a second reward is reported to the subject by the at least one of a display and an auditory feedback element whenever the subject maintains the biomarker values above the computed threshold Thr (t) during a time longer than a time boosting parameter for an up training protocol, or whenever the subject maintains the biomarker values below the computed threshold Thr (t) for a time longer than said time boosting parameter for a down training protocol.

18. The method for biofeedback training of a subject according to claim 10 , wherein initial values of Thr (t) are computed from a previous session or from a series of biomarkers computed from a bio-signal obtained under a given condition.

19. The method for biofeedback training of a subject according to claim 10 , further comprising:

removing artefacts from the bio-signal of the subject before computing the time-series of biomarker values.