FIELD OF THE INVENTION

The present invention relates to the field of biofeedback, especially neurofeedback for helping in reducing a person's musculoskeletal pain. More specifically, the invention relates to a ElectroEncephaloGraphy (EEG) based neurofeedback system, a method for providing neurofeedback to a person, and a method for reducing a person's musculoskeletal pain. BACKGROUND OF THE INVENTION

Musculoskeletal pain is pain that affects the muscles, ligaments, tendons and bones. An example of such pain is a 'tennis elbow'. The indirect socioeconomic costs due to chronic musculoskeletal pain have far exceeded those estimated for heart disease, cancer and diabetes. Therefore, non- pharmacological treatments of chronic musculoskeletal pain are highly desirable.

Examples are seen in the literature, where reduction of a person's pain is achieved by neurofeedback, where patients are trained to control specific EEG waves to reduce their sensation of pain. However, these methods only work for central neuropathic pain or migraines, not for musculoskeletal pain.

SUMMARY OF THE INVENTION

Thus, according to the above description, it is an object of the present invention to provide an efficient method and a device for reducing and possibly eliminating musculoskeletal pain of a person. In a first aspect, the invention provides a neurofeedback system comprising

- an EEG sensor arranged for mounting on a person, so as to sense

electroencephalogram brain waves of the person, and to output data values accordingly,

- a display arranged to provide visual feedback, and - a processor system arranged to receive data values from the EEG sensor, and to calculate a value indicative of a power magnitude within a limited frequency band, accordingly, wherein the processor system comprises a processor programmed to execute a control algorithm arranged to cause the processor system to:

- determine a baseline value indicative of a power magnitude in response to data from the EEG sensor within a limited frequency band, such as the EEG alpha frequency band (8-13 Hz),

- determine, in real-time, such as sampled at a rate of 250 Hz, values indicative of a power magnitude in response to data values from the EEG sensor within a limited frequency band, such as the EEG alpha frequency band (8-13 Hz), and

- display on the display a real-time visual indication of said values compared to said baseline value, wherein at least one attribute, such as color, of said visual indication is determined, in real-time, in response to said values being above or below a threshold which is determined in response to said baseline value.

Such neurofeedback system is advantageous, since it has been found in

experiments that it is possible to reduce a person's musculoskeletal pain (e.g. tennis elbow) by giving the person real-time feedback of a level of the person's specific EEG band, e.g. the alpha band (8-13 Hz), the beta band (14-29 Hz), and also the gamma band (>30 Hz) are relevant, during performing of a body movement causing the painful sensation. Thus, training such painful body movements using the neurofeedback system for providing feedback may have the effect of reversing maladaptive brain plasticity associated with pain conditions.

The invention is based on the inventors' insight that perceived musculoskeletal pain caused by arm/finger movement correlates especially with measured EEG alpha band (8-13 Hz) signal power for a person: larger alpha power corresponds to a more painful movement. The EEG signal can be measured non-invasively. It has been found that it is possible for a person to train by performing a painful movement, e.g. using a forced feedback instrument, and at the same time watching the corresponding EEG (alpha or other band) power level in real-time. Hereby, the person can subconsciously train a reduction in EEG (alpha or other band) power level associated with the painful movement, and thereby reduce pain sensation by reversing maladaptive brain plasticity associated with the pain condition.

The alpha band power level has been found to be useful for certain types of pain. However, for other types of pain, the relevant limited frequency band and/or the relevant time window (relative to onset of the movement) to be used for the baseline, to be used for the baseline can be selected by identifying specific type of such limited frequency bands and/or time window signature specific for the type of pain in order to provide the most significant baseline and thus the most efficient neurofeedback stimulation to be applied to patients suffering from the pain. E.g. such baseline signature can be derives from a group of patients suffering from this type of pain and compare their EEG responses during performing the painful movement with a healthy control group to identify the most significant differences, thereby revealing the frequency and/or time signature specific for the type of pain which in turn can provide the basis for the baseline frequency band and/or time window to be monitored and provided as feedback to the patients. Preferably, in general the EEG sensor comprises a plurality of EEG electrodes to be located at different positions, e.g. 3-10, electrodes on the patient's scalp, since position specific frequency and time signatures may be significant to be used for a specific type of pain. However, once identified for the specific type of pain, the most relevant one or few EEG electrode positions can be used in final neurofeedback system.

Such baseline signature as mentioned above may be provide by initial trials for one individual patient, i.e. where EEG values are monitored during the patient performing the painful movement, and where an automatic or semi-automatic selection algorithm is applied to the data obtained to identify a limited EEG frequency band and/or time windows which generate the most specific (e.g. high or low) energy compared to a healthy subject. A frequency analysis can be performed in time windows to allow a frequency versus time analysis, e.g. to identify frequency and time segments with high or low energy compared to prestored data for healthy subjects.

EEG signals from the training person in real time can be captured non-invasively by a wired or wireless EEG sensor device, and thus in a low cost version, the total system can be implemented by such EEG sensor device connected to a normal computer, a tablet or even a smartphone programmed to perform the

neurofeedback training algorithm. The person suffering from pain is motivated to train with such system, since the visual attribute change can be seen as a "reward" as in a game. Further aspects can be implemented in the control algorithm to "reward" the person for achieving low EEG power magnitude levels. E.g. attribute changes may include displaying symbols or photos, and/or showing numbers or graphs indicating the obtained EEG power magnitude results.

Some embodiments comprise a feedback device arranged to at least partly follow a predetermined body movement of a person, and comprising a sensor, e.g. an electric switch, arranged to sense a start of a predetermined body movement, and being arranged to output a timing signal to the processor accordingly. This allows calculating the baseline value as an average value over a period of time before the person initiates the body movement, or in other words what can be called the preparation phase just before the body movement. Especially, the control algorithm may be arranged to calculate a value indicative of a power magnitude in response to data from the EEG sensor averaged over a period of time comprising a time interval corresponding to a preparation phase before a start of said predetermined body movement. Especially, said time interval has a length of 0-5 seconds, such as 0.2-3 seconds, such as 0.5-2 seconds. The system may comprise a feedback device comprises an exercising instrument arranged to guide the person in performing a predetermined body movement causing musculoskeletal pain, such as comprising an element providing a load against performing said movement. With such exercising instrument, it may be easier for the person to perform the same movement each time, and in version with a load against performing the movement, e.g. a spring or another type of load providing device, it is possible to adjust the load to provide a reasonable amount of pain for the person during the movement. This may be gradually increased, once the pain reduction effect of the training has set in, and the person can trains with a higher load. Especially, the feedback device may comprise a controllable actuator arranged to provide a controllable load against performing said body movement. In some versions, the processor system is arranged to control the controllable actuator, and wherein the control algorithm is arranged to control the controllable actuator, so as to control a load against performing said body movement in response to received data values from the EEG sensor, such as to control the load against performing said movement adaptively in response to data values from the EEG sensor. The feedback device is preferably arranged to at least partly follow a

predetermined hand or arm movement of the person. However, it is to be understood that the system may be used also for other body movements, and thus a feedback device suited to the specific type of musculoskeletal pain is preferred.

The limited EEG frequency band preferably comprises at least a part of the EEG alpha frequency band, i.e. at least a part of the band 8-13 Hz. E.g. the limited EEG frequency band is the EEG alpha frequency band 8-13 Hz, or comprises at least 80% of the EEG alpha frequency band 8-13 Hz. The EEG alpha band is preferred, since it has been proven experimentally to correlate with the person's sensation of pain, at least with respect to some types of pain.

However, in general to be able to cover various types of pain, the limited frequency band to be used for calculation of the baseline value may be selected in response to the specific type of pain. Thus, for a specific person, said limited frequency band may be selected in response to a specific type of pain which the person suffers from. This can be either based on trial EEG data obtained for the individual person or patient, or based on predetermined EEG data obtain for a group of persons suffering from the specific type of pain. E.g. EEG signals in the beta or gamma frequency band can be relevant for some types of pain. Further, specific time windows relative to onset of the person performing the painful movement can be specific for a specific type of pain, and thus specific frequency- time signatures can be relevant for specific types of pain in order to provide the most effective neurofeedback to a person suffering from the specific type of pain. The system may have access to tabulated frequency-time signatures relevant to be used for baseline value calculation for a set of specific types of pain, and thus the relevant signature can be selected by software in response to a type of pain being entered to the system. Specifically, the control algorithm may causes the processor to determine said limited frequency band in response to a selection algorithm based on sensing data values from the EEG sensor over a period of time during the person performing a predetermined movement. Specifically, said limited frequency band may be selected by the selection algorithm in response to a spectral analysis performed on said data values from the EEG sensor collected over the period of time. Said period of time may be divided into a plurality of time windows, and said period of time may be selected to start from a certain period, such as 1-5 seconds, before the person performing the movement, until a certain period, such as 1-5 seconds, after the movement has been performed. The selection algorithm may be arranged to select one or more time windows and said limited frequency band in response to the person performing the movement, e.g. the selection algorithm may be arranged to select one or more time windows within the limited frequency band in response to the person performing the movement. The EEG sensor may comprise a plurality of electrodes, such as 3-10, arranged for being positioned at different positions on the person's scalp. Said limited frequency band may be selected from an analysis of individual EEG values collected from the plurality of individual electrodes. A limited time window relative to the person performing a movement is selected to be used for calculation of the base line value in response to a specific type of pain which the person suffers from. Said limited frequency band and said limited time window relative to the person performing a movement are selected in response to a specific type of pain which the person suffers from. The control algorithm may be arranged to select said limited frequency band in response to EEG data obtained for the individual person performing a painful movement in a test trial. The control algorithm may be arranged to select said limited frequency band in response to a specific type of pain by selecting one among a plurality of predetermined limited frequency band data based on EEG data obtained from an experiment with persons suffering from the specific type of pain. Specifically, the limited frequency band is selected as a limited band within the frequency interval 8-30 Hz. Specifically, the limited time window for analysing for time signatures, may have a length of such as 0.1-10 seconds, such as 0.1-1 second.

The processor system is preferably arranged to display a bar on the display, wherein a height of said bar is continuously adjusted in real-time according to a power magnitude in response to data from the EEG sensor. Especially, the processor system is arranged to change a color of said bar in real-time in response to said values indicative of a power magnitude in response to data values from the EEG sensor being above or below said threshold determined in response to said baseline value. E.g. the bar may be displayed with a green color when below said threshold, and wherein the bar is then displayed with a red color, when above said threshold. This provides the training person with an immediate feedback on the EEG level performance. The processor system may be arranged to display a bar on the display, wherein a height of said bar is adjusted according to said baseline level.

The baseline value may be determined in response to data values from the EEG sensor during the person performing a plurality of repetitions of said

predetermined body movements. E.g. as an average power magnitude value calculated for 5-50 repetitions, such as 5-30 repetitions. It is to be understood that the baseline value may simply be the average value, or the average value subtracted by a threshold, so as to make the task for the training person more more or less difficult for the person by adjusting the threshold.

By providing the visual real-time indication in response to EEG power values being above or below a threshold based on the baseline value, the threshold can be adjusted to be above or below the baseline value, and thus allowing the point where a visual attribute, e.g. color of a bar, changes. This can be used to adjust how difficult the person perceives the task of obtaining the goal of changing the visual attribute to "good" or "acceptable". Especially, of course the threshold may be set to be equal to the baseline value, and thus the visual attribute changes around the calculated baseline value, e.g. a baseline value calculated as an average of the EEG alpha band power magnitude obtained in a plurality of trials. Especially, the processor system may be implemented as a computer, a tablet, a smartphone, or a dedicated device. Further, such devices also has a display which can be used as the display for the visual feedback during training.

The EEG sensor can be a standard low cost component. Especially the EEG sensor may comprise a plurality of single EEG sensors arranged to be placed at various positions on the person's head, e.g. a plurality of single sensors spatially distributed on a cap to be placed on the person's head.

The interface for transferring EEG data from the EEG sensor to the processor system may be wireless or wired.

In a second aspect, the invention provides a method for providing neurofeedback to a person, the method comprising

- receiving real-time EEG data indicative of electroencephalogram brain waves of the person,

- calculating real-time value indicative of a power magnitude within a limited frequency band, such as the EEG alpha frequency band (8-13 Hz), in response to the EEG data,

- determining a baseline value indicative of a power magnitude in response to the EEG data within a limited frequency band, such as the EEG alpha frequency band (8-13 Hz), and

- displaying a real-time visual indication of said real-time values compared to said baseline value, wherein at least one attribute, such as color, of said visual indication is determined, in real-time, in response to said values being above or below a threshold which is determined in response to said baseline value. In preferred embodiments, the method is a method for providing neurofeedback to a person, the method comprising

- sensing real-time electroencephalogram brain waves of the person during the person performing a predetermined body movement causing musculoskeletal pain with an EEG sensor, - calculating real-time value indicative of a power magnitude within a limited frequency band, such as the EEG alpha frequency band (8-13 Hz), in response to data from the EEG sensor,

- determining a baseline value indicative of a power magnitude in response to data from the EEG sensor within a limited frequency band, such as the EEG alpha frequency band (8-13 Hz), in response to the person performing a predetermined body movement causing musculoskeletal pain, and

- displaying a real-time visual indication of said real-time values compared to said baseline value, during the person repeatedly performing said predetermined body movement causing musculoskeletal pain, wherein at least one attribute, such as color, of said visual indication is determined, in real-time, in response to said values being above or below a threshold which is determined in response to said baseline value. In a third aspect, the invention provides a computer executable program code, or a programmable- or fixed hardware, and/or combination hereof, arranged to perform the method according to the second aspect, when executed on a processor. The computer executable program code may be stored on a data carrier. The program code may be implemented to function on any type of processor platform.

In a fourth aspect, the invention provides a method for reducing musculoskeletal pain of a person, comprising the person training by performing a painful body movement along with receiving neurofeedback according to the method according to the second aspect.

It is appreciated that the same advantages and embodiments described for the first aspect apply as well for the second, third, and fourth aspects. Further, it is appreciated that the described embodiments can be intermixed in any way between all the mentioned aspects.

BRIEF DESCRIPTION OF THE FIGURES

The invention will now be described in more detail with regard to the

accompanying figures of which FIG. la illustrates EEG alpha band power for pain patients, and for healthy person with induced pain: prior to, and following injections of hypertonic saline,

FIG. lb illustrates for the same patients as illustrated in Fig. la, where pressure pain thresholds PPT prior to and at the end of the task were monitored using the VAS,

FIG. 2 illustrates an example of an algorithm for a neurofeedback training session, FIG. 3 shows a photo of an experimental setup with a person performing with a training session using a hand movement feedback device,

FIG. 4 shows an example of a display layout with colored bars for visual feedback of the alpha EEG power level compared to a baseline value,

FIG. 5 illustrates alpha EEG power levels for a person across three training days, FIG. 6 illustrates sensory-motor-rhythm for a person during a movement execution, where levels at different EEG wave frequencies are illustrated versus time,

FIG. 7 illustrates a sketch of elements of a system embodiment,

FIG. 8 illustrates steps of a method embodiment,

FIGs. 9a and 9b show results of a control study where no feedback was given to patients during 4 days,

FIG. 10 shows results of three sessions each with three runs where patients were given neurofeedback during performing the painful movement,

FIG. 11 shows time-frequency analysis on EEG data for healthy (top), patients (middle) and difference between the two groups (lower),

FIG. 12 illustrates time-frequency analysis of EEG signals from various sensor positions on the scalp as differences between patients with pain and healthy persons,

FIG. 13 shows graphs of Movement Related Cortical Potentials (MRCP) versus time for control condition (top), and after saline injection inducing tennis elbow pain (lower),

FIG. 14 shows a graph of MRCP for the experiment of FIG. 13 indicating variability calculated for the planning phase of the movement,

FIG. 15 shows, again for the experiment of FIG. 13 MRCP, the time of peak negativity associated with the initiation of the movement, where pain patients are included for reference. The figures illustrate specific ways of implementing the present invention and are not to be construed as being limiting to other possible embodiments falling within the scope of the attached claim set. DETAILED DESCRIPTION OF THE INVENTION

FIGs. la and lb illustrate basic observations, which form the foundation of the present invention, namely a correlation with the sensation of pain, Pressure Pain Threshold PPT, as the Visual Analog Scale VAS. Pain patients 'Pain', healthy person with induced pain prior to ΉηΡ', and following ΉννΡ 'injections of hypertonic saline. Fig. la indicates EEG alpha power from one electrode location (C3) across all subjects. Alpha power was significantly increased in the pain patients Pain compared to HnP (p<0.05), while induced pain HwP showed a similar trend. This was correlated to the participants perceived pain VAS, which shows a high correlation between pain sensation and EEG alpha power level.

FIG. 2 outlines an example of an algorithm for a neurofeedback method to reduce musculoskeletal pain, which has been tested in a preliminary version on one pain patient.

FIG. 3 shows a photo of a person with EEG sensors mounted on top of her head. She uses her hand in a feedback device used to do wrist extensions, and a display is used to provide the visual neurofeedback of measured EEG alpha power levels. For the first 20 trials subjects were asked to perform wrist extensions with a light weight held in their hand. The power within the EEG alpha frequency band was subsequently calculated and served as the baseline value in the following trials, during which the alpha band power was continuously displayed to the subject in the left panel of the feedback screen (display). A green bar indicated a decrease in power (a desynchronization) while a red bar referred to an increase in power (a synchronization).

FIG. 4 shows an example of a screen dump of the display at a random time, merely to illustrate the basic layout with bars, and where the height of the bars indicate the magnitude of EEG alpha power level relative to the baseline value. Subjects were asked to try to maintain the bars green. Upon movement performance, the system fed back to the subject the power within the same band but only for the preparation phase of the movement, right bar of the feedback screen, see FIG. 4. A successful trial entailed that this bar was green, thus that power was maintained below the baseline value. The first of three blocks of 50 trials were performed and the percentage displayed on the screen indicated to the subject how many successful trials were completed. During the second block, the baseline values were adjusted to those obtained for the previous block. For the third block the baseline values were adjusted based on the second block. In this way the task difficulty increased for each block, ensuring the subjects were trained appropriately. Three patients have been exposed to this neurofeedback system, to date.

FIGs. 5 and 6 show results from one patient. FIG. 6 shows the change in sensory- motor-rhythm (SMR) during that the movement execution was evident, which may be difficult due to the lack of colors. However, the EEG alpha power (8-13 Hz) was decreased (dark area) within a session but also across the three sessions (performed on separate days), see FIG. 5. More importantly, these decreases in power were accompanied by decreases in the VAS scale (from 5.6 after session one to 0 at the end of session three) indicating that the patient felt less pain by session number three. FIG. 7 shows a sketch of basic elements of a system embodiment, namely a neurofeedback training system arranged to reduce a person's musculoskeletal pain. Inputs from the involved person are shown as dashed arrows towards the sensing devices, whereas the visual feedback to the person is indicated by a dashed arrow pointing away from the display arranged to provide visual feedback to the person. One or more non-invasive EEG sensor(s) is/are arranged for mounting on a person, so as to sense electroencephalogram brain waves of the person, and to output data values accordingly. The sensors may have a wired or wireless interface to output digital data, e.g. sampled at 50-1000 Hz, e.g. 100- 500 Hz, such as 250 Hz or around 250 Hz. The processor system, e.g. a computer, a tablet or a smartphone, is arranged to receive data values from the EEG sensor. E.g. via USB or another wired or wireless interface, e.g. at a sample rate of 250 Hz. The processor system comprises a processor, and an interface arranged to receive the EEG data from the EEG sensor(s). In response to the EEG data, the processor system is arranged to calculate a value indicative of a power magnitude within a limited frequency band, accordingly, e.g. as the magnitude of the EEG signal squared. The processor system is arranged to control visual output on the display. Further, the processor system is arranged to receive input from the feedback device which comprises a sensor arranged to sense a start of a predetermined body movement, and being arranged to output a timing signal to the processor accordingly. E.g. the sensor in the feedback device may be implemented as an electric switch or an optical arrangement, for detecting e.g. a wrist movement onset of the person's training body movement.

The processor in the processor system is programmed to execute a control algorithm arranged to cause the processor system to determine a baseline value indicative of a power magnitude in response to data from the EEG sensor(s) within the EEG alpha frequency band (i.e. 8-13 Hz), or another limited band including at least a major portion of the EEG alpha frequency band, in response to the person performing a predetermined body movement causing musculoskeletal pain.

The control algorithm is arranged to determine, in real-time (i.e. updated at a frequency of more than 1 Hz, preferably more than 2 Hz), values indicative of a power magnitude in response to data values from the EEG sensor within a limited frequency band, e.g. the EEG alpha frequency band (8-13 Hz). The control algorithm is further arranged to cause the processor system to display on the display a real-time visual indication of said values compared to said baseline value, during the person repeatedly performing said predetermined body movement causing musculoskeletal pain. At least one attribute of said visual indication is determined, in real-time, in response to said values being above or below a threshold which is determined in response to said baseline value. This attribute can be a color, a size of a bar or other object or symbol etc. indicating the difference from the baseline value. The control algorithm is preferably arranged to calculate a value indicative of a power magnitude in response to data from the EEG sensor averaged over a period of time comprising a time interval corresponding to a preparation phase before a start of said predetermined body movement. The preparation phase timing is calculated in response to the timing signal from the feedback device. This has been found to be a suitable EEG value for feeding back to the training person for optimal training result. This preparation phase period of time may have a length of 0-5 seconds, such as 0.2-3 seconds, such as 0.5-2 seconds, and thus the processor system preferably stores EEG sensor data and calculates the

preparation time average EEG power magnitude, once the start timing signal from the feedback device is known.

Especially, two time intervals are considered for providing the EEG average feedback value, namely 1-2 seconds prior to movement execution, and 1 second to onset of movement execution. However, this may in cases also be

individualized to a time window of 500 ms starting anywhere between 2 seconds prior to movement until movement onset. This is since each individual person will commence the movement in a slightly different manner with variations in the duration of the preparation phase. This will also be related to the complexity of the executed movement.

The preferred number of repetitions of the body movement is a minimum of 10 for the first block but up to 30. For the training blocks, which are typically 3, between 20-25 repetitions has been found to be preferred.

In a specific embodiment, EEG feedback is provided visually via the display by two bars to the person: 1) one bar showing real-time EEG alpha power level continuously, and 2) one bare only showing EEG alpha power level during execution of the painful movement. The precise timing of the feedback has been found to be an important factor for optimal performance, as discussed above in relation to selection of preparation phase time interval. The height of the bars indicate EEG alpha power magnitude. The latter is indicated as green ("good") if the alpha level is below the baseline level for initial movement execution, and red ("bad") if the EEG alpha level is above this baseline level. In subsequent training sessions, the threshold can be adjusted downwards, i.e. below the baseline level, such that the task of having a "green bar" becomes increasingly difficult. The training session can be completely automatically performed by an automated control algorithm, thus allowing the processor system to control the training session. It is expected that a patient suffering from pain has to train several times, perhaps at regular intervals, over a period of time to experience a reduce pain.

The feedback device may comprise an exercising instrument arranged to guide the person in performing the body movement. Also, the feedback device may include a forced feedback actuator can which can provide the type of movement (finger, arm, leg etc.) that causes the pain in the patient.

FIG. 8 illustrates basic steps of a method for providing neurofeedback to a person, especially during the person performing a painful movement. The method comprises receiving R_EEG a real-time EEG data indicative of

electroencephalogram brain waves of the person, calculating C_RTV real-time value indicative of a power magnitude within a limited frequency band, e.g. the EEG alpha frequency band (8-13 Hz) or a pain type specific band, in response to the EEG data. Next, determining D_BLV a baseline value indicative of a power magnitude in response to the EEG data within the limited frequency band, and displaying D_RTB a real-time visual indication of said real-time values compared to said baseline value, wherein at least one attribute, such as color, of said visual indication is determined, in real-time, in response to said values being above or below a threshold which is determined in response to said baseline value. The method may include initially receiving an input regarding the type of pain the person suffers from, and selected the limited frequency band in response thereto.

A control study with three patient was performed to evaluate whether an observed change in EEG power and pain ratings after the neurofeedback treatment were due to the treatment itself. The patients participated in three recording sessions in which they performed the following session protocol :

1) VAS pain intensity recorded

2) 20 wrist extensions, EEG continuously recorded

3) VAS pain intensity recorded 4) 1-3 minutes break

5) two repetitions of 2)-4)

FIG. 9a shows the result of this experiment, where no feedback was given to the three patients. EEG alpha power was recorded during the wrist movement, and the graph illustrates for the results for the three blocks B1-B3 on three days. FIG. 9b shows the corresponding result for the VAS pain intensity for each of the three patients which was recorded in the control study without feedback on the four days. The dashed line indicates the average level which is seen to remain constant from start to finish of the experiment. These results are opposite to what was observed with neurofeedback. When neurofeedback was provided to pain patients, their EEG alpha power decreased over the three sessions and this was associated with a decrease in the pain rating. FIG. 10 shows results of EEG alpha power for an experiment where neurofeedback was applied to the pain patient during performing the painful movement. As seen (data for one patient), a significant reduction in EEG alpha power is seen over the three sessions (each including three runs). This result is opposite to what was observed without neurofeedback (FIGs. 9a and 9b). When neurofeedback was provided to pain patients, their EEG alpha power decreased over the three sessions, and this was associated with a decrease in the pain rating, thereby indicating that the training with neurofeedback helps to reduce the patient's sensation of pain. For some types of pain, e.g. the EEG alpha frequency band is relevant to use as the limited EEG frequency band for the baseline values to be presented for the patient in the neurofeedback. However, for other types of pain, e.g. tennis elbow or other pain types, the inventors have found that the definition of frequency bands, together with the definition of time windows associated with performing the movement varies, and it is difficult to associate these definitions with a neurophysiological description. Taking this into account, and with the attempt of optimizing the neurofeedback approach, a "floating window" depending on the specific EEG time and frequency signatures of pain, and not on a fixed definition based on the literature, was identified. Data from 16 patients and 13 control participants have been analyzed in order to identify the frequency bands and time windows associated with the pain induced by the movement.

FIG. 11 shows a time-frequency analysis for EEG channel C3 for patient with tennis elbow pain. Average data from all subjects are shown. Note that the time scale is relative to start of movement, i.e. start of movement corresponds to time 0 seconds. The top graph displays data from healthy controls, the graph in the middle indicates data from patients, and the bottom graph shows the significant difference between the two groups, with larger differences indicated in dark color. These results indicate a difference in EEG power around 10 Hz, i.e. a portion of the alpha band, between controls and patients. This is always presents and therefore is likely unrelated to the movement, although still due to pain.

A difference between the two groups can also be observed in the higher frequency bands, above 25 Hz, which is part of the beta and gamma bands. This component seems to be related to the movement, as it is visible only during movement preparation and execution, around time 0 seconds.

FIG. 12 shows graphs similar to the lower graph of FIG. 11, for the same experiment, but here the results are split into the separate EEG channels monitored. Channel FP1 was used as ground. The same vertical frequency axis, horizontal time axis and color code has been used as in FIG. 11. It can be seen that the 25-30 Hz components can be observed in various EEG channel locations, but seem to be more evident in the frontal areas, i.e. in channels Fz and F4, indicated by dashed circle. EEG signals from these areas are traditionally associated with the planning of the movement. The time constant frequency band around 10 Hz is seen to be most prominent in channels C3, P3 and P4, and these are also indicated by dashed circles. Thus, based on the shown results for elbow pain, the neurofeedback system will use the relevant frequency band and time information to provide the relevant EEG based feedback to the patients, i.e. a visualization on their own power in the time and frequency band that was Identified as signature of the specific type of pain. Specifically for tennis elbow pain, a constant component (not movement related ) around i0 Hz, part of the alpha band, and a movement related above 25 Hz, part of the beta and gamma bands. These might be different for a different type of pain. Preferably, the time and frequency identification is not based on the time window (movement preparation, execution etc) of frequency bands (alpha, beta,, gamma etc.) defined by literature or being fixed,, but preferably it is oniy a portion of it or a combination of them, depending on the difference induced by the pain. This helps to provide the most effective feedback for a specific type of pain. This means that before the neurofeedback is applied, the "signature" of the specific pain will have to be identified using a group of patients and controls. Alternatively, it may be based on initial test trials on the specific patient,

FIG. 13 shows graphs indicating characteristics of the signal in the time domain have also been analyzed for persons performing a movement. When saline injection is performed, thus simulating muscleskeletal pain similar to tennis elbow pain, on heathy participants, an increased variability can be observed in the Movement Related Cortical Potentials (MRCP). This is evident in the planning phase of the movement and in the reafferentiation phase (after the movement has been executed), thus before and after time 0 in FIG. 13. Upper graph shows MRCP for the control condition, while lower graph shows MRCP results after saline injection.

FIG. 14 shows the average variability and Standard Deviation of the MRCP in all subjects of the movement planning phase of the experiments explained in relation to FIG. 13. FIG. 15 shows, still for the saline experiment, the time of peak negativity of the MRCP, associated with the initiation of the movement for the control group, the saline injected group as well as real pain patients. As seen, this was also different between healthy participants, healthy participant with induced pain (saline injection) and patients. The results indicate that the peak negativity occurs earlier in presence of pain.

Altogether, the system is advantageous, since it requires rather simple standard components: a computer and a display (e.g. only a tablet), and a low cost EEG sensor connected to the computer (tablet). The feedback device is optional, but at least a feedback device with a timing sensor for feeding back timing related to the start of the body movement, especially the preparation phase, has been found to improve efficiency of the training. In more expensive versions, the feedback device may comprise a dedicated feedback instrument arranged to help the person to perform controlled movements and optionally with a forced feedback actuator which can be controlled in response to the obtained EEG alpha power values, at least for certain types of pain, a fixed limited EEG frequency band is sufficient for efficient feedback. In general, the system may alternatively be designed to use EEG frequency-time signatures specific for a specific type of pain, thus providing a more specific feedback to a person tailored to the specific pain type the person suffers from, as explained in the foregoing. This can be

implemented in the system in several ways, but it has been found that different types of pain exhibit different EEG frequency-time signatures a more efficient training can be obtained if the feedback is based on baseline values calculated taking into account the specific type of pain. Such frequency-time signatures may be derived for a group of patients suffering from a specific pain, or alternatively, it may be derived for the individual patient based on test trials with the patient performing the painful movement and collecting EEG data in a time window covering the preparation phase, the movement itself, and a certain time after the movement. These data can be subsequently analyzed and compared to expected data for healthy persons, so as to arrive at specific limited EEG frequency bands and limited time windows where the difference from healthy persons is observed. These limited EEG frequency bands and limited time windows can then be used as baseline values for neurofeedback during training of the pain patients. To sum up: the invention provides a neurofeedback system which can be used by a person for training, so as to reduce musculoskeletal pain. The system comprises a non-invasive EEG sensor to sense EEG signal from the person. A processor system, e.g. a tablet or the like, receives data values from the EEG sensor and calculates a EEG power magnitude within the EEG alpha band accordingly. The processor system is programmed to determine a baseline value indicative of EEG alpha band power, and determine, in real-time, EEG alpha band power in response to data from the EEG sensor. On a display, the processor system displays a real-time visual indication of the EEG alpha band power values compared to said baseline value. At least one attribute, e.g color, of the visual indication is determined in response to the real-time values being above or below a threshold which is determined in response to said baseline value. In preferred embodiments, a feedback device with a sensor for sensing a start of body movement of the person provides timing feedback to the processor system to allow calculation of EEG alpha power in the preparation phase before start of the movement, and to use this as the visual EEG feedback. Specific frequency-time signatures for a specific type of pain may be used for the baseline value used in the feedback to the person during training.

Although the present invention has been described in connection with the specified embodiments, it should not be construed as being in any way limited to the presented examples. The scope of the present invention is to be interpreted in the light of the accompanying claim set. In the context of the claims, the terms "including" or "includes" do not exclude other possible elements or steps. Also, the mentioning of references such as "a" or "an" etc. should not be construed as excluding a plurality. The use of reference signs in the claims with respect to elements indicated in the figures shall also not be construed as limiting the scope of the invention. Furthermore, individual features mentioned in different claims, may possibly be advantageously combined, and the mentioning of these features in different claims does not exclude that a combination of features is not possible and advantageous.