JP6598319B2 - Rehabilitation support system - Google Patents

Description

The present invention relates to a support system for supporting the rehabilitation to restore motor function of cerebral paralyzed limb.

When cerebral infarction or cerebral hemorrhage develops, brain neurons are damaged, and after the acute symptoms are cured, sequelae such as paralysis of the limbs often remain due to brain damage.
The present inventors have previously proposed an exercise support apparatus (arm robot) that supports rehabilitation of a patient whose upper limb is paralyzed (Patent Document 1 below).
As shown in FIG. 5, this apparatus has an upper arm drive arm 11 having a fixing part 111 for fixing a patient's upper arm and a forearm drive arm 12 having a fixing part 121 for fixing a forearm. The arm 12 and the upper arm drive arm 11 are configured to move the patient's forearm and upper arm according to a set program. FIG. 6 shows a state where the arm of this arm robot is attached to a patient.

  When rehabilitating a patient's paralyzed limb using such a support device, the patient is conscious of trying to move the paralyzed part of the patient, rather than starting training passively while the patient is moved by the robot. It is known that rehabilitation is more effective when robots receive exercise support.

In consideration of such points, Patent Document 2 below discloses a rehabilitation system that drives an exercise support device when an electroencephalogram representing exercise intention is observed from a patient's electroencephalogram.
In this system, as shown in FIG. 7, motorized devices 30, 31 and a motor 32 for extending a patient's finger dynamically are mounted on the patient, and the movement intention to move the finger in response to a signal from the occupational therapist The patient is instructed to hold When the patient has a finger movement intention, an ERD signal related to the movement intention is detected from the brain wave of the patient, and this detection activates the motor 32 of the electric brace 30 to extend the patient's finger. Therefore, the peg P that has been picked falls. Further, in accordance with the driving of the electric appliance 30, electrical stimulation is applied from the muscle stimulation electrode 33 to the patient's muscle involved in finger extension.

JP 2017-153706 A Japanese Patent Laid-Open No. 2016-13182

As described in Patent Document 2, the technology of “Brain Machine Interface (BMI)” that identifies specific information from an electroencephalogram and drives a machine based on the identification information has been studied in various fields. Yes.
However, it is quite difficult to identify specific information from the brain waves of patients with brain damage.

The electroencephalogram is usually observed through a plurality of electrodes arranged on the scalp. The mounting position of the electrode on the scalp is determined by the international 10/20 method. An electroencephalogram is a recording of an impulse (action potential) due to nerve activity in the brain via multiple electrodes on the scalp. When specific nerve activity occurs in the brain, the effect is affected by the scalp. Observed at the top electrodes.
FIG. 8 shows a cross-correlation between electrodes of signals detected at each electrode when a beta wave associated with consciousness at awakening is generated in the brain of a healthy person. The thicker the line connecting the electrodes, the more similar signals are detected at those electrodes.
On the other hand, FIG. 9 shows a cross-correlation between the electrodes when a beta wave is generated in the brain of a patient with a damaged brain.

As can be seen from FIGS. 8 and 9, the brain wave of a patient with a damaged brain is significantly different from the brain wave of a healthy person, and its characteristics are difficult to grasp compared to the brain wave of a healthy person. In addition, if the degree of brain damage is different, the difference in brain waves between patients is also increased.
Therefore, it is practically difficult to identify features from spontaneous brain waves of a patient with a damaged brain and to drive a rehabilitation exercise support device based on the identification result.

The present invention has been made in view of such circumstances, and an object thereof is to provide a system that supports rehabilitation of a patient whose body and limb is paralyzed due to a brain disorder.

The present invention is a system for supporting rehabilitation of a patient whose limb is paralyzed due to a brain disorder, and an exercise support apparatus that the patient wears to receive support for movement of the paralyzed limb, and a patient An electroencephalogram data analysis apparatus that analyzes the electroencephalogram data of the brain, and a control unit that controls driving of the exercise support apparatus based on the analysis result of the electroencephalogram data analysis apparatus.
The electroencephalogram data analysis apparatus identifies an electroencephalogram acquisition unit that acquires electroencephalogram data, and motor recall data that appears when the motion of the limb part is intended from the electroencephalogram data, and the human limb part is not paralyzed. A patient who wears an exercise support device, machine-learns the boundary between the first region in which the motion recall data is distributed and the second region in which the motion recall data of a human whose limbs are paralyzed is distributed. Machine learning means for outputting information on a region to which the motor recall data identified from the electroencephalogram data belongs, or information indicating a distance between the motor recall data and the boundary line.
The control unit instructs driving of the exercise support apparatus when the movement recall data of the patient wearing the exercise support apparatus is within a predetermined distance from the boundary line or in the first region.
In this system, when data similar to the motion recall data of a healthy person is detected to some extent from the spontaneous brain waves of a patient who wears the motion support device and performs function recovery training, the motion support device is driven. Therefore, based on the patient's willingness to exercise, the paralyzed limb moves with the support of the exercise support device, and the effect of rehabilitation is improved. Further, by continuing such training, the patient can output a spontaneous electroencephalogram capable of detecting motor recall data equivalent to that of a healthy person.

In the rehabilitation support system of the present invention, a support vector machine (SVM) is used as the machine learning means.
SVM uses the brain waves of a large number of healthy subjects and patients as sample data, analyzes the position of the electrode where the signal changes when recalling the motion from the brain waves of the healthy subject, and represents the healthy person represented by the signal of the electrode at that position. Machine learning is performed on the boundary line that separates the motor recall data and patient motion recall data. When the brain wave of the patient wearing the exercise support device is input, the distance between the patient's exercise recall data and the boundary line is calculated and output.

  In the rehabilitation support system of the present invention, the brain wave data of a healthy person may be used as SVM sample data, or instead, the brain wave data on the non-paralyzed side of a hemiplegic patient may be used.

Further, in the rehabilitation support system of the present invention, the control unit is provided with a storage unit for storing a manual motion when the therapist moves the paralyzed limb portion of the patient wearing the exercise support device with a hand, When driving the exercise support device based on the patient's willingness, it is desirable to reproduce the manual exercise stored in the storage unit.
Based on the patient's voluntary intention, exercise suitable for the patient's symptoms is performed with the support of the exercise support device.

  In the rehabilitation support system of the present invention, the arm robot shown in FIG. 5 can be used as the exercise support device.

  According to the present invention, it is possible to effectively perform rehabilitation of a patient whose body limb is paralyzed due to a disorder in the brain.

The block diagram which shows the structure of the rehabilitation assistance system which concerns on embodiment of this invention. The figure explaining operation of SVM The flowchart which shows the procedure of the rehabilitation assistance method which concerns on embodiment of this invention. Diagram showing an example of manual exercise therapy The figure which shows the arm robot which supports upper limb movement The figure which shows the state which mounted | wore the patient with the arm robot of FIG. Diagram showing a conventional rehabilitation system Diagram showing cross-correlation of EEG data detected by each electrode (Healthy person) Diagram showing the cross-correlation of EEG data detected by each electrode (Brain disabled)

FIG. 1 is a block diagram showing the configuration of a rehabilitation support system according to an embodiment of the present invention.
This system drives an arm robot 10 based on an analysis result of an exercise support apparatus (arm robot) 10 that supports rehabilitation of a patient with paralyzed upper limbs, an electroencephalogram analysis apparatus 20 that analyzes an electroencephalogram, and the electroencephalogram analysis apparatus 20. And a control unit 30.
The arm robot 10 is shown in FIG. 5, and the upper arm driving arm 11 having a fixing part 111 for fixing the upper arm, the forearm driving arm 12 having a fixing part 121 for fixing the forearm, and not shown in FIG. 5. However, based on an instruction from the control unit 30, the drive unit 13 that drives the upper arm drive arm 11 and the forearm drive arm 12, and the exercise therapy information that acquires information on the movement of the upper arm drive arm 11 and the forearm drive arm 12 as exercise therapy information. And an acquisition unit 14.

The electroencephalogram analysis apparatus 20 includes an electroencephalogram information acquisition unit 21 that acquires an electroencephalogram via an electrode disposed on the scalp, an electroencephalogram information recording unit 22 that records the acquired electroencephalogram, and an electroencephalogram information that analyzes the recorded electroencephalogram information. And an analysis unit 23. The electroencephalogram information analysis unit 23 will be described in detail later.
In addition, the control unit 30 includes an exercise therapy information recording unit 31 that records exercise therapy information sent from the exercise therapy information acquisition unit 14 of the arm robot 10, and the speed and power of joint motion from the recorded exercise therapy information. The exercise therapy information analysis unit 32 that analyzes the direction of movement, the drive determination unit 33 that determines whether to drive the arm robot 10 based on the analysis result of the electroencephalogram information analysis unit 23, and the drive determination unit 33 include the arm robot 10 A movement therapy reproduction control unit 34 that causes the drive unit 13 to reproduce the movement of the arm robot 10 analyzed by the movement therapy information analysis unit 32.

The electroencephalogram information analysis unit 23 of the electroencephalogram analysis apparatus 20 corresponds to “machine learning means” in the claims, and performs machine learning using a support vector machine (SVM) to analyze the electroencephalogram. The electroencephalogram information analysis unit 23 is realized by a computer executing processing defined by a program.
For machine learning of the electroencephalogram information analysis unit 23, first, an electroencephalogram information acquisition unit 21 acquires an electroencephalogram when a large number of healthy persons imagine the movement of the upper limb (or when an actual movement of the upper limb is performed). Is done.
The electroencephalogram information analysis unit 23 learns the positions of a plurality of electrodes having a high cross-correlation, in which a signal change appears when the healthy person recalls the upper limb movement from the large number of electroencephalograms. Then, from the brain wave when the healthy person recalls the upper limb motion, the motion recall data of the healthy person whose component is the power spectrum of the brain wave signal detected by these electrodes is obtained.
Similarly, the electroencephalogram information analysis unit 23 learns by using, as sample data, a large number of electroencephalograms when a person with a brain disorder imagines upper limb movement (or when an actual movement of the upper limb is performed). Motor recall data of “a person with a brain disorder” having the power spectrum of an electroencephalogram signal detected by the electrodes as a component is obtained.

Next, the electroencephalogram information analysis unit 23 performs machine learning on a boundary line that separates the motor recall data of the healthy person from the motor recall data of “a person with a brain disorder”.
Now, it is assumed that the motion recall data has as its component the power spectrum of an electroencephalogram signal detected by two electrodes having a high cross-correlation.
In this case, the motion recall data of a healthy person and a “human with a brain disorder”, as shown in FIG. 2, the X axis represents the magnitude of the power spectrum component of the electroencephalogram signal detected by one electrode, and Y The axis can be plotted in a feature space representing the magnitude of the power spectrum component of the electroencephalogram signal detected by the other electrode.

In order to obtain a boundary line that separates the motion recall data of the healthy person displayed in the feature space and the motion recall data of the “human with a brain disorder”, the electroencephalogram information analysis unit 23 obtains a margin (subtracted boundary line and data). Set the boundary line so that the shortest distance between
Here, a two-dimensional feature space has been described as an example. However, if the number of electrodes having high cross-correlation increases, the number of dimensions of the feature space increases accordingly, and the number of electrodes may be one. For example, the feature space becomes a one-dimensional space.

After performing such pre-processing, the electroencephalogram information analysis unit 23 acquires the electroencephalogram of the patient attached to the arm robot 10, and the patient's motor recall using the power spectrum of the electroencephalogram signal detected by the learned electrode as a component. Data is calculated and plotted in the feature space. Then, the patient's motion recall data is compared with the boundary line, and the patient's motion recall data is in the first region where the motion recall data of the healthy person in the feature space exists, or “a person with a brain disorder” It is identified whether the motion recall data is in the second region where it exists. If it is in the second region, the shortest distance between the patient's motion recall data and the boundary line is calculated.
The electroencephalogram information analysis unit 23 transmits the identification result and information on the shortest distance from the boundary line to the drive determination unit 33 of the control unit 30.

When the patient motion recall data is in the first region of the feature space, or when the shortest distance from the boundary line is less than the predetermined distance, the drive determination unit 33 moves the arm robot 10 to the motion therapy reproduction control unit 34. Instruct to drive.
Receiving this instruction, the exercise therapy reproduction control unit 34 causes the drive unit 13 of the arm robot 10 to reproduce the movement of the arm robot 10 analyzed by the exercise therapy information analysis unit 32.

The flowchart of FIG. 3 shows the procedure of exercise support by this system.
Using the electroencephalograms of many healthy subjects as sample data, identify the electrodes where EEG changes appear when you intend to move the upper limbs, and set the EEG data detected by those electrodes as motor recall data for upper limb movements (Step 1). Collect a lot of motor recall data of healthy persons and motor recall data detected by the electrodes when a person with paralyzed upper arms (paralyzed person) intends to move the upper limbs. Plotting is performed in the feature space to obtain a boundary line that separates the motion recall data of the healthy person and the motion recall data of the paralyzed person (step 2).
Next, in order to support the rehabilitation of a patient whose upper limb is paralyzed, the arm robot 10 is attached to the patient (step 3), and manual exercise therapy is performed in which the therapist holds and moves the patient's upper limb. FIG. 4 shows the situation at this time. At this time, information on the movement of the arm robot 10 is acquired by the exercise therapy information acquisition unit 14 and recorded in the exercise therapy recording unit 31 of the control unit 30 (step 4).

Next, the brain wave of the patient wearing the arm robot 10 is measured (step 5).
From the electroencephalogram when the patient intends to move the upper limb, the patient's motion recall data detected by the electrodes specified for motion recall data is measured. Then, the relationship between the patient motion recall data and the boundary line obtained in step 2 in the feature space is analyzed by the electroencephalogram information analysis unit 23 (step 6).
If the patient's motion recall data is in the first region where the motion recall data of healthy persons in the feature space exists (Yes in step 7), motion support by the arm robot 10 is started (step 10). If the patient's motion recall data is not in the first region (No in step 7), it is analyzed whether the shortest distance between the patient's motion recall data and the boundary line is less than the threshold, and if it is less than the threshold (step 8). Yes), exercise support by the arm robot 10 is started (step 10).

  If the shortest distance between the patient's motion recall data and the boundary is not less than the threshold (No in step 8), the patient is instructed how to image the upper limb to move, or in step 8. After adjustment to lower the threshold (step 9), the procedure from step 5 is repeated.

In the case of Yes in step 7 or step 8, the exercise therapy reproduction control unit 34 of the control unit 30 causes the arm robot 10 to reproduce the manual exercise therapy recorded in the exercise therapy information recording unit 31 by the arm robot 10. The drive unit 13 is controlled.
When the time or number of manual exercise therapy reaches a specified value (Yes in step 11), the procedure performed in the upper limb exercise therapy is recorded (step 12), and the process is terminated. If step 11 is No, the procedure from step 5 is repeated.

  As described above, this system is configured to drive the arm robot 10 when data similar to the motion recall data of the healthy person is detected to some extent from the brain waves of the patient wearing the arm robot 10. When the arm robot 10 is driven, the patient will experience a simulated experience that the upper limb has moved as expected. Accumulation of such experiences enhances the effectiveness of rehabilitation and creates a virtuous cycle in which motor recall data detected from the patient's brain waves is closer to that of healthy individuals.

The threshold value in step 8 may be reduced according to the progress of rehabilitation.
Further, in step 6, the motor recall data of the healthy person and the motor recall data of the patient are described as being composed of electroencephalogram data detected by electrodes at the same position. For example, in FIG. 9, when the motor recall data of a healthy person is composed of brain wave data detected by the electrodes F4 and C4 of the right hemisphere, the patient's motor recall data is also detected by the electrodes F4 and C4. It was explained that it is a component.
However, the patient's motion recall data includes not only motion recall data whose components are brain wave data detected by the electrodes F4 and C4 in the right hemisphere, but also brain waves detected by the corresponding electrode F3 and electrode C3 in the left hemisphere. You may include the movement recall data which uses data as a component.
This is because there is a phenomenon in which the function of the brain on one side lost due to cerebral infarction or the like is “shouldered” by the brain on the other side.

In addition, instead of the brain waves of healthy persons in steps 1 and 2, the brain waves on the non-paralyzed side of the hemiplegic patient may be used as sample data.
Here, rehabilitation of a patient whose upper limb is paralyzed has been described, but the present invention can also be applied to rehabilitation of a patient whose other limbs are paralyzed.

  INDUSTRIAL APPLICABILITY The rehabilitation support system and the support method of the present invention can effectively perform rehabilitation of a patient whose limb is paralyzed due to a disorder in the brain, and can be widely used in rehabilitation facilities, hospitals, and the like.

10 Exercise support device (arm robot)
DESCRIPTION OF SYMBOLS 11 Upper arm drive arm 12 Forearm drive arm 13 Drive part 14 Exercise therapy information acquisition part 20 EEG analysis apparatus 21 EEG information acquisition part 22 EEG information recording part 23 EEG information analysis part 30 Control unit 31 Exercise therapy information recording part 32 Exercise therapy information analysis 32 Unit 33 drive determination unit 34 exercise therapy reproduction control unit 111 fixing unit 121 fixing unit

Claims (6)

A system that supports rehabilitation of patients who have paralyzed limbs due to brain damage,
An exercise support device worn by the patient to receive support for exercise of the paralyzed limb part; and
An electroencephalogram data analysis apparatus for analyzing the electroencephalogram data of the patient;
A control unit for controlling the driving of the exercise support device based on the analysis result of the electroencephalogram data analysis device;
With
The electroencephalogram data analysis apparatus comprises:
An electroencephalogram acquisition unit for acquiring electroencephalogram data;
From the brain wave data, identifying motor recall data that appears when the movement of the limb part is intended, a first region in which the motion recall data of a human whose limb part is not paralyzed is distributed, The movement recall data identified from the brain wave data of the patient wearing the exercise support device, by dividing the second region in which the movement recall data of the human with paralyzed limbs is distributed and machine learning the boundary line Machine learning means for outputting region information or information indicating the distance between the motion recall data and the boundary;
Have
The control unit instructs to drive the exercise support device when the exercise recall data of the patient wearing the exercise support device is within a predetermined distance from the boundary line or in the first region. Rehabilitation support system. The rehabilitation support system according to claim 1,
A rehabilitation support system, wherein the machine learning means is a support vector machine (SVM). The rehabilitation support system according to claim 2,
The SVM is a rehabilitation support system in which the boundary line is machine-learned using brain wave data of a healthy person and human brain wave data in which the body part is paralyzed. The rehabilitation support system according to claim 2,
The SVM is a rehabilitation support system in which the boundary line is machine-learned by using the electroencephalogram data of the non-paralyzed side of the hemiplegic patient as the electroencephalogram data of the human body part that is not paralyzed. The rehabilitation support system according to claim 1,
The control unit includes a storage unit that stores a manual motion when a therapist moves a paralyzed limb portion of a patient wearing the exercise support device with a hand, and the motion recall data of the patient Is within a predetermined distance from the boundary line, or is in the first region, the rehabilitation support system that instructs the exercise support device to reproduce the manual movement stored in the storage unit . The rehabilitation support system according to claim 1,
The exercise support apparatus includes: an upper arm drive arm that is attached to the upper arm of the patient whose upper limb is paralyzed; a forearm drive arm that is attached to the forearm of the patient; and a drive unit that drives the upper arm drive arm and the forearm drive arm. , A rehabilitation support system.