シリンダとフットレストの幾何学的関係から, 要求仕様である ストロークと推力を満足する シリンダを選定した.要求仕様は 足関節リハビリ機器のフットレストの回転軸から背屈方向に20度回転するのに必要なストロークと,女性が手で押す力の平均値である40Nの力[*]を足裏に加えることが出来る推力とした.足関節可動角度,推力と足裏にかかる力の関係を計算値,実測によって調査し比較した. Figure 4に計算で使用するパラメータを示す.Nは足裏反力,Fはシリンダ推力,Rはフットレストの回転軸から計測位置までの距離,rはフットレストの回転軸からシリンダロッドとフットレストの取り付け部,pはシリンダ取り付け部からフットレストの回転軸までの距離,qはシリンダ長,θはqとrのなす角とする.ここで,R, r, pはフットレストが回転しても不変であり,既知のパラメータである.
Fig. 4 Parameters for calculation
式7にθを余弦定理より求める算出式を,式8にモーメントのつり合いにより足裏反力Nを求める式を示す.
\(\cos\theta=\frac{q^2+r^2-p^2}{2qr}\) (7)
\(N=\frac{F\sin\theta\cdot r}{R}\) (8)
シリンダのストロークはこの幾何学的な関係から押し方向では,43.6㎜,引き方向では98.2mmのストロークが必要であることが明らかになった.また,シリンダに必要な推力は 本来ならば, 押し方向と引き方向双方に40N必要なため復動型シリンダを選定すべきところであるが,本研究において,脳梗塞患者の足関節の 痙縮 を対象としていること,小型軽量化による可搬性を有する,という点から,今回は背屈動作のみを繰り返す単動型シリンダ (M24D150.0,Airpel製) とした.このシリンダは,印加圧力0.20MPa時の推力は 理論(計算)値で40Nとなり,実験によって充分満足できる推力を発生していることを確認している.
3.3 Design for control system
足関節リハビリ機器が理学療法士のリハビリ動作を制御設計によって実現するために,位置制御から力制御に切り替える制御系設計を行った.この制御は,健常者や足関節の拘縮が弱い使用者には位置制御により目標の訓練角度を繰り返し行い,拘縮により位置制御を行っても目標角度に達しない患者には位置制御から力制御に切り替える制御となっている.Figure 5にこの制御のフローチャートを示す.
Fig. 5 Flowchart of switching control
まず始めにプログラムを起動すると,目標角度になるように位置制御を行う.ここで目標角度に達するとそのまま位置制御を継続し設定時間の間その角度を維持する.5秒経過しても目標角度に達しないと力制御に切り替わるように設定されている.力制御部では位置制御時にかかっていた力の10%上乗せした力がかかるように加圧させ,理学療法士の足関節が動かなくなったところで少し力を加えるという動作を模している.リハビリ機器は位置制御,力制御ともにPID制御により実現するためP,I,Dのパラメータを決定し,切り替え制御を検証する制御実験を行った.足関節リハビリ機器にストッパを用いて角度制限を行い,目標角度に到達しない場合を再現した.今回の実験では,角度制限時と非制限時の違いを確認しやすいように,位置制御での目標角度を-45度から+20度までとし,0度の位置で角度制限を行った.また,計測開始から1秒後に動作するように設定した.角度非制限時の結果をFigure 6に,角度制限時の結果をFigure 7に示す.
Fig. 6
Fig. 7
Figure 6より,目標角度に到達すると,そのまま継続して位置制御が行われていることが確認できる.また,Figure 7において,角度制限により位置制御を行っても1秒から6秒の間に目標角度に到達できず,6秒以降には位置制御時の空気圧の増加が確認でき,位置制御から力制御に切り替わっていることが分かる.実際の計測結果において,角度制限時での1から6秒までの印加圧力が0.072MPaであり,足裏反力が計算値で29.6Nとなるのに対し,6秒以降の印加圧力が0.08MPa,足裏反力は32.9Nと計算されることから,力制御に切り替わることで足裏にかかる力が10%増加していることが確認できる.
4. Ankle neuro rehabilitation system.
4.1 System setup
Figure 8 demonstrates the ankle neurorehabilitation system.
Fig. 8 Ankle neuro rehabilitation system
The subject the head cap that contains Ag-AgCl active electrodes that connect to the amplifier (Active Two, Biosemi Inc., Netherland) with which electroencephalogic (EEG) signal was measured from the brain scalp. The task was to employ motor imagery of the right ankle by 30 degrees. The EEG signals during the task as well as resting period were to be detected. The EEG signals were sent to the amplifier where A/D conversion was employed. The signal was sent through USB receiver (via optic fiber) to PC, where data were processed by LABVIEW (LABVIEW2015, National Instruments Inc. USA). The sampling frequency was 2048Hz. Fuzzy Template Matching (FTM) algorithm was employed and optimal value was obtained (FTM learning). When the FTM processed EEG signal surpassed the threshold, a triggering signal was generated, which was sent to data acquisition device (DAQ: USB-6000, National Instruments, USA), and sent by the trigger sending device (MARQ, Kissei Com. Japan, sampling frequency 2048Hz) to the Trigger Receiver. This receiver was connected to another PC where the software (Matlab2016, Mathworks, USA) processes controlled triggering signals. Here, the triggering signal sends commands where the rehabilitation device operates the pullup by 30 degrees for a duration of 20 seconds. Note thresholds were set by eye, initially to lower to detect motor imagery signals, and refined to exclude weaker signals of FTM signals of resting EEG.
The rehabilitation device is activated by the air pressure that is generated by compressed air supply through the air-compressor (YC-4, Yaezaki-Kuuatsu Inc.) to the electro-pneumatic regulator (RTR-200-1, Koganei Inc.). Control signals were sent through I/O board (MF634, HUMUSOFT). The rotation angle was measured by the potentiometer (SVO1, Murata Manufacturing Co. Ltd). Data from the potentiometer were sent through the I/O board to the PC. The system continues to be activated for a certain time (20 seconds) once it obtains a triggering signal. Only after activation is finished, the next triggering signal becomes effective.
4.2 Experimental procedure
Five volunteers (age: 22 to 22 years old, all healthy male students on the campus) participated in the experiment. The experiment was followed by the Kwansei Gakuin University regulation of ethics for the Protection of Human Subjects of Medical Research, which was approved by the campus committee. Informed consent was acquired by each subject following the Declaration of Helsinki. The subject was seated on the chair in a relaxed position with the right leg on the foot rest (the ankle bended in a natural position). The left leg was kept on the floor. Electroencepalogic signal was detected by the Active Two system, in which eight electrodes (F3, F4, C3, Cz, C4, P3, Pz, and P4) were used based on the conventional 10-20 method. Only alpha (8~12Hz) and beta bandwidth (13~30Hz) was used for FTM detection.
4.3 Experimental protocol
Using the above system, first we assured that the signal of motor imagery of ankle uplift movement was indeed detected, which in turn operated the machinery. The experiment was consisted of by three stages:
(1) development and practice of motor imagery for the participant, (2) system tune-ups to determine and parameter learning process of motor imagery EEG, and (3) the testing stage for the motor imagery EEG to use it for activating machinery. As the motor imagery development, first voluntary movement of right ankle uplift was employed (voluntary task: 30 trials). (I) Motor imagery of the voluntary movement followed immediately after the voluntary movement (imagery task: 30 trials). Furthermore, the machine produced autonomous ankle movement that simulates human rehabilitation behavior (autonomous movement task: 30 trials). Each task was initiated by a trigger LED. Each trial consisted of by four seconds period. EEG was measured during this session. (II) The second process of system tune-ups consisted of the two sub-steps (a) EEG signals of the motor imagery was detected (2minutes). As a comparison, a baseline (resting) EEG signal was measured during the resting period (2 minutes). (b) Fuzzy Template Matching Algorithm was applied to determine parameters in order to discriminate the two states of motor imagery EEG and the resting EEG. The calculation of FTM took 60 seconds. (III) Finally, as the third step, the FTM logic was tested by detecting a real motor imagery EEG. A trigger signal was to be generated as the output FTM signal. Therefore, as soon as the system detected the motor imagery EEG, the triggering signal was supposed to be generated, which was to be transferred to the robot system. In order to test if this is true, two tasks were employed: (1) the participants were asked to employ motor imagery when the LED light was illuminated. (2) the participants were asked to rest (rest task) to test the resting EEG signal was valid.
As the EEG independent component analysis, EEGLAB was used. The EEG signal was first notch filtered (60+/-0.5Hz as well as 120+/-0.5Hz bandwidth) to exclude artifacts from electrical current, which was in turn band-pass filtered with the frequencies between 0.04 to 200 Hz (FIR filtering). Note spatial care was taken to minimize the notch filter’s artifact in determining an optimal bandwidth.
4.4 Results
Figure 9 depicts the FTM outputs and thresholds, triggering position, and rotation angle of the system.
Fig. 9 Output value and response of ankle rehabilitation devise
When the output value of FTM surpasses the threshold, a trigger was generated, and the rehabilitation device was operated. Note the system was programmed as such that the rotation angle was kept constant for 20 seconds once the trigger is generated. Extra triggers during this period was neglected. After the 20 second period, another trigger could successfully operate the system.
次に一般に行われている患者の意思とは関係なく行われる他動運動の状態と本研究で開発されたL-FTMによるリハビリ訓練を行った状態をICA解析の結果をトポグラフ表示 により比較を行う. この表示は脳の賦活している箇所が赤く表示されている.
Results of ICA analysis indicates that overall EEG power was stronger during the test period when ankle rehabilitation system using L-FTM was employed during motor imagery (Figure 10 upper one) as compared to the period during the passsive motion task (Figure 10lower one). Namely, the activity around Cz was more conspicuous the movement with motor imagery. This demonstrates that the brain signals of EEG increased with motor imagery, which was utilized to the rehabilitation system to be operated.
(他動運動との比較にしました.運動野(C3, Cz)のみに絞ってカウントし比較するのはどうでしょうか.下は田中先生案です.)
Results of ICA analysis indicates that overall EEG power was stronger during the test period when autonomous machine movement was employed during motor imagery (Figure 10 upper one) as compared to the period during the autonomous movement task (Figure 10lower one). Namely, the activity around Cz was more conspicuous the movement with motor imagery. This demonstrates that the brain signals of EEG increased with motor imagery, which was utilized to the rehabilitation system to be operated.
Fig.10 ICA results
(五人分のデータを表にして、その内容を日本語で記述してください by田中)
報告書:FTM を用いた足関節ニューロリハビリテーションの検証実験 参考
>運動野(C3, Cz)のみ絞って議論する.(時間もないので 現状のデータから)・・・後で差し替える.
5. Conclusions
足関節を対象とした従来とは異なり特定の脳波特徴を検出しなくても運動意志を読み取りリハビリテーション機器が動作するニューロリハビリテーションシステムを開発した.
・ERD,ERSを検出せず,脳波のα波,ベータ波を対象にして,単純に脳波パワーを'High'か'Low'かのみの2つのラベルを基にしたL-FTMを用いて運動意図の検出可能なことが確認できた.また,訓練を必要とせず,運動意図の検出に,およそ1時間以内という短時間で実現できた.
・空気圧シリンダを用いた足関節リハビリ機器を開発し,拘縮による関節剛性(硬さ)の度合いに応じて,位置(角度)制御から力制御に切り替わる制御系を設計,目標動作が実現できていることが確認できた.
・L-FTMで検出された運動意図をトリガ信号にして,足関節リハビリ機器が動作し,運動に関係し脳波が賦活される部位を確認できた.
今後は,より精度を高める手法の検討と実際の脳梗塞患者の方への実証実験を進めていきたい.
Acknowledgements
This work was supported partially by MEXT(The Ministry of Education, Culture,Sports,Science and Technology)-Supported Program for the Strategic Research Foundation at Private Universities, 2014-2018(Grant No. S1411038).