System design
The general framework for the hindlimb gait control system based on CPG for rats with SCI is shown in Fig. 7. The initial value of the stimulator is provided according to the stride length and frequency set by the host computer. The stimulator provides positive and negative pulse signals to the one-port electrode, which is connected with the CPG site, this induces bilateral gait movement of the hindlimbs of the rats. The unilateral hind limb gait of the rats was measured by detecting the knee joint angle of the hindlimb of rats, and the data were displayed on the screen. In line with the model of stimulating signal parameters for stride length and frequency, when the knee joint angle is extremely large or small, the pulse amplitude is reduced or increased. When the step frequency is extremely high or low, the stimulation signal period is lengthened or shortened resulting in automatic control of hindlimb movement in rats.
The stimulator utilizes an STM32F103 and communicates with the host computer by ZigBee[16]. STM32F103 with a voltage output digital-analog converter (DAC). The 12-bit digital input DAC, configured by the system, uses two timers to control the voltage amplitude of output signal. The voltage follower and subtraction circuit, composed of two LM358Ns, generate positive and negative pulse signals to stimulate the CPG site. To obtain the relationship model between the gait angle data and signal parameters, the stimulator is equipped with buttons to control the signal parameters and record angle data. The knee joint angle information was obtained using two JY901 sensors, and the data transmitted to the stimulator via a serial port. The JY901 sensor, integrated with high-precision gyroscope, derives the motion posture in real time[17]. In a dynamic environment, the measurement angle error is only 0.1°, which meets the requirements of knee joint angle measurement. The stimulation electrode is either in contact with the epidermis or is oxidized when stimulating the CPG site; thus the impedance measurement part is added to the system. As the impedance measurement and stimulation are both required at the same key CPG site, a single tungsten wire electrode is used as an impedance measurement probe and a stimulation electrode to achieve a dual-purpose function. In the circuit design, a relay is used to make or break the connection between the impedance measurement circuit and the stimulation circuit and thus prevent the influence of the stimulation circuit on impedance measurement[18]. An AD5933 was used as the impedance measurement chip, and a two-point calibration method was used to measure impedance. The starting frequency was 29,000Hz, the number of scanning points was 16, and the frequency increment was 62.5Hz. Each frequency point was repeated four times to obtain an average value of the real and imaginary parts, and thus obtain the corresponding amplitude and phase. The calculated impedance value displayed on the screen[19]. According to the range of the spinal cord impedance, the calibration resistance was set to 1\(K\Omega\)[20].
The operation of the system is divided into two parts. First, confirming that the impedance is within the set threshold, the system shows the green light and selects the initial stimulus signal according to the preset stride and frequency. Second, a set of gait angle information is detected. The hindlimb movement is automatically adjusted according to the mapping model of the signal and angle. In the first part, if the impedance is not within the set range, a red light prompts the user to check the problem. In the second part, when the automatic adjustment of the hind limb fails to reach the reasonable range of the reference frame, the stimulator is stopped, indicating that the rats may have muscle fatigue.
Amplitude-angle model
Three rat subjects with fixed electrodes were used to build the stimulation model. The gait is derived by measuring the knee joint angle of the unilateral hindlimb of SD rats. The sensor position and knee joint angle θ are shown in Fig. 8. The correlation between the stimulus signal voltage amplitude and gait angle data is converted into a mapping relationship. The CPG site was stimulated by pressing the button to change the voltage amplitude of the stimulus signal, and the knee joint angle of the unilateral hindlimb was measured to establish the knee joint angle and signal parameter model[21–23].
During the stimulation experiment on SCI rats, it was observed that by changing the amplitude of stimulation signal in 0.161 V steps, the movement of hindlimbs can be regulated more smoothly and accurately. When the amplitude of the stimulus signal is too weak then no movement can be observed. When the amplitude is extremely strong, precise regulation cannot be achieved. For voltage amplitudes between 1.289–2.255V, the gait changes of the hindlimbs of rats is clearly observed, which is in accordance with the closed-loop control of the system.
In order to get the relationship between amplitude parameters of stimulus signal and knee joint angle, the method of separating variables is used[24]. The mapping between the different voltage amplitudes and knee joint angles are shown in the Fig. 9, Fig. 10. When the CPG of anesthetized SCI rats was not stimulated, the knee joint angle of different rats was generally stable at between 118 ° and 122 °. In Fig. 9, the positive pulse amplitude is constant at 1.289 V, and the negative voltage changes from − 1.289 V to -2.255 V in step of 0.161 V. Table 1 shows the minimum knee joint angle of each rat and the mean of the three rats. In Fig. 10, the negative pulse amplitude constant at minus 2.255 V, and the positive voltage changes from + 1.289 V to + 2.255 V in step of 0.161V. Table 2 shows the maximum knee joint angle of each rat and the mean of the three rats. The experimental analysis showes that the positive and negative pulse amplitudes have different effects on the angle change.
Table 1
Minimum and mean values of the knee joint angles of the three rats with the increase of negative voltage
|
电压(V)
|
-1.289
|
-1.450
|
-1.611
|
-1.772
|
-1.933
|
-2.094
|
-2.255
|
编号
|
|
①
|
110.41°
|
107.32°
|
106.45°
|
104.60°
|
101.28°
|
98.63°
|
97.64°
|
②
|
109.52°
|
107.61°
|
106.69°
|
104.39°
|
102.43°
|
99.87°
|
97.53°
|
③
|
109.28°
|
108.56°
|
105.06°
|
102.87°
|
99.23°
|
97.86.°
|
96.37°
|
Mean
|
109.73°
|
107.83°
|
106.06°
|
103.95°
|
100.98°
|
98.78°
|
97.18°
|
Table 2
The maximum and mean values of the knee joint angles of the three rats with the increase of positive voltage
|
电压(V)
|
+ 1.289
|
+ 1.450
|
+ 1.611
|
+ 1.772
|
+ 1.933
|
+ 2.094
|
+ 2.255
|
编号
|
|
①
|
126.32°
|
126.85°
|
128.13°
|
129.94°
|
130.9°
|
131.32°
|
131.61°
|
②
|
127.13°
|
127.76°
|
128.82°
|
129.62°
|
130.35°
|
130.93°
|
131.42°
|
③
|
126.95°
|
128.19°
|
129.35°
|
129.54°
|
130.25°
|
130.75°
|
131.17°
|
Mean
|
126.8°
|
127.6°
|
128.8°
|
129.7°
|
130.5°
|
131.0°
|
131.4°
|
The experimental results show that the positive and negative pulse amplitudes have different effects on the angle variation. As shown in Fig. 11, β represents the angle variation range of the knee joint of rats, Step represents the amplitude of the pulse signal to control the Step size, α represents the initial angle of rats without stimulation, θ1 represents the angle variation range of positive impulse stimulation, and θ2 represents the angle variation range of negative impulse stimulation. The relationship between parameters is shown as follows. The relationship model between amplitude and angle is established.
The variation range of knee joint angle β ∈ (96°~132°),
Step size of amplitude regulation Step = 0.161V,
The initial angle α∈ (118°~122°),
θ 1 is the control range of positive pulse change,
θ 2 is the control range of negative pulse change,
Under stimulation by a positive pulse signal, the right hindlimb is extended, with knee joint θ1 was approximately 1 °. Under stimulation by a negative pulse signal, the right hindlimb is flexed, with knee joint θ2 was approximately 2 °. The angle of the knee joint gradually returned to normal when the stimulation was removed.Relationship model between amplitude and knee joint angle is is shown in Eq. 2 − 1 and 2–2:
θ 1 = α + γ1 ± Nstep (γ1 is the angle of extension of hind limbs) (2 − 1)
θ 2 = β + γ2 ± 2Nstep (γ2 is the angle of flexion of the hind limbs) (2–2)
The results showed that the stimulation effect of negative pulse on CPG site was more obvious[25], and the range of positive and negative pulse signals between 1.289 and 2.255V maintained the range of knee joint angle between 96° and 132° when the CPG site was stimulated.
The reference angle of the rat hind limb movement automatically controlled by the system was set reasonably according to the actual knee angle. The actual angle was obtained by stimulating the CPG of SCI rats with different stimulus signals. At the same time, the sampling frequency was set to 10Hz to enable comparition of the actual angle and the reference angle [26]. The knee joint angle of SCI rats was automatically adjusted according to the reference angle set by the voltage amplitude and interval time. The reference angle is expressed as piecewise functions namely positive pulse stimulation, negative pulse stimulation, and stimulus interval. A reasonable reference angle range was set according to the actual movement angle range of the CPG stimulation. There are many types of knee joint angle changes, as shown in Table 3 and Table 4. These show knee joint angle forms of reference gait change, which correspond to positive and negative pulse stimulation, respectively, and the corresponding reference frame function is synthesized. The lowest stimulation period of the system is 2.1 s(There are 20 positive pulse waveforms with a pulse width of 200us and 20 negative pulse waveforms with a pulse width of 200us. The pulse interval is 30ms, and the minimum interval of positive and negative pulse is 0.9s).
Table 3
The coordinate of positive pulse reference system is taken
t/s
|
kT + 0.8
|
kT + 0.9
|
kT + 1
|
kT + 1.1
|
kT + 1.2
|
kT + 1.3
|
kT + 1.4
|
kT + 1.5
|
kT + 1.6
|
kT + 1.7
|
106/127
|
120
|
121
|
123
|
126
|
127
|
126
|
122
|
122
|
121
|
119
|
102/129
|
118
|
122
|
125
|
127
|
129
|
128
|
121
|
120
|
119
|
121
|
96/132
|
119
|
121
|
123
|
126
|
132
|
129
|
127
|
122
|
121
|
120
|
Table 4
coordinates of negative pulse reference system
t /s
|
kT + 1.7S + T’
|
kT + 1.7S + T’+0.1
|
kT + 1.7S + T’+0.2
|
kT + 1.7S + T’+0.3
|
kT + 1.7S + T’+0.4
|
kT + 1.7S + T’+0.5
|
kT + 1.7S + T’+0.6
|
kT + 1.7S + T’+0.7
|
kT + 1.7S + T’+0.8
|
kT + 1.7S + T’+0.9
|
106/127
|
120
|
118
|
107
|
107
|
106
|
107
|
108
|
110
|
115
|
120
|
102/129
|
122
|
115
|
108
|
104
|
102
|
104
|
108
|
116
|
118
|
119
|
96/132
|
121
|
118
|
110
|
101
|
96
|
99
|
102
|
115
|
117
|
118
|
The reference frame is fitted in the first period range. When entering the gait control range of the second period, the collection time of the system is set as the initial value, which is equivalent to continuing to cycle the control with the reference angle of the first period.
In the first cycle (K = 0), the fitting function of knee joint angle under positive and negative pulse stimulation is shown in Eq. 2–3.
λ 2 = f (x) = a1*sin (b1*x + c1) + a2*sin (b2*x + c2) + a3*sin(b3*x + c3) (2–3)
When the reference frame of 106 ~ 127° is selected, the positive pulse fitting function is shown in Fig. 12.
a1 = 850.4,b1 = 2.29,c1 = 0.62,a2 = 725.7,b2 = 2.48,c2 = 3.68,a3 = 2.48,b3 = 11.42,c3 = 3.37.
The negative pulse fitting function is shown in Fig. 13.
a1 = 122.9,b1 = 1.67,c1 = 0.82,a2 = 42.18,b2 = 6.8,c2 = 1.62,a3 = 15.6,b3 = 8.09,c3 = 4.05.
In the time interval between positive and negative pulses, the fitting function is shown in Eq. 2–4.
y∈[118,122],x∈[kT + 1.7,kT + 1.7 + T'] (2–4)
Equation 2–4 indicates that in the interval time, the angles from 118° to 122° meet the regulation requirements. These functions are divided into three segments, and combined into a reference angle system to regulate the hindlimb gait of rats.
The positive pulse reference frame coefficient at 102 ~ 129°:
a1 = 779.3,b1 = 2.16,c1 = 0.3,a2 = 655.4,b2 = 2.35,c2 = 3.32,a3 = 2.25,b3 = 13.4,c3 = 2.08.
The negtive pulse reference frame coefficient at 102 ~ 129°:
a1 = 263.9,b1 = 2.24,c1 = 0.74,a2 = 161.8,b2 = 3.23,c2 = 3.5,a3 = 0.66,b3 = 18.84,c3 = 1.23.
The positive pulse reference frame coefficient at 96 ~ 132°:
a1 = 367.8,b1 = 2.61,c1 = 0.38,a2 = 242,b2 = 3.28,c2 = 3.21,a3 = 5.29,b3 = 10.64,c3 = 3.08.
The negtive pulse reference frame coefficient at 96 ~ 132°:
a1 = 136.5,b1 = 2.01,c1 = 0.68,a2 = 46.35,b2 = 5.80,c2 = 2.13,a3 = 6.11,b3 = 8.18,c3 = 3.96.
In the closed-loop control of the system, to ensure that the sampled knee joint angle does not deviate from the reference angle, reference points are set for the actual angle information, which are the starting position and the minimum point in each cycle, as shown in Fig. 14. The initial calibration point is correctly set when the angle return value is stable within 118° to 122° for a period of time. The minimum calibration point within a time period is the minimum value resulting from the angle data collected within the time period of the signal. The sampling period of the sensor is set as 0.025s to ensure the correctness of the sampling point.
Data recording
Two gait angle sensors are calibrated in the same environment, and the data is sent to the processor through serial port. The communication protocol is shown in Table 5. The system sets the angle output rate to 40Hz and the serial port baud rate to 9600bps.
Table 5
JY901 angle Transport Protocol table
Start bit
|
The X-axis Angle
|
The Y-axis Angle
|
The Z-axis Angle
|
Temperature
|
Checksum
|
0x55 0x53
|
DegxL + DegxH
|
DegyL + DegyH
|
DegzL + DegzH
|
TL + TH
|
SUM
|
According to Table 5, the angle of y-axis is obtained as follows:
Degy=((DegyH<<8)| DegyL)/32768*180°
The processor receives the real-time data of the y-axis direction of the two sensors and analyzes the real-time data of the knee joint angle of the rat.
Basso, Beattie and Bresnahan (BBB) motor abilities were scored on days 3, 5, 7, 10, 14, 21 and 28 after surgery in SCI and SCI + ES rats to evaluate hindlimb motor functions. In the BBB tests using two non-experimenters, the rats were allowed to move freely and the motor function of both hindlimbs was recorded with the highest score on the BBB score, being 21. The higher the score, the better the motor function. Experimental animal groups were also used for independent evaluation (double-blind method), and the average values were recorded [27, 28].
Somatosensory evoked potentials(SEP) was recorded on days 7, 14, 21 and 28. When SEP was recorded, the hindlimb ankle was stimulated by the stimulation signal generated by Keypoint Table Electromyography. The hindlimb somatosensory triggered signal goes to the brain and is monitored by the electrode implanted in the head [29]. The stimulation electrode was triggered by an electric pulse of 4mA, pulse width of 0.1ms, frequency of 3Hz, and repeated 200 times. The stimulation intensity was set to make the toes of the hind limbs twitch slightly. The latency of the applied potential and wave amplitude is used (peak downward is P, upward is N) to judge the recovery of hindlimbs. The latency period reflects the conduction distance, velocity, and synaptic delay time of the action potential with a shorter latency period, indicating better recovery. The amplitude (\(uV\)) reflects the number of synchronous firing neurons with higher amplitude, indicating better recovery.
The BBB score and SEP data in SCI + ES rats were analyzed, and compared with SCI rats to judge the recovery of lower limb gait.
The methods for evaluating the recovery of motor function combines behavior, kinesiology and physiology. Initial assessment for movement characteristics was used BBB scales [30]. EMG, kinematics [31], grip strength [32] was conducted to precisely assess hindlimb function recovery ability. SEP [33–35] was used to detect the electrophysiological function of rats as a means to evaluate recovery of spinal pathways.