Closed-loop electrical block of vagus nerve scales from rodent to porcine cardiac models

Figures

• 

  Figure 1.. Experimental setup. For both rat and pig experiments, a bipolar stimulating electrode was placed on the proximal right side vagus nerve to deliver the perturbing stimulation, increasing activity on the vagus. The DC current-controlled source was connected to a carbon separated interface nerve electrode (CSINE), which is then ionically connected via a saline bridge to the vagus, distal to the perturbation electrode; this setup allows for the safe delivery of the DC blocking current. For the rats, the right-side nerve was crushed, and the left-side nerve was cut, to remove any basal efferents and any afferent effects caused by the stimulation, which isolated the system to out inputs and increased system stability. In the pig experiments, the vagus nerve on both sides remained intact to more closely model how this therapy would be used in a clinical setting. The fuzzy logic controller (FLC) was fed a measurement signal (electrocardiogram (ECG) or blood pressure (BP)), which was used to calculate the heart rate. The FLC determined what the output would be and directed the current source to output the appropriate blocking current.

• 

  Figure 2.. Illustration of the fuzzy logic controller. Subplot (A) shows the input membership functions for the error in heart rate (i.e. actual value—setpoint, bpm). Subplot (B) shows the input membership functions for the slope of the HR error (bpm/s). Subplot (C) shows the output membership functions for the controller (change in mA, where a positive value increases the amount of cathodic blocking current). Subplot (D) shows the list of controller rules, relating the input and output functions. The controller uses a ‘center of sums’ defuzzification method. The main theme of this controller is to increase the block if the HR is above the setpoint, decrease the block if the HR is below the setpoint, and to slow the changes to the block if the HR is trending in the correct direction. A unique feature of this controller is that the blocking current is lowered when the HR is near the setpoint, to counteract the property of DC block that causes it to increase in efficacy over time.

• 

  Figure 3.. Example controller trial in a rat. The subplots (A) show the heart rate and block for a typical controller trial. The baseline heat rate for any therapy is applied is defined as the ‘100% HR’ for analysis. Vagus stimulation is then turned on, and the HR after 2 min is the ‘0% HR.’ After this, the static or controlled block is turned on (while leaving the VNS on) for 5 min. After the block was turned off, the vagus stimulation was left on to watch the HR recover to the 0% HR. The stimulation was left on for at least 2 min after the cessation of block. In the analysis, the time to 50% was the time from the time the block or controller started to the point when the HR crossed the ‘50% HR’ (halfway between the 0% and 100% HRs). The ‘time at 50%’ was defined as when the heart rate was within the light green box (from 40% to 60% HR). The ‘recovery time’ was the time from the cessation of block to the point when HR has returned to the 0% HR level. After determining recovery time, the VNS was stopped, and the HR recovered to the baseline HR. Some oscillation is seen in both the heart rate and the blocking current, indicating that the gain may have been too high in this system. The…

• 

  Figure 4.. Block thresholds for each set for both rat and pig. Subplot (A) shows example trials during block thresholding. Vagus stimulation was applied to lower the HR to the 0% level, and the block is applied to raise the HR back to the baseline (100%) level in 60 s. The red plot shows the HR rising too quickly, the blue line does not reach 100%, and the green line reaches 100% in exactly 60 s, and is deemed the block threshold. Subplot (B) shows the block thresholds for both the rat and pig models. The black line indicates the mean block threshold. The rat block thresholds were slightly lower than the that of the pig (means 1.18 and 2.50 mA, respectively) but had a larger spread (standard deviations 1.35 and 1.00, respectively). These models were deemed to behave similarly enough for the controller to successfully port from the rat model to the pig, despite the differences in nerve size and electrode tips.

• 

  Figure 5.. Combined results from the rat trials. Plots (A)–(C) show the normalized HR for the 10% BT60, 50% BT60, and controller trials. Each light trace shows an individual trial, while the darker trace shows the mean trace across all time points. The 10% BT60 trials show the least amount of block, with most being under 25%; however, a few trials had significant induction effects and were able to completely block back to the 100% baseline. The 50% BT60 trials show a wide array of responses, with significant amounts of trials achieving both little-to-no block and achieving complete block. The spread of these trials indicates very low predictability. The FLC controller trials show a leveling off of the HR at 50% for most trials, with a substantially smaller standard deviation. Several trials showed spontaneous changes in the heart rate throughout the trial; however, in most cases, the controller was able to adapt and bring the heart rate back to the setpoint during the duration of the block. Plots (D)–(F) show the combined normalized %BT for the 10% BT60, 50% BT60, the controller trials, respectively. The static trials seen in (D) and € are simply flat lines, as the same normalized amplitude…

• 

  Figure 6.. Output measurements for the rat. Black brackets indicate a statistically significant difference in the Wilcoxon test for that pair. Subplot (A): time at 50% heart rate for the rat trials. The plot above shows the time at 50% BT60 for each of the trial types, with the controlled trials spending significantly more time at the 50% HR (median 398.3 s) than the uncontrolled trials (10% median 0 s; 50% median 19.2 s). Subplot (B): injected charge from the block electrode for the rat trials. This plot shows distributions for the injected charge of the block electrode for each the trial types, with the 50% BT BT60 having the most on average, but with a larger range in the controller trials. The 10% static trials had a median of 62.2 mC, the 50% had a median of 301.0 mC, and the controller trials had a median of 115.0 mC. Subplot (C): time to recovery after end of block for the rat trials. This plot shows the time to recovery for each of the trial types, with the 50% BT60 having a considerably longer recovery time (median 36.8 s) compared to the 10% static (median 0.0 s) and the controlled trials (6.0 s).

• 

  Figure 7.. Results from the pig trials. Plots (A)–(D) show the mean normalized HR for the 10% BT60, 50% BT60, the 2x gain controller, and the 10x gain controller trials, respectively, with the lighter lines representing individual trials, and the darker line representing the mean value at each time point. The 10% BT60 shows the least amount of change with large spread. The 50% BT60 plot show that the average trial goes up to the 100% HR in about 5 min, and begins to level off from there, albeit with a large amount of trial-to-trial variance. The 2x gain controller trial shows an increase to about 75% HR, and then a correction back to the 50% setpoint, with substantially smaller trial-to-trial variance. The 10x gain controller shows a slightly better-controlled curve, reaching and settling at the 50% setpoint quickly. Plots (E)–(H) show the combined normalized %BT for the 10% BT60, 50% BT60, the 2x gain controller, and the 10x gain controller trials, respectively. The static trials seen in (E) and (F) are simply flat lines, as the same normalized amplitude was used for every trial. The 2x gain controller (G) increases the block to a peak neat 60% and then reduces to about 30% by the end. The…

• 

  Figure 8.. Output measurements for the pig. Black brackets indicate a statistically significant difference in the Wilcoxon test for that pair. Subplot (A): time at 50% heart rate for the pig trials. The plot above shows the time at 50% HR for each of the trial types, with both controllers performing better on average than either of the static trials. The 10x gain controller (median 508.0 s) performed slightly better than the 2x gain trials (median 318.8 s), likely due to their ability to rise faster. While the average for the 50% BT60 static block trials is quite low (median 6.0 s), there were a few trials that did appear to stabilize around 50% HR for an extended period. None of the 10% BT trials spent any significant period near our target zone (median 0.0 s). Subplot (B): injected charge from the block electrode for the pig trials. This plot shows distributions for the injected charge of the block electrode for each the trial types, with the 50% BT60 static block (median 648.0 mC) having the most on average, followed by the controllers (10x median 467.0 mC; 2x median 226.1 mC), with the 10% BT60 static block coming in with the lowest charge (median 129.9 mC). Subplot (C: time to recovery…