Co-activation of the diaphragm and external intercostal muscles through an adaptive closed-loop respiratory pacing controller

EDITED BY

Chiara Veneroni,

Polytechnic University of Milan, Italy

REVIEWED BY

Gabi Mueller,

Swiss Paraplegic Research, Switzerland

Kristi Streeter,

Marquette University, United States

CORRESPONDENCE

Ranu Jung

[email protected]

RECEIVED 03 April 2023

ACCEPTED 20 June 2023

PUBLISHED 07 July 2023

CITATION

Adury RZ, Siu R and Jung R (2023)

Co-activation of the diaphragm and external

intercostal muscles through an adaptive

closed-loop respiratory pacing controller.

Front. Rehabil. Sci. 4:1199722.

doi: 10.3389/fresc.2023.1199722

open-access article distributed under the term s

of the Creative Commons Attribution License

(CC BY). The use, distribution or reproduction in

other forums is permitted, provided the original

author(s) and the copyright owner(s) are

credited and that the original publication in this

journal is cited, in accordance with accepted

academic practice. No use, distribution or

reproduction is permitted which does not

comply with these terms.

Co-activation of the diaphragm

and external intercostal muscles

through an adaptive closed-loop

respiratory pacing controller

Rabeya Zinnat Adury

1,2

, Ricardo Siu

2,3

and Ranu Jung

2,4

Department of Applied Physiology and Kinesiology, University of Florida, Gainesville, FL, United States,

Department of Biomedical Engineering, Florida International University, Miami, FL, United States,

Department of Physical Medicine and Rehabilitation, Case Western Reserve University, Cleveland, OH,

United States,

Department of Biomedical Engineering, The Institute for Integrative and Innovative

Research (I

R), University of Arkansas, Fayetteville, AR, United States

Introduction: Respiratory pacing is a promising alternative to traditional

mechanical ventilation that has been shown to signiﬁcantly increase the survival

and quality of life after the neural control of the respiratory system has been

compromised. However, current pacing approaches to achieve adequate

ventilation tend to target only the diaphragm without pacing external intercostal

muscles that are also activated during normal inspiration. Furthermore, the

pacing paradigms do not allow for intermittent sighing, which carries an

important physiological role. We hypothesized that simultaneous activation of

the diaphragm and external intercostal muscles would improve the efﬁciency of

respiratory pacing compared to diaphragm stimulation alone.

Materials and Methods: We expanded an adaptive, closed-loop diaphragm pacing

paradigm we had previously developed to include external intercostal muscle

activation and sigh generation. We then investigated, using a rodent model for

respiratory pacing, if simultaneous activation would delay the fatigability of the

diaphragm during pacing and allow induction of appropriate sigh-like behavior

in spontaneously breathing un-injured anesthetized rats (n = 8) with pacing

electrodes implanted bilaterally in the diaphragm and external intercostal

muscles, between 2nd and 3rd intercostal spaces.

Results: With this novel pacing system, we show that fatigability of the diaphragm

was lower when using combined muscle stimulation than diaphragm stimulation

alone ( p = 0.014) and that combined muscle stimulation was able to induce

sighs with signiﬁcantly higher tidal volumes compared to diaphragm stimulation

alone ( p = 0.014).

Conclusion: Our ﬁndings demonstrate that simultaneous activation of the

inspiratory muscles could be used as a suitable strategy to delay stimulation-

induced diaphragmatic fatigue and to induce a sigh-like behavior that could

improve respiratory health.

KEYWORDS

closed-loop system, ventilatory control system, stimulation, respiratory pacing, sighs,

augmented breaths, stimulation-induced fatigue

Introduction

Following acute cervical cord injury, brainstem disease, or stroke, autonomic control of

respiration may be compromised (1). Though mechanical ventilation has a prominent role as

a lifesaving tool in the intensive care unit and for acute respiratory assistance, as little as 18 h

TYPE Original Research

PUBLISHED 07 July 2023

DOI 10.3389/fresc.2023.1199722

Frontiers in Rehabilitation Sciences 01 frontiersin.org

of diaphragm inactivity and mechanical ventilation can result in its

atrophy (2). Electrical stimulation of the phrenic nerve or the

diaphragm has been considered as an alternative to mechanical

ventilation both in the acute setting and for long-term use (3–7).

Contrary to the natural muscle ﬁber recruitment order during

muscle contraction, direct electrical stimulation of the nerves or

intramuscular simulation of the neuromuscular junction results

in recruitment of muscle ﬁbers in the reverse order, recruiting

larger highly excitable fast fatigable ﬁbers ﬁrst, followed by

smaller slower fatigue resistant ﬁbers (8). The rate of activation

of ﬁber units during functional electrical stimulation is also

much higher (25–75 Hz) compared to voluntary ﬁber unit

recruitment (5–10 Hz) (9). Thus, because of both reverse

recruitment and the increase in ﬁber unit recruitment rate,

electrical stimulation can result in muscle fatigue at a faster rate

than under natural conditions.

In ord er to mai ntain sufﬁcient muscle contraction to

maintain a desired f unctional outcome, e.g., breath volume f or

adequate ventilation, with the onset of fatigue, the electrical

stimulation intensity would have to be adjusted. A closed-loop

adaptive control system could compensate for the fatigue by

increasing stimulation without requiring constant manual

tun in g . However comm e rciall y availa b l e paci n g system s do not

offer this a bility. Also, during di aphragm muscle pacing alone,

diaphragmatic contraction causes inward movement of the rib

cage due to inactivity o f the intercostal muscles, reducing the

intrathoracic volume. This reduction in volume diminishes the

efﬁcienc y of diaphragm stimulation, leading to a lower tidal

breath volume (10). Canine studies have indicated that

combining upper thoracic stimulation in conjunc tion wit h

diaphragmatic stimulation can ameliorate this and elicit a tidal

volume greater than the sum of the volume elicited by

stimulating each respiratory muscle alone (10, 11).

Another feature that is unaccounted for is the ability to

induce periodic sighs. Sighing has profound physiological and

psychological beneﬁts (12, 13). An adult human being sighs

approximately once every 5 min (12). Sighs can reset regular

breathing, improve gas exchange, and prevent the progressive

collapse of alveoli by expanding them during a long, deep

breath (14). In acute respiratory failure patients, mechanical

ventilation incorporated with periodic sighs can decrease

lung strain, ventilation heterogeneity, and increased gas

exchange (13, 15), but the shortcomings associated with

mechanical ventilation, such as the risk of muscle atrophy,

impairment in mobility, and the necessity for tracheal

intubation remain.

We recently developed an adaptive closed-loop intramuscular

diaphragm (Dia) pacing system that can adapt the stimulation

parameter to automatically personalize stimulation patterns and

adapt to achieve adequate ventilation for meeting metabolic

needs (16, 17). To delay the onset of diaphragm muscle fatigue

during long-term pacing and include sigh-like augmented breaths

periodically, we have expanded the capabilities of the system by

the addition of external intercostal (EIC) muscle stimulation that

supports thoracic stabilization and expansion of the upper rib

cage to prevent inward thoracic movement. Adding adaptive

external intercostal stimulation could lower the stimulation

charge required for diaphragm contraction to attain the desired

tidal breath volume, thus delaying stimulation-induced fatigue of

the diaphragm. Additionally, sighs could be induced with lower

charge delivery to the diaphragm.

Thus, we have introduced a novel approach to personalized

pacing that can counteract stimulation-induced diaphragmatic

fatigue during respiratory pacing, induce sigh-like behavior, and

that has the potential to adapt to changes in metabolic needs.

Such a control system offers the ability to signiﬁcantly enhance

the use of respiratory pacing systems for weaning patients from

mechanical ventilation and for long-term chronic use.

Materials and methods

Study design

Experiments were conducted on n = 8 anesthetized,

spontaneously breathing, adult male Sprague Dawley rats

weighing 400 ± 80 g. Rats were maintained under anesthesia with

subcutaneous urethane (50 mg/kg) injection followed by

supplementary isoﬂurane (0.5%–3%) inhalation. Toe pinch reﬂex

and respiratory rate were used as an indicator of the proper level

of anesthesia. The body temperature was maintained around 37°

C with the help of a closed-loop thermal pad (TCAT-2DF

controller, Physitemp instruments Inc., NJ, USA) and a rectal

thermometer. 30G stainless steel electrodes were inserted

subcutaneously in chest muscles to record the electrocardiogram

and monitor heart rate. A chest-mounted respiratory belt was

utilized to monitor the breathing pattern and breathing rate of

the animal. To avoid dehydration, a lactated ringer solution was

injected subcutaneously every 2–3h.

Ethics statement

Rats were housed individually in the university animal care

facility with a 12-hour light/dark (reverse with natural cycle)

cycle with food and water ad libitum. All the procedures were

approved and are in accordance with the guidelines established

by the Institutional Animal Care and Use Committee at Florida

International University.

Tracheostomy and intramuscular electrode

implantation in the diaphragm, external

intercostal muscle

After tracheostomy and placement of a custom tracheal tube, a

small pneumotachometer (PTM type HSE-73-0980) was directly

connected to the tracheal tube to measure airﬂow. This ﬂow was

integrated (PI-1000, CWE Inc. Ardmore, PA, time constant =

0.2 s) to obtain the breath volume on a breath-by-breath basis. A

capnograph (CapStar-100, CWE Inc., Ardmore, PA) measured

end-tidal CO

(etCO

) throughout the experiment to ensure

Adury et al. 10.3389/fresc.2023.1199722

Frontiers in Rehabilitation Sciences 02 frontiersin.org

normocapnic etCO

(35–45 mmHg) was maintained. To implant

pacing electrodes in the diaphragm muscle, a 2–2.5 cm incision

along the linea alba was made caudal to the xiphoid process. A

hand-operated stimulator (DigiStim 3 plus, Neuro Technologies),

providing a single monophasic pulse at 1 Hz, was used to map

the motor point location in each hemidiaphragm. Custom-made

stainless-steel wire electrodes [Diameter 0.002″ (bare), 0.0045″

(coated), with a 2–3 mm active region] were implanted bilaterally

approximately 1 mm from the point of stimulation that elicited

the strongest diaphragm contraction. A study in the adult canine

model showed that the most substantial increase in inspiratory

volume could be induced by pacing 2nd intercostal spaces

compared to stimulation of other intercostal spaces (10). Hence,

after an incision was made along the midline of the chest,

muscle layers (trapezius and latissimus dorsi) were removed by

blunt dissection to identify the 2nd intercostal space and the

external intercostal muscles exposed. The hand-operated

stimulator was used to map the muscles and bilateral electrodes

were implanted intramuscularly.

Diaphragm, external intercostal muscle

pacing

The twitch threshold current (the minimum current required

to elicit visible muscle contraction) was determined for each

electrode. Stimulation pulses were sent using a programmable

constant-current stimulator (FNS-16, CWE Inc., Ardmore, PA)

delivering cathodic ﬁrst, biphasic current pulses of 80 μs/phase at

75 Hz (18), varying the current amplitude. The maximum

current amplitude was set as 1.5–2 times the twitch threshold of

the muscle, resulting in a stimulation current limit of 2–4 mA.

The stimulation amplitude was determined by the adaptive

controller whereas the twitch threshold facilitated us to set the

maximum allowed amplitude for each muscle during the

stimulation. Adaptive pacing used the same stimulation

parameters as described above. The previously designed

controller’s capability [developed by Siu et al. (16)] was

expanded by adding two independent stimulation channels while

keeping the original conﬁguration consisting of a pattern

generator for setting the respiratory rate and a pattern shaper to

set the breath volume pattern same (Figure 1). The single

Pattern Shaper thus allowed simultaneous adaptive control of the

stimulation of the diaphragm and the external intercostal muscles.

Experimental protocol

The adaptive controller modulates the stimulation parameters

to match the desired breath volume on a breath-by-breath basis

with a ﬁxed breath duration (cycle period). Each trial consisted

of at least 1 min of spontaneous breathing, followed by at least

5 min and at best 15 min of adaptive stimulation. For each trial,

the desired breath volume pattern and cycle period during pacing

were obtained by averaging the pre-stimulation recorded breath

cycles except for non-breathing behavior cycles, i.e., sighs. As

intact rats have an intrinsic breathing pattern, successful

entrainment of the stimulation-assisted breath with the intrinsic

breath requires the stimulation-assisted breath to have a larger

tidal volume. Hence, the desired breath volume needed was set at

FIGURE 1

The adaptive neuromorphic controller architecture. Electrodes were implanted bilaterally in the diaphragm and external intercostal muscles and

controlled using an adaptive controller following a pattern generator (PG)/pattern shaper (PS) scheme. The PG is a ﬁxed oscillator that generates a

ﬁxed breath duration (cycle period) for each breath. The PS is an artiﬁcial neural network that adapts stimulation output (current amplitude) to all

muscles based on the instantaneous error between the desired and measured breath volume. The maximum allowed current amplitude for

stimulating each muscle is independent and accounts for differences in individual muscle activation properties and electrode impedances.

Adury et al. 10.3389/fresc.2023.1199722

Frontiers in Rehabilitation Sciences 03 frontiersin.org

an ad-hoc 120% of the baseline volume (16). Experimental trials

were spaced 20–30 min apart in order to allow the inspiratory

muscles to recover from stimulation-induced fatigue. We

recorded one trial for each type of stimulation and the order of

the trials was randomized across animals.

We performed a separate set of trials for eliciting a targeted

breath volume pattern with sighs ( Figure 2). The evoked sigh

was set t o occur after every 30 breaths cycles. For inducing

sighs, the controller stored the PS outputs (time-shifted

weighted summation of neuronal output ) of each update for

the previous breath cycle and sent out twice the magnitude of

PS output to the c ontroller. The adaptive learning in the

controller was paused automatically every 30 cycles, during the

insertion of the sigh and the following cycle. The doubling of

the amplitude of stimulation on each time step (every 40 ms)

caused the inspiratory muscles to contract by an additional

degree.

Strength-duration curve determination

In order to assess the optimal pulse width for stimulation,

strength duration (SD) curves were generated by plotting

the twitch threshold current required (strength levels) versus the

pulse widths (duration of the stimulating pulse). The stimulation

frequency was kept the same for both diaphragm and external

intercostal muscle pacing. Decreasing pulse widths from 500 to

100 µs in steps of 100 µs and 100 µs to 10 µs in steps of 10 µs

were used (8). We allowed a 60s rest period between consecutive

pulse stimuli to prevent carryover effects of the stimulation on

muscle recruitment. In general, from the SD curves, we

determined the rheobase (twitch threshold current at inﬁnite

pulse duration; here 500 µs) and a range of chronaxie (stimulus

duration at the point where the twitch threshold current is twice

the rheobase) in order to assess the excitability of the muscle.

The PG/PS controller

The controller was originally developed by Abbas and

Chizeck (19), used in an incomplete spinal cord injury (iSCI)

rodent model for cyclic limb movement (18),andinaniSCI

rodent model for functional stimulation of the diaphragm

muscle to attain a desired breath volume pattern (16, 17). The

controller is a neural network system that can provide

automated customization of the stimulation parameters to

drive the desired action. It has two components: a pattern

generator (PG) and a pattern shaper (PS). The PG has a

pattern-generating capability based on neurophysiological

models and se ts the timing of the breath. In this study, a

frequency oscillator was used to produce a ﬁxed respiratory

frequency, with a ﬁxed period (16). The PS is a single-layer

adaptive artiﬁcial neural network that modulates the current

amplitude to elicit a desired breath volume trajectory. This

adaptation is driven by the instantaneous error across the

breath cycle between the measured breath volume and

the desired breath volume. For a more detailed description of

the PG/PS controller please refer to Siu et al. (16). For ou r

experiments, one PS unit was used to stimulate both sets o f

muscles (Diaphragm and external intercostal ) with the output

to each muscle being modiﬁe d by an in dependent gain factor

determined via twitch threshold as described above.

FIGURE 2

Experimental sequence diagram. The arrows indicate consecutive steps of experiment. One trial for each stimulation type was recorded. Separate color

schemes have been used for different stimulation approaches.

Adury et al. 10.3389/fresc.2023.1199722

Frontiers in Rehabilitation Sciences 04 frontiersin.org

Performance measures

To evaluate performance of the controller, a percent inspiratory

root-mean-squared error was calculated based on the desired

breath volume and the measured breath volume at a given

instant (16). Charge delivered per cycle was calculated as the

summation of the product of the width and amplitude of each

pulse sent during one cycle. We conducted a Wilcoxon signed

rank test comparing the average charge delivery per cycle to the

diaphragm muscles among the animals for both Dia and Dia +

EIC stimulation. As electrical stimulation or the charge delivered

to the muscle is responsible for fatigability, a metric for

measuring fatigability, fatigue index (FI) was measured that

represents relative changes in charge delivered to the muscle.

Fatigue index for each trial was calculated as the difference in the

average initial charge delivered per cycle and ﬁnal charge

delivered per cycle, normalized by the initial charge per cycle

(18). As muscles get fatigued during the time course of trials,

higher level of charge is delivered during stimulation to maintain

the same functional level. This relative increase in charge delivery

or fatigue index was used as a measure of muscle fatigability

during the stimulation in a given trial. For calculating the fatigue

index, we chose the initial and ﬁnal cycle sets, each consisting of

50 cycles and approximately 400 cycles apart. These sets were

considered in such a way that there is a minimal number of

cycles without entrainment. We considered the initial set of

cycles when the measured breath volume and desired breath

volume started to achieve entrainment with a minimal phase

difference. The initial charge per cycle set and the ﬁnal charge

per cycle set was approximately 400 cycles apart. For animal 1

and animal 2, the number of available cycles for analysis was not

sufﬁcient, hence the fatigue index was not calculated for these

animals. Thus, the fatigue index analysis includes n = 6 animals.

Since the desired tidal volume varied among all animals and

sometimes for different trials for the same rat, we normalized

the sigh tidal volume with the targeted tidal volume for each

trial. We denoted t his normalized value as the “sigh tidal

volume fa ctor.” For animal 7, only trials for inducing regular

breat h patterns could be recorded, but trials for inducing sighs

could not be recorded and were not included in the data

analysis. Thus, the sigh tidal volume data ana lysis included n =

7animals.

Data acquisition and data analysis

Each recording trial was 5–15 min long, consisting of at least

60 s of intrinsic breathing, followed by stimulation-augmented

breathing. At least a 20–30-min rest period was allowed between

trials. Data recording included EKG, breath volume, charge

delivered, and PS output (Normalized simulation output). All the

data from the PG/PS controller were collected at 25 Hz, and

other recorded data were collected at 10 kHz, which later was

down-sampled to 25 Hz. A one-way Wilcoxon signed rank

analysis, that assumes non-normal distribution of the dataset and

accounts for small sample sizes, was conducted in SPSS on

fatigue index data with α = 0.05 signiﬁcance level to test if the

charge delivered per cycle to the diaphragm muscle was

signiﬁcantly lower with Dia + EIC muscle stimulation. Besides,

we performed one way Wilcoxon signed rank test on the average

charge delivered per cycle across the animals for Dia and Dia +

EIC stimulation trials. The tidal volume factor of sighs was

averaged across all the sighs for each observation for Dia + EIC

muscle stimulation and Dia muscle stimulation separately. The

one-way Wilcoxon signed-rank test was performed on the

averaged tidal volume factor for Dia and Dia + EIC to test if tidal

volume generated for the sighs was signiﬁcantly larger in Dia +

EIC muscle stimulation than Dia muscle stimulation.

Results

Overall, the following sections present data from n = 8 animals.

In 2 animals the trials were short (5 min long) therefore for the

fatigue index analysis, data from n = 6 animals were included in

the analysis. We could not run the trials for inducing sigh in one

animal. Hence, data from n = 7 animals were included for the

sigh tidal volume analysis. Our results indicate that adding

adaptive external intercostal stimulation could lower stimulation

required by the diaphragm to attain the desired tidal breath

volume, thus delaying stimulation-induced fatigue of the

diaphragm. Besides the addition of external intercostal muscle

pacing could facilitate induction of periodic augmented breaths.

The following sub-sections report details.

Determination of pulse width of stimulation

by strength-duration curve parameters

Strength duration curves were generated by plotting the

minimum intensity (amplitude) of an electrical stimulus required

to produce a twitch with different pulse width durations up to

500 µs. Figure 3 shows a nonlinear hyperbolic relationship

between the minimum stimulus strength at each pulse width of

stimulation. The optimum stimulus pulse width was deﬁned by

the chronaxie value determined from the SD curves. For all four

muscles of both the animals, rheobase value ranged from 0.35 to

1.3 mA, and the chronaxie values ranged from 35 to 100 µs. The

average chronaxie value across all the muscles for both animals

were calculated to be 80.75 ± 27.52 µs. Based on this calculated

chronaxie guideline, we used 80 µs as our stimulation pulse width.

Adaptive closed-loop control can achieve

desired tidal volume pattern

To assess the effect of synergistic muscle activation by

respiratory pacing, trials in which only the diaphragm was

stimulated and trials in which both the diaphragm and 2nd

intercostal muscles were stimulated were carried out. Figure 4

shows the outcome during the stimulation of respiratory muscles

with the adaptive closed-loop controller. The stimulation output

Adury et al. 10.3389/fresc.2023.1199722

Frontiers in Rehabilitation Sciences 05 frontiersin.org

initially starts at zero when the controller is turned on for the

diaphragm-only stimulation (Figure 4A), and combined muscle

stimulation (Figure 4C). The controller adapts stimulation

output to match the stimulation assisted breath volume pattern

with the desired breath volume pattern by increasing the charge

delivered. After approx. 15 min of pacing, the assisted breath

volume pattern continues to match the desired breath volume

pattern for both the diaphragm-only stimulation (Figure 4B) and

combined muscle stimulation (Figure 4D). Figure 4E shows an

example of the controller’s pattern-shaping capability for the ﬁrst

100 s of a 15-minute-long trial for combined muscle stimulation.

The vertical light blue dashed line at the beginning of the trial

indicates turning on of the controller. The inspiratory root mean

square error (iRMSE) value increases to account for the

difference in volume between the stimulation-assisted breath and

the desired breath. Based on the error value in each instance, the

stimulation amplitude adapts to match the desired breathing

pattern. Once the stimulation-assisted breath volume pattern

starts to match the desired breathing pattern, iRMSE, as well as

the stimulation output, starts decreasing. For our investigations,

when the iRMSE value decreased to 10% or less and remained

there for at least 20 pacing cycles, the controller was considered

to have “adapted,” which is indicated in Figure 4E by the second

vertical dark blue dashed line.

Combined muscle stimulation reduces

diaphragmatic fatigue

Figure 5A shows the average charge per cycle delivered (mC)

to the diaphragm muscle during Dia stimulation (in blue) or

Dia + EIC stimulation (in red) across all animals for the ﬁrst 550

cycles. These traces show that charge delivered per cycle to

diaphragm muscle increases in the later cycles of the trial (cycles

300–550), suggesting that to attain the desired volume pattern

more charge is required to maintain a similar breath volume.

One way Wilcoxon signed rank test showed that the average

charge per cycle delivered to the diaphragm muscle during Dia +

EIC stimulation was signiﬁcantly lower compared to Dia

stimulation (p < 0.001). The decrease in charge delivered per

cycle to attain the desired volume pattern indicates that the

diaphragm muscle’s susceptibility to fatigue decreases when

paired with intercostal stimulation, thus the results suggest that

Dia + EIC stimulation leads to a slower onset of fatigue.

Figure 5B shows %iRMSE averaged over the six animals for the

ﬁrst 550 cycles of Dia stimulated and Dia + EIC muscle

stimulated trials (blue and red in color, respectively). It shows

that initially, %iRMSE was high, but once the controller

facilitates matching the desired breathing pattern, %iRMSE value

decreased. Although, because of occasional loss of entrainment,

the %iRMSE cycled between a high and low value from one

breath to the next. In order to assess whether combined muscle

stimulation facilitates faster entrainment and better alignment

between the stimulation-induced breath volume and the desired

breath outcome compared to single muscle stimulation, we

quantiﬁed the number of breath cycles required to achieve

entrainment. For the analysis, we set ad-hoc criteria for the

iRMSE to be less than 20% for at least 20 breaths. There was one

animal who didn’t achieve entrainment for Dia stimulation and

another animal who didn’t achieve entrainment for Dia + EIC

stimulation, these two animals were excluded from the analysis.

Thus, we conducted statistical analysis to determine the number

of cycles needed to achieve entrainment for each condition for

the remaining 6 animals. A Wilcoxon signed rank test showed

that in combined muscle stimulation, number of cycles required

to achieve entrainment were signiﬁcantly lower than for

diaphragm stimulation [51 ± 20 vs. 114 ± 47 cycles needed

(average ± standard error) to achieve entrainment, p = 0.04].

Figure 5C shows the fatigue index (FI) plotting for the

diaphragm muscle during diaphragm muscle stimulation and

combined muscle stimulation for six animals. The diagonal blue

line through the origin (0, 0) represents the line where the

FIGURE 3

Strength-duration curves for bilateral stimulation of the diaphragm and external intercostal muscles. The panels show strength duration curves for two

different animals. Rheobase refers to the lower limit of stimulus intensity needed to excite the muscle ﬁber at a very long stimulus duration (500 µs here).

Chronaxie refers to pulse duration at which the threshold is twice the rheobase value. Strength duration curves for stimulation of the left or right

diaphragm, or the left or right external intercostal muscles are shown for two animals. Also illustrated are horizontal dashed lines that identify the

rheobase value and vertical dashed lines that identify the chronaxie for the Right EIC Strength duration curves; blue = Dia left, orange = Dia right,

yellow = EIC left, violet = EIC right.

Adury et al. 10.3389/fresc.2023.1199722

Frontiers in Rehabilitation Sciences 06 frontiersin.org

FIGURE 4

The ability of the adaptive PS controller to assist attaining the desired breath volume pattern in intact animal during diaphragm muscle stimulation or

combined muscle stimulation. (A,C) Response to adaptive closed-loop control of diaphragm muscle stimulation or combined muscle stimulation

respectively in a rat. Stimulation output progressively increases to counteract the mismatch between the measured stimulation assisted breath

volume trajectories (solid line) and the desired breath volume trajectory (dotted line). The vertical light blue dashed line shows when the controller

was turned on. (B,D) Continuation of A and C respectively showing the last 10 s of the trial for diaphragm muscle stimulation and combined muscle

stimulation. In both cases, the stimulation assisted breath volume pattern overlaps the desired breath volume pattern. (E) The adaptive closed-loop

controller implementation for the combined muscle stimulation (same animal, same trial as C,D) for ﬁrst 100 s. In the beginning, iRMSE was high; to

match the stimulation elicited instantaneous volume with the desired instantaneous volume, stimulation output adapted and initially increased cycle

by cycle. The second vertical dark blue dashed line shows where the stimulation assisted breath volume started to match the desired volume. iRMSE

value decreased to <10%, and the stimulation output remained stable for the duration of the trial.

Adury et al. 10.3389/fresc.2023.1199722

Frontiers in Rehabilitation Sciences 07 frontiersin.org

fatigue index is the same for the diaphragm-only stimulation (FI

dia

)

or combined stimulation of the diaphragm and external intercostal

(FI

Dia + Eic

). A positive fatigue index indicates that greater charge

delivery was required during the latter part of the trials, while a

negative fatigue index indicates that a lower charge delivery was

required for the latter part. The region under the straight line

represents the area where the fatigue index was higher when only

the diaphragm muscle was stimulated. Figure 5C illustrates that

all the data was under the diagonal line, which suggests that the

diaphragm muscle was more fatigued during diaphragm-only

muscle stimulation than during combined muscle stimulation.

The statistical test showed that FI values of the diaphragm

muscle were signiﬁcantly lower during combined muscle

stimulation (p = 0.014).

Pacing can be used to elicit sigh-like

behavior

To assess the controller’s ability to induce sighs, we performed

Dia muscle stimulation and Dia + EIC muscle stimulation

separately as in the previously described trials, however,

stimulation at twice the value of the previous cycle was delivered

every 30 cycles to induce augmented breaths, or sighs.

Figures 6A–C shows intrinsic sighs, sighs induced with

diaphragm-only stimulation, and sighs induced via combined

stimulation, each compared to the previous non-augmented

breath. Fig ure 6D shows the sigh tidal volume factor

(measured sigh tidal volume/desired non-sigh t idal volume)

obtained during diaphragm-only muscle stimulation (in blue)

and combined muscle stimulation (in red) for seven rats. Even

though the trial duration was the same during Dia + EIC or

Dia stimulation, th e number of sighs co uld be different

because of different respiratory rates. For the mean ± SD

calculation, within one animal, the smaller number of sighs

between the two trials (Dia + EIC or Dia)wasutilized.One-

tailed paired sample t-test for within subject average tidal

volume factors shows that in 5 out of 7 cases, average tidal

volu me f ac t ors were sign i ﬁcantly large r in combined muscle

stimulationthanindiaphragm-onlystimulation.Oneway

Wilcoxon signed rank test for between subject average ti dal

volume factor indicated that the volume factor for combined

muscle pacing was signiﬁcantly higher than that f or

diaphragm pacing -only (p = 0.014).

FIGURE 5

Averaged charge delivered per cycle to the diaphragm muscle and averaged %iRMSE over the animals and fatigue index representation during combined

muscle pacing and diaphragm pacing. (A) Averaged charge delivered per cycle to the diaphragm muscle for diaphragm-only muscle stimulation (in blue)

and combined muscle stimulation (in red). (B) Averaged %iRMSE over the breath cycles of six animals for diaphragm muscle stimulation alone (in blue) or

combined muscle stimulation (in red). At the beginning of a trial, %iRMSE was high, but once the desired volume trajectory was obtaine d, the %iRMSE

value decreased. (C) Fatigue index for diaphragm-only muscle stimulation (FI

Dia

) and combined muscle stimulation (FI

Dia + Eic

). The diagonal line going

through origin presents the ratio of FI = 1; any point below the line (the light blue area) suggests less fatigue susceptibility and the points above the

line (the light red area) suggests greater fatigue susceptibility of the diaphragm muscle to stimulation.

Adury et al. 10.3389/fresc.2023.1199722

Frontiers in Rehabilitation Sciences 08 frontiersin.org

Elicitation of sighs can reset

synchronous entrainment

Figure 7A illustra tes the ability of sighs to reset synchronous

entrainment when the intrinsic breathing pattern and that elicited by

inspiratory stimulation were not synchronous. If there were at least 3

breaths where the measured and desired breath initiation was out of

phase, we considered that as loss of synchrony before a sigh. During

stimulation assisted breathing, sighs aided in the alignment and

synchroniza tion between the desired and measured tidal volume

patterns. In our experiments, in 12 sighs during different trials in all

animals, we observed 11 sighs (91%) reset the alignment.

Whilewewereabletoconsistentlyinducesighs,weobservedrare

instances of loss of synchro ny after the induced sigh ( Figure 7B)

during combined muscle stimulation. As combined muscle

stimulation induced a larger sigh volume compared to single muscle

stimulation, the expira tion (downw ards phase in a breath cycle)

takeslongertocompletethanthedesired expiration, resulting in a

phase mismatch between the volume pattern of the induced and

desired breaths. Synchrony took several breaths to reoccur, in some

cases lasting until the following induced sigh.

Discussion

This investigation presents the viability of a novel approach to

ventilatory pacing by simultaneous diaphragm and external

intercostal muscle stimulation using an adaptive closed-loop

FIGURE 6

Eliciting sigh-like behavior by diaphragm muscle stimulation and combined muscle stimulation. Representative breath volume traces showing a regular

breath followed by an intrinsic sigh during spontaneous breathing (A), a paced sigh using diaphragm-only stimulation (B), and a paced sigh using

combined muscle stimulation ( C). (D) Average tidal volume factors for sigh in diaphragm-only muscle stimulation (in blue) and combined muscle

stimulation (in red) for all the animals, where *indicated p < 0.05.

FIGURE 7

Elicitation of sigh can reset synchrony. (A) Shows an example of sigh functioning as a re-setter to restore synchrony between desired breathing and

induced breathing. (B) Represents an example whe re after a sigh elicitation, loss of synchrony occurred between the desired and induced breath pattern.

Adury et al. 10.3389/fresc.2023.1199722

Frontiers in Rehabilitation Sciences 09 frontiersin.org

controller to achieve a desired respiratory breath pattern including

sighs. We showed that it is possible to attain a targeted breath

volume pattern using this synergistic approach of combined

muscle stimulation, that the target could be reached earlier, that

this strategy leads to a reduction in diaphragmatic fatigue onset,

and that the strategy allows for inclusion of efﬁcient sigh-like

augmented breaths within the breathing pattern. Combined,

these factors could lead to improved respiratory health in

individuals that depend on diaphragmatic pacing.

Diaphragmatic fatigue can occur due to electrical stimulation

and hence a need to modulate the stimulation parameters to

achieve the desired effect on ventilation has previously been

identiﬁed (20). Besides changing stimulation parameters to

overcome diaphragmatic fatigue, one can also activate additional

inspiratory muscles to attain the desired ventilation, thereby

reducing the need for stronger diaphragmatic contractions.

Indeed, when we combined diaphragmatic and external

intercostal muscle stimulation and used the adaptive controller,

we found that the fatigue index of the diaphragm muscle

calculated during the long trials (>550 cycles) of stimulation was

signiﬁcantly lower during combined muscle stimulation while

also being able to achieve the desired volume pattern. The lower

charge delivery utilized for diaphragmatic contraction likely

resulted in fewer diaphragm muscle ﬁbers being recruited by

electrical stimulation and hence reduced overall diaphragm

muscle fatigue. Presumably, the contraction of the external

intercostal muscle allowed the expansion and stabilization of the

upper rib cage during the combined muscle stimulation. The

additional expansion of the upper rib cage increases the volume

of the thoracic cavity and decreases intra-alveolar pressure and

more air is drawn into the lungs. Thus, the external intercostal

muscle pacing facilitates inhalation and enhances respiratory

mechanics, thereby reducing the need for higher charge delivery

to the diaphragm to achieve the desired volume. Hence, this

approach can be used to reduce stimulation-induced fatigue of

the diaphragm muscle.

Since the animals were anesthetized with isoﬂurane, a

respiratory depressant, the respiratory rate decreases and results

in an elevation of the PaCO

. The increased breath volume not

only assists in triggering the Herring-Breuer reﬂex and

promoting entrainment, but also alleviates the increased PaCO

that results from slower breathing. During this study, we did not

observe signiﬁcant deviations from normative end-tidal CO

(etCO

) values, and thus did not study the relationship between

etCO

and stimulation parameters. The second aim of our

investigation was to assess the closed-loop controller’s ability to

generate sigh-like behavior by diaphragm-only or combined

diaphragm and external intercostal muscle stimulation. Currently,

some mechanical ventilation systems incorporate periodic sighs

for the re-aeration of collapsed alveoli (15). However, the

shortcomings of mechanical ventilation persist. Aside from

mechanical ventilation, available phrenic nerve pacing systems

can also be used to provide sigh breaths by an intermittent

increase of stimulus frequency (5). However, since all the phrenic

nerve pacing systems are open-loop systems, manual adjustment

of stimulus parameters is needed for inducing sighs. Our results

indicate that combined muscle stimulation can produce a larger

tidal volume than diaphragm-only muscle stimulation. Periodic

breaths with larger volume can help inﬂate more alveoli and thus

might prevent atelectasis (13, 21). Following the sigh, the

adaptive controller adjusted the stimulation parameters

automatically to return the ventilatory pattern to the desired

ventilatory pattern (5, 15). Thus, our approach provided an

automatic cyclic increase in stimulation amplitude to elicit a

large tidal volume without the need for human intervention.

Another beneﬁt of including periodic sighs into the pacing

paradigm is related to the entrainment of the paced breaths to

the intrinsic breaths. When diaphragmatic pacing at a ﬁxed cycle

period is implemented over an underlying spontaneous breathing

rhythm then the two rhythms must entrain themselves;

introduction of sighing was found to improve this synchrony.

When a loss of synchrony between the intrinsic breathing and

stimulation assisted breathing occurred before a sigh, the sigh

helped in resetting the intrinsic breathing pattern and aligning

the desired and measured volume patterns. Previous studies also

indicate that sighs function as a re-setter for intrinsic breathing

by restoring lung resistance and compliance back to a normal

level (21). The intrinsic central pattern generator gets feedback

during sighing since the large breath during a sigh activates

pulmonary stretch receptors. While we were able to induce

augmented breaths, we did observe instances of loss of synchrony

after the induced sigh. This occurred possibly due to the shorter

expiratory time in the induced sigh compared to the intrinsic

sighs. It may be possible to prevent this loss of synchrony by

extending the duration of the sigh, as entrainment with

mechanical ventilators has been associated with longer breaths

with lower ﬂow rates and higher volumes (22). Further

experiments are required to assess if post sigh loss of synchrony

is observed in animal models that have impaired breathing such

as after incomplete spinal cord injury, and if a longer induced

sigh duration would be beneﬁcial for re-synchronization.

Our empirical study in an animal model showed that the

synergistic approach of combined muscle stimulation of multiple

inspiratory muscles can be used to reduce diaphragmatic fatigue

onset, and for efﬁcient sigh-like breath elicitation. During the

implementation of diaphragm stimulation in clinical

environments for partients, the range of stimulation frequencies

typically spans from 20 Hz to a maximum of 50 Hz (23).

However, a stimulus frequency of 20 Hz is often preferred. When

diaphragm pacing is prolonged, it becomes necessary to readjust

the stimulation parameters to prevent fatigue caused by

stimulation. Typically, a decrease in stimulation frequency is

employed as a preventive measure against fatigue since higher

frequencies have the potential to induce muscle fatigue (24).

Low-frequency stimulation has been found to enhance the

endurance properties of electrically stimulated muscles by

transforming the composition of muscle ﬁbers from a mixture of

types to a predominantly type I ﬁber population. However, this

transformation also leads to a notable decrease in ﬁ

ber diameter

as well as decreases the capacity for maximum force generation

(23). Consequently, the ability to generate maximum inspired

volume is compromised.

Adury et al. 10.3389/fresc.2023.1199722

Frontiers in Rehabilitation Sciences 10 frontiersin.org

In our study, to optimize the ability to generate maximum force, a

stimulation frequency of 75 Hz was selected. Furthermore, to mitigate

fatigue induced by stimulation, incorporating additional pacing to the

external intercostal muscles has shown to be effective. This approach

enables the generation of higher tidal volumes while reducing the

amount of charge delivered to the diaphragm muscle. Besides, the

inspiratory cycle tidal volume was augmented by muscle pacing,

whereas expira tion was passive. As a result, expir ation-muscle-

assisted beha viors like coughing that are essential for clearing the

airway s of mucus, dust and germs, cannot be inﬂuenced. Hence, an

abdominal muscle stimulation paradigm could be an additional

component to the controller. This additional component would

allow pacing to offer a complete set of desired ventilatory behaviors

that included regular breaths, periodic sighs, and user triggered

coughs to assure good respiratory health.

The combined respiratory muscle pacing could be a promising

option for ventilator-dependent individuals with spinal cord injury.

In patients with a functional phrenic nerve, combined external

intercostal and diaphragm pacing might maintain full-time

ventilatory support. Abdominal muscle pacing would offer on-

demand cough. The adaptive controller with multiple muscle

pacing could also provide support by supplementing inspiratory

volumes in individuals with partial phrenic nerve lesions. Clinical

trials with human participants will be required to validate the

closed-loop adaptive control for respiratory pacing and verify the

beneﬁts of this novel approach.

Author summary

Restoration of ventilation through respiratory muscle pacing has

been shown to be a promising alternative to traditional mechanical

ventilation. Though there are multiple muscles involved in our

regular breathing, currently respiratory pacing only targets the

diaphragm muscle. The pacing is associated with diaphragm

muscle fatigue. Additionally, the pacing does not allow

intermittent sighs like in natural breathing. Here, we present an

adaptive respiratory pacing scheme that combines diaphragm

pacing with the pacing of external intercostal muscles, which also

contributes to inspiration. We investigate this system using a

rodent model and demonstrate that the fatiguability of the

diaphragm is signiﬁcantly lower when the diaphragm and external

intercostal muscle are paced together than diaphragm pacing

alone. Besides, this multiple-muscle pacing scheme assists in the

production of signiﬁcantly larger sigh-like breaths. Our approach

could be a promising option for long-term ventilation-dependent

individuals to improve overall respiratory health by delaying

respiratory muscle fatigue and introducing a sigh-like behavior.

Data availability statement

The raw data supporting the conclusions of this article will be

made available by the authors, without undue reservation.

Ethics statement

The animal study was reviewed and approved by Institutional

Animal Care and Use Committee at Florida International

University.

Author contributions

All work was developed and conducted at the Adaptive

Neural Systems Laboratory at Florida International University.

RA conduct ed the animal studies and processed and analyzed

all data. RS developed the computational model and provided

critical input on the design of the study. RJ supervised

the whole study and provided crucial feedback. RA, RS, and

RJ contributed to the manuscript writing and editi ng. All

authors contributed to the article and approved the submitted

version.

Funding

The work was supported by the National Institutes of Health

R01-NS086088 and the Agence Nationale de la Recherche ANR-

13-NEUC-0001 under the US-French Collaborative Research in

Computational Neuroscience program.

Acknowledgments

The authors would like to acknowledge Zachary C. Danziger

and Jacob McPherson for their suggestions and feedback during

the study, and Jefferson Gomes for his assistance during

experiments.

Conﬂict of interest

The authors declare that the research was conducted in the

absence of any commercial or ﬁnancial relationships that could

be construed as a potential conﬂict of interest.

Publisher’s note

All claims expressed in this article are s olely those of

the a uthors and do not necessarily represent those of their

afﬁliated organizations, or those of the publisher, the

editors and the reviewers. Any product that may be

evaluated in this article, or claim that may be made by

its manufacturer, is not guaranteed or endorsed by the

publisher.

Adury et al. 10.3389/fresc.2023.1199722

Frontiers in Rehabilitation Sciences 11 frontiersin.org

References

1. Kelly BJ, Luce JM. The diag nosis and manag ement of neuromuscular diseases

causing respiratory failure. Chest. (1991) 99(6):1485–94. doi: 10.1378/che st.99.6.

1485

2. Levine S, Nguyen T, Taylor N, Friscia ME, Budak MT, Rothenberg P, et al. Rapid

disuse atrophy of diaphragm ﬁbers in mechanically ventilated humans. N Engl J Med.

(2008) 358(13):1327–35. doi: 10.1056/NEJMoa070447

3. Costanzo MR, Ponikowski P, Coats A, Javaheri S, Augostini R, Goldberg LR, et al.

Phrenic nerve stimulation to treat patients with central sleep apnoea and heart failure.

Eur J Heart Fail. (2018) 20(12):1746–54. doi: 10.1002/ejhf.1312

4. Dean JM, Onders RP, Elmo MJ. Diaphragm pacers in pediatric patients with

cervical spinal cord injury: a review and implications for inpatient rehabilitation.

Curr Phys Med Rehabil Rep. (2018) 6(4):257–63. doi: 10.1007/s40141-018-0200-2

5. DiMarcoAF. Phrenic nerve stimulation in patients with spinal cord injury. Respir

Physiol Neurobiol. (2009) 169(2):200–9. doi: 10.1016/j.resp.2009.09.008

6. Posluszny JA Jr, Onders R, Kerwin AJ, Weinstein MS, Stein DM, Knight J, et al.

Multicenter review of diaphragm pacing in spinal cord inj ury: successful not only in

weaning from ventilators but also in bridging to independent respiration. J Trauma

Acute Care Surg. (2014) 76(2):303–10. doi: 10.1097/TA.0000000000000112

7. Romero FJ, Gambarrutta C, Garcia-Forcada A, Marín MA, Diaz de la Lastra E,

Paz F, et al. Long-term evaluation of phrenic nerve pacing for respiratory failure

due to high cervical spinal cord injury. Spinal Cord. (2012) 50(12):895–8. doi: 10.

1038/sc.2012.74

8. Ichihara K, Venkatasubramanian G, Abbas JJ, Jung R. Neuromuscular electrical

stimulation of the hindlimb muscles for movement therapy in a rodent model.

J Neurosci Methods. (2009) 176(2):213–24. doi: 10.1016/j.jneumeth.2008.09.015

9. Malešević NM, Popović LZ, Schwirtlich L, Popović DB. Distributed low-

frequency functional electrical stimulation delays muscle fatigue compared to

conventional stimulation. Muscle Nerve. (2010) 42(4):556–62. doi: 10.1002/mus.21736

10. Walter JS, Joseph P, Raymond D, Robert DS, Scott S, Kiratipath I, et al.

Stimulation of abdominal and upper thoracic muscles with surface electrodes for

respiration and cough: acute studies in adult canines. J Spinal Cord Med. (2018) 41

(3):326–36. doi: 10.1080/10790268.2017.1335447

11. Walter JS, Wurster RD, Zhu Q, Laghi F. Respiratory muscle pacing with

chronically implanted intramuscular permaloc electrodes: a feasibility study.

J Rehabil Res Dev. (2011) 48(2):103. doi: 10.1682/JRRD.2010.05.0086

12. Li P, Janczewski WA, Yackle K, Kam K, Pagliardini S, Krasnow MA, et al. The

peptidergic control circuit for sighing. Nature. (2016) 530(7590):293–7. doi: 10.1038/

nature16964

13. Patroniti MD, Foti G, Cortinovis B, Mag gioni E, Bigatello LM, Cereda M, et al.

Sigh improves gas exchange and lung volume in patients with acute respiratory

distress syndrome undergoing pressure support ventilation. Anesthesiology. (2002)

96(4):788–94. doi: 10.1097/00000542-200204000-00004

14. Severs LJ, Vlemincx E, Ramirez J-M. The psychophysiology of the sigh: i: the

sigh from the physiological perspective. Biol Psychol. (2022) 170:108313. doi: 10.

1016/j.biopsycho.2022.108313.2022/03/11

15. Mauri T, Eronia N, Abbruzzese C, Marcolin R, Coppadoro A, Spadaro S, et al.

Effects of sigh on regional lung strain and ventilation heterogeneity in acute

respiratory failure patients undergoing assisted mechanical ventilation*. Crit Care

Med. (2015) 43(9):1823–31. doi: 10.1097/CCM.0000000000001083

16. Siu R, Abbas JJ, Hillen BK, Gomes J, Coxe S, Castelli J, et al. Restoring

ventilatory control using an adaptive bioelectronic system. J Neurotrauma. (2019)

36(24):3363–77. doi: 10.1089/neu.2018.6358

17. Siu R, Abbas JJ, Hillen BK, Gomes J, Coxe S, Castelli J, et al. Autonomous

control of ventilation through closed-loop adaptive respiratory pacing. Sci Rep.

(2020) 10(1):21903. doi: 10.1038/s41598-020-78834-w

18. Fairchild MD, Kim SJ, Iarkov A, Abbas JJ, Jung R. Repetetive hindlimb

movement using intermittent adaptive neuromuscular electrical stimulation in an

incomplete spinal cord injury rodent model. Exp Neurol. (2010) 223(2):623–33.

doi: 10.1016/j.expneurol.2010.02.011

19. Abbas JJ, Chizeck HJ. Neural network control of functional neuromuscular

stimulation systems: computer simulation studies. IEEE Trans Biomed Eng . (1995)

42(11):1117–27. doi: 10.1109/10.469379

20. Jammes Y, Collett P, Lenoir P, Lama A, Berthelin F, Roussos C. Diaphragmatic

fatigue produced by constant or modulated electric currents. Muscle Nerve. (1991) 14

(1):27–34. doi: 10.10 02/mus.880140106

21. Li P, Yackle K. Sighing . Curr Biol. (2017) 27(3):R88–R89. doi: 10.101 6/j.cub.

2016.09.006

22. de Vries HJ, Jonkman AH, Tuinman PR, Girbes ARJ, Heunk s LMA. Respiratory

entrainment and reverse triggering in a mechanically ventilated patient. Ann Am

Thorac Soc. (2019) 16(4):499–505. doi: 10.1513/AnnalsATS.201811-767CC

23. DiMarco AF, Kowalski KE. High-frequency spinal cord stimulation of

inspiratory muscles in dogs: a new method of inspiratory muscle pacing. J Appl

Physiol. (2009) 107(3):662–9. doi: 10.1152/japplphysiol.00252.2009

24. Elefteriades JA, Quin JA, Hogan JF, Holcomb WG, Letsou GV, Chlosta WF, et al.

Long-term follow-up of pacing of the conditioned diaphragm in quadriplegia. Pacing

Clin Electrophysiol. (2002) 25(6):897–906. doi: 10.1046/j.1460-9592.2002.00897.x

Adury et al. 10.3389/fresc.2023.1199722

Frontiers in Rehabilitation Sciences 12 frontiersin.org