The Application of Thoracic Impedance-Based End-Tidal Carbon Dioxide Estimate in Cardiopulmonary Resuscitation: A Rat Study

Abstract

Thoracic impedance (TI) correlates with end-tidal carbon dioxide (ETCO₂) in large animals. This pilot study in Sprague-Dawley rats investigated whether TI can estimate ETCO₂ and dynamically guide compression depth. A dataset of TI and ETCO₂ measurements in rats was established to analyze the correlation between the two and construct a regression model. TI peak was strongly and positively correlated with ETCO₂ (r = 0.78, p < 0.001) and exhibited a progressive decay during prolonged compression. This pilot study demonstrated the feasibility of using TI-estimated ETCO₂ to guide compression depth in a rat model. The TI-guided strategy maintained ETCO₂ closer to the target value of 20 mmHg; however, no significant differences were observed between groups in ROSC rates, survival rates, blood gas parameters, or histopathological damage. Larger-scale studies are needed to evaluate clinical efficacy.

1. Introduction

Cardiac arrest (CA) is a life-threatening emergency. Without timely intervention, it rapidly progresses to irreversible multiple organ failure, particularly in oxygen-sensitive organs such as the brain and heart [1]. Out-of-hospital cardiac arrest (OHCA) represents a significant public health challenge, with a global average adult incidence of 55 cases per 100,000 individuals [2]. Yet survival remains dismal: the global average survival rate is only 7% [1], 8% in Europe [3], and 10.4% in the United States even after system-wide optimization of resuscitation care [4].

The American Heart Association (AHA) guidelines for cardiopulmonary resuscitation (CPR) and cardiovascular emergency care state that CPR is a critical intervention for treating CA patients [5]. To improve resuscitation quality, some studies have suggested using feedback devices during CPR [6,7]. In 2015, the AHA guidelines permitted the use of visual and auditory feedback devices during CPR in an effort to optimize the effectiveness of resuscitation in real time; these devices can be used to guide chest compressions and correct compression depth and rate [8]. Feedback on chest compressions can be obtained using accelerometers or force sensors; however, these sensors have not yet been widely adopted, particularly outside of research settings [9]. Beyond mechanical feedback, physiology-guided resuscitation—in which coronary perfusion pressure (CPP), diastolic blood pressure (DBP), and end-tidal carbon dioxide (ETCO₂) are monitored in real time to titrate compression quality—has gained considerable attention [10,11]. Incorporating individual physiological responses into the resuscitation protocol may improve rates of return of spontaneous circulation (ROSC) and survival [12,13]. Nevertheless, invasive hemodynamic monitoring is impractical during ongoing chest compressions in OHCA because it requires specialized personnel, stringent surgical conditions, and—critically—time that delays the start of monitoring and may interrupt CPR [14].

ETCO₂ is widely regarded as a key physiological parameter during resuscitation. It predicts ROSC [15,16], and the 2025 AHA guidelines affirm that ETCO₂ ≥ 20 mmHg is a predictor of ROSC [17]. Moreover, ETCO₂ is sensitive to compression quality: each 10 mm increase in depth raises ETCO₂ by approximately 1.5–2.5 mmHg [18,19], allowing it to serve as a feedback indicator of resuscitation effectiveness. Unfortunately, ETCO₂ monitoring requires endotracheal intubation and mechanical ventilation, procedures that demand skilled personnel and introduce a setup delay of roughly 4 min after the rapid response team arrives [20]. Consequently, ETCO₂ feedback is often unavailable during the critical early phase of out-of-hospital resuscitation.

A non-invasive, rapidly deployable alternative that correlates with ETCO₂ and hemodynamic indices would therefore be valuable. Thoracic impedance (TI)—electrical bioimpedance acquired through defibrillation pads—may fulfill this role. Previous work has demonstrated strong correlations between TI amplitude and cardiac output (CO) [21], CPP [22], and ETCO₂ [21,23,24], as well as between TI and compression depth [22,23,24]. As for TI itself, since most basic defibrillators can acquire TI signals via defibrillation pads, this may make it one of the earliest signals available in OHCA [25]. Therefore, TI has the potential to serve as an alternative indicator for monitoring the quality of chest compressions and estimating physiological parameters during CPR (particularly ETCO₂, which is relatively easy to obtain). This noninvasive measurement may facilitate the development of new CPR feedback algorithms [21]. However, current research remains focused on investigating the relationship between TI and physiological parameters, and has not yet succeeded in transforming TI from a monitoring parameter into a feedback control tool.

Previous studies have found a strong intra-animal correlation between TI and ETCO₂ in porcine models [23]. We aim to evaluate whether a similar relationship exists in rats, another commonly used animal model for CPR, and further investigate whether TI gradually decreases over time during compression. We chose a rat CPR model because it is well established, cost-effective, and provides a reproducible platform for testing feedback algorithms. Finally, this study aimed to establish a proof-of-concept framework for using TI as a feedback control variable to regulate compression depth.

2. Materials and Methods

2.1. Experimental Overview and Animal Allocation

We used male Sprague-Dawley rats (300 ± 20 g) obtained from Beijing Huafukang Biotechnology Co., Ltd. (Beijing, China). Animals were housed in a clean-grade facility at 25 °C under a 12 h light/dark cycle (2–3 rats per cage) with ad libitum access to food and water. Following a 1-week acclimatization period, rats were fasted for 12 h before the experiment but allowed free access to water.

The rat experimental protocol was approved by the Tianjin University Ethics Committee (Approval No.: TJUE2025-A-S-064). The protocol was completed and the data reported in accordance with the ARRIVE guidelines.

This study consisted of three phases using a total of 44 male Sprague-Dawley rats. The phases were designed to progressively establish the TI-ETCO₂ relationship, characterize signal behavior, and validate the control strategy.

Table 1. Overview of experimental phases, inclusion/exclusion criteria, and animal allocation.

Item	Phase I: Dataset Creation	Phase II: 5 min Compression	Phase III: Animal Testing
Purpose	Establish TI-ETCO2 correlation; construct regression model	Characterize time-dependent decay of TI during 5 min compression	Compare TI-guided dynamic compression vs. fixed-depth strategy
Experimental procedure	TI and ETCO2 recorded; establishing a simple linear regression model for TI-ETCO2	Compression depth fixed at d0 for 5 min continuous compression; TI recorded	Asphyxial CA; CPR for 10 min max; collect physiological data
Inclusion criteria	Body weight 300 ± 20 g; stable baseline heart rate and blood pressure	Body weight 300 ± 20 g; stable baseline heart rate and blood pressure	Body weight 300 ± 20 g; good TI signal quality; stable baseline heart rate and blood pressure
Exclusion criteria	Severe respiratory depression or circulatory instability post-anesthesia; failed or poor-quality TI/ETCO2 signal recording; surgical failure	Surgical failure; signal loss > 10% during 5 min recording	Anesthesia failure or complications; intubation failure or tracheal injury; catheterization failure; uncontrolled bleeding during asphyxia or CPR; failure to achieve ROSC within 10 min with incomplete data recording
Animals initially included	28	6	10
Animals excluded (reason)	5 (2 respiratory depression; 2 TI electrode detachment; 1 surgical failure)	0	0
Animals analyzed	23	6	10 (Control: 5; Experimental: 5)

summarizes the purpose, criteria, and animal allocation for each phase.

2.2. Dataset Creation

TI was measured using a custom-built bioimpedance acquisition circuit built around the MAX30009 analog front-end (Analog Devices, Inc., Wilmington, MA, USA). A four-electrode (tetrapolar) configuration was employed, with a constant-current excitation of 1.28 mA at a frequency of 50 kHz applied through the outer current-injection electrodes, while voltage was sensed via the inner pair of electrodes to minimize contact resistance errors [26,27,28]. ETCO₂ was measured using a small animal monitor (Vet12, Shenzhen Mindray Animal Medical Technology Co., Ltd., Shenzhen, China).

We set four compression depths—d₋₂, d₋₁, d₀, and d₁—using one-third of each rat’s anteroposterior thoracic diameter (d₀) as the reference [29,30,31]. Specifically, d_-−2 is 2 mm shallower than d₀, d₋₁ is 1 mm shallower than d₀, and d₁ is 1 mm deeper than d₀. Chest compressions were performed sequentially at depths d₋₂, d₋₁, d₀, and d₁, with each depth maintained for 10 s, resulting in a total of 40 s of data acquisition. We extracted 5 s segments from the middle portions of the TI signals at each of the four compression depths (Figure 1) and computed the mean TI value. ETCO₂ was recorded manually at the midpoint of each 10 s compression-depth interval.

2.3. Processing and Utilization of TI

We applied a sliding window peak detection algorithm [32] to identify peaks or troughs in the TI within each window (using peaks as an example here). Let the amplitude of the i-th peak be PA_i, and let N be the total number of peaks detected within the window. The mean peak PA is computed as:

$$PA={\displaystyle \frac{{\sum }_i^N{PA}_i}{N}}$$

(1)

In our study, the window size was fixed at 128 points, and the TI sampling rate was 100 Hz. Since the compression rate was 210 bpm, the window could accommodate four TI peaks. We averaged four consecutive TI peaks to generate a smoothed PA.

Compression rate was estimated from the same peak sequence. Let P₁ and P_N represent the sample indices of the first and last peaks, respectively, and Fs the sampling frequency. The rate in beats per minute (bpm) is then given by:

$$bpm={\displaystyle \frac{(N-1)\times Fs\times 60}{P_N-P_1}}$$

(2)

2.4. Analysis of TI Characteristics During 5 Min Compression

Lee et al. demonstrated that mechanical impedance decreased significantly with repeated chest compressions in a porcine model, attributing this attenuation to decreased elastic modulus of the thoracic structure, possibly involving partial rib fracture or joint damage [33]. However, there are very few high-quality studies reporting whether electrical impedance decreases during chest compressions. Therefore, we decided to investigate this issue.

Six rats were included in the study, and all six completed the 5 min continuous pressing test without any dropouts. The compression depth was set at one-third of the anteroposterior diameter of the rat’s thorax and remained constant throughout the compression process, with TI signals continuously recorded. The analysis period spanned 5 min following the start of compression. To quantify the trend of TI over time, the absolute error (AE) was defined as:

$$AE=P_1-P_N$$

(3)

where P₁ and P_N denote the first and final TI peaks, respectively.

Define the relative error (RE) as:

$$RE={\displaystyle \frac{AE}{P_1}}\times 100\%$$

(4)

Given the small sample size (n = 6) and paired design (repeated measurements within the same animal), we compared initial and terminal TI values using the exact paired-samples Wilcoxon signed-rank test (two-tailed $\alpha$ = 0.05). We summarized decay magnitude by computing the median absolute decay, median relative decay rate, and their corresponding interquartile ranges (IQR). The 95% confidence interval (CI) for the median absolute decay and median relative decay rate was estimated using the bias-corrected and accelerated (BCa) bootstrap method with 10,000 resamples. The statistical analyses were performed using MATLAB R2023b (MathWorks, Natick, MA, USA).

2.5. An Investigation into the Relationship Between TI and ETCO₂

We first assessed the correlation between TI and ETCO₂ using Pearson correlation analysis. We then fitted a simple linear regression model to quantify their association [34]. This method assumes linearity, independent residuals, normality, and homoscedasticity [35]. The model was specified as: ${ETCO}_2 ={ \beta }_0 + {\beta }_1 \times TI+ \epsilon$. Where ${\beta }_0$ denotes the intercept, ${\beta }_1$ the slope, and $\epsilon$ the random error term [36]. The ordinary least squares (OLS) method was used to estimate the regression coefficients, which minimizes the sum of squared residuals [37].

We verified the assumptions of simple linear regression through residual diagnostics. Residual plots were examined for patterns indicating non-linearity or heteroscedasticity. We used the Breusch-Pagan test to further assess the homoscedasticity [38]. Normality of residuals was evaluated using the Lilliefors test and Q-Q plots.

Given that multiple measurements (four compression depths) were obtained from each rat, the independence assumption of OLS regression was potentially violated due to within-subject correlation. We therefore computed cluster-robust standard errors (SE) via the Huber–White sandwich estimator, clustering at the rat level [39,40]. To gauge the impact of clustering, we examined the ratio of cluster-robust to conventional SE and calculated the intraclass correlation coefficient (ICC) to quantify within-rat correlation [41].

2.6. Animal Testing

We randomly allocated ten rats to the control (n = 5) and experimental (n = 5) groups by lottery. This study was designed as a pilot feasibility investigation. We selected the sample size (n = 5 per group) based on prior experience and practical constraints without conducting a formal power analysis. All rats completed the protocol; none were excluded.

Table 2. Baseline physiological data in rats.

	Control Group (n = 5)	Experimental Group (n = 5)	U Value	p Value
Weight (g)	301 (293–310.5)	304 (296–311.5)	11	0.84
Anteroposterior Chest diameter (cm)	3.4 (3.4–3.5)	3.5 (3.4–3.5)	10	0.63
Heart rate (bpm)	338 (325.5–382.5)	351 (336–366.5)	12	1.00
MAP (mmHg)	131 (121–141)	121 (117–128)	19	0.22

presents the baseline characteristics of the two groups. We observed no significant between-group differences in body weight, anteroposterior chest diameter, heart rate, or mean arterial pressure (MAP) (all p > 0.05).

We induced anesthesia with a smallanimal anesthesia machine (R500, Shenzhen Ruowode Life Science Technology Co., Ltd., Shenzhen, China) and maintained it with intraperitoneal sodium pentobarbital (60 mg/kg). Next, chest height and body weight were recorded. We performed endotracheal intubation with a 14G tube (with a metal guidewire) under direct visualization using a light source. A small animal ventilator (VentStar, Shenzhen Ruoward Life Science Co., Ltd., Shenzhen, China) was set to volume-controlled mode (fraction of inspired oxygen, 21%; ventilation rate, 70 bpm; tidal volume, 7 mL/kg). We monitored the electrocardiogram (lead II) using a multi-channel physiological acquisition system (MP160, BIOPAC, Goleta, CA, USA). Arterial and venous access was established using 24-gauge indwelling needles (pre-flushed with 5 IU/mL sodium heparin solution). Following 10 min of stabilization, we recorded baseline heart rate and arterial blood pressure. Subsequently, vecuronium bromide (1 mg/kg) was administered intravenously, and the trachea was clamped 2 min later. CA was initiated when the MAP dropped to 20 mmHg. CPR was performed immediately following 7 min of CA [42].

The control group used a small animal CPR device (KW-XF, Nanjing Calvin Biotechnology Co., Ltd., Nanjing, China), with the compression rate set to 210 bpm and the compression depth fixed at one-third of the anteroposterior diameter of the thorax. The experimental group used the same CPR device at the same rate, but the compression depth adjusted based on ETCO₂ estimated from TI. We administered adrenaline (30 μg/kg) starting 2 min after CPR onset and repeated the dose every 2 min up to a maximum of three doses. The maximum CPR duration was 10 min; if the rat did not achieve ROSC within 10 min, resuscitation was discontinued. ROSC was defined as a MAP exceeding 60 mmHg sustained for 10 min [43]. After 2 h of monitoring, the rat was weaned from mechanical ventilation and its incision was closed. For postoperative analgesia, the animals received a single subcutaneous injection of buprenorphine sustained-release (1.0 mg/kg) for postoperative analgesia.

In the experimental group, compression depth was adjusted every 10 s based on the TI-derived ETCO₂ estimate displayed to the operator. A single trained operator performed all adjustments, blinded to actual ETCO₂ values. At each 10 s interval, the operator reviewed the sequence of displayed ETCO₂ estimates and applied the following adjustment protocol: (1) if the estimated ETCO₂ was <18 mmHg, increase depth by 0.5 mm; (2) if the estimated ETCO₂ was >22 mmHg, decrease depth by 0.5 mm; (3) if within the target range (18–22 mmHg), maintain current depth. Adjustments were made using the KW-XF small animal CPR device, which permits precise depth modulation via mechanical control. The adjustment range was bounded by d₋₂ (2 mm shallower than d₀) and d₁ (1 mm deeper than d₀).

To minimize bias, all personnel involved in outcome assessment and data analysis were blinded to group allocation. Specifically:

(1) Physiological outcome assessors: The researcher performing arterial blood gas measurements and recording ETCO₂ values was unaware of group assignment, with samples and recordings labeled only by animal identification numbers.

(2) Histopathological assessors: Independent pathologists evaluated hematoxylin-and-eosin (HE)-stained brain and lung tissue sections. We coded the sections with random numbers before examination, and the pathologists remained unaware of group allocation during scoring.

(3) Data analysts: Researchers conducting statistical analyses used coded identifiers (Group A and Group B) during initial data processing. We disclosed group allocation only after all analyses were complete.

Primary outcome: The proportion of animals that achieved and maintained the target ETCO₂ level (≥20 mmHg) during compression. This endpoint directly reflects the effectiveness of the dynamic compression strategy guided by TI in maintaining adequate pulmonary blood flow and ventilation-perfusion matching.

Secondary endpoints: (1) ROSC rate; (2) survival rates at 24, 48, and 72 h post-ROSC; (3) arterial blood gas parameters; (4) histopathology at 72 h post-ROSC.

For arterial blood gas analysis, we used Origin 2024 (OriginLab Corporation, Northampton, MA, USA) software for data analysis. Quantitative data are presented as the median (IQR). We compared groups using the exact Mann–Whitney U test (two-tailed). A p-value of <0.05 was considered statistically significant.

3. Results

3.1. TI Analysis Results

The results of the MATLAB peak detection algorithm identifying TI peaks are shown in Figure 2a, where all peaks in approximately 38 s of TI signal were successfully identified (the green triangles indicate the locations of the TI peaks). Figure 2b shows a magnified view of the local TI signal waveform within the red dashed box in Figure 2a, with the peak amplitudes labeled.

The results of the MATLAB calculations are shown in Figure 3. The compression rate was 210.3 ± 0.9 bpm, with an average frequency error of 0.3 bpm and a coefficient of variation of 0.4%; the peak value was 2.6874 ± 0.0629 Ω, with a coefficient of variation of 2.3%. These results indicate that the peak detection algorithm is capable of calculating both the compression rate and the TI peak.

3.2. Analysis of TI Characteristics During Prolonged Compression

TI gradually declined during continuous compression (Figure 4). The AE of TI ranged from 0.21 to 0.66 Ω, with a RE of 7.0–19.0% (

Table 3. TI error.

	Initial TI (Ω)	Terminal TI (Ω)	AE (Ω)	RE (%)
1	3.04	2.55	0.49	16.1
2	3.18	2.61	0.57	17.9
3	2.97	2.56	0.41	13.8
4	3.54	3.06	0.48	13.5
5	2.99	2.78	0.21	7.0
6	3.47	2.81	0.66	19.0

). The paired-samples Wilcoxon signed-rank test confirmed a statistically significant decrease from initial to terminal measurements over the 5 min period (W = 21.0, exact p = 0.031). The median absolute decay was 0.485 Ω (IQR: 0.410–0.570); the 95% BCa bootstrap CI was 0.310–0.570. The median relative decay rate was 14.96% (IQR: 13.56–17.92%); the 95% BCa bootstrap CI was 7.0–17.9%.

Time-dependent TI decline during chest compressions has also been observed in porcine models [33]. Notably, the measured impedance above was mechanical, whereas our study assessed electrical TI, which reflects tissue conductivity. Although both parameters decrease during prolonged compression, the underlying mechanisms may differ. Mechanical impedance decline is attributed to structural damage to the thoracic cage (e.g., rib fracture, cartilage fatigue) and viscoelastic energy dissipation of the chest wall [44]. In contrast, the decrease in electrical TI observed in the present study may reflect progressive alveolar collapse, pulmonary edema, or altered intrathoracic fluid distribution, all of which reduce lung gas content and increase electrical conductivity, thereby lowering impedance [45]. Regardless of the specific mechanism, a key implication of the decline in TI is that the mechanical efficacy of chest compression may be diminishing over time, meaning that compression at a fixed depth cannot compensate for this attenuation. Therefore, real-time monitoring of TI—using changes in TI as a physiological feedback indicator rather than mechanical depth—allows for automatic adjustment of depth when attenuation occurs, thereby maintaining effective compression and avoiding excessive or insufficient compression caused by a “one-size-fits-all” depth setting. This study aims to elevate TI from a “monitoring indicator” to a “control target,” providing a proof-of-concept foundation for future closed-loop resuscitation strategies.

3.3. Regression Model

Figure 5 depicts the linear association between TI peak and ETCO₂. TI peak was strongly correlated ETCO₂ (Pearson r = 0.776, 95% CI: 0.679–0.846, p < 0.001). The OLS regression fit (red solid line) is described by: ETCO₂ = −5.759 + 8.560 × Peak. We computed all inferential statistics with cluster-robust SE to account for within-animal correlation. The intercept was −5.759 (cluster-robust SE = 1.660, 95% CI: −9.202, −2.317, p = 0.002), and the slope was 8.560 (cluster-robust SE = 0.602, 95% CI: 7.312, 9.808, p < 0.001). For every 1 Ω increase in TI peak, ETCO₂ increased by an average of 8.56 mmHg. The model was highly significant overall (Wald χ²(1) = 202.236, p < 0.001) and explained 60.2% of the variance (R² = 0.602, 95% CI: 0.542–0.686, based on 2000 bootstrap replicates with resampling at the rat level). Notably, cluster-robust SE were slightly smaller than conventional OLS SE (deflation factor: intercept 0.76, slope 0.82), likely due to weak or negative within-animal residual correlation arising from physiological ETCO₂ fluctuations. The blue dashed line indicates the 95% CI based on cluster-robust SE. The prediction accuracy was root mean square error = 3.90 mmHg, mean absolute error = 3.22 mmHg.

Figure 6 presents the regression diagnostics. Residuals were randomly distributed around zero, with no discernible funnel or curved patterns, supporting the assumptions of homoscedasticity and linearity; the Breusch–Pagan test yielded p = 0.207. Residuals approximated a normal distribution (Lilliefors test, p > 0.05). Q-Q plots displayed minor tail deviation, suggesting mild fat tails, which did not compromise the primary inferences. Intraclass correlation was extremely weak (ICC(1) = −0.237).

For each ETCO₂ estimate, four consecutive TI peaks were averaged, and the mean value was input into the regression equation (ETCO₂ = −5.759 + 8.560 × TI peak) to generate a ETCO₂ estimate. This estimate was displayed on-screen to the operator, refreshed approximately every four compression cycles (1.28 s). The target ETCO₂ was set at 20 mmHg, corresponding to a TI peak target of 3.0 Ω.

3.4. Results of Animal Studies

Figure 7 presents representative arterial pressure and electrocardiogram tracings during asphyxia and CPR. The upper panel of Figure 7a shows a continuous decline in arterial pressure in rats during asphyxia, while the lower panel is a close-up of the same data. Figure 7b illustrates the CPR phase leading to ROSC: compressions generated elevated systolic and diastolic pressures, epinephrine administration produced a marked pressor response, and ROSC was marked by a sudden sustained rise in arterial pressure. Figure 7c shows the situation where ROSC was not achieved. In rats that did not achieve ROSC, blood pressure following chest compressions exhibited low diastolic and systolic pressures, and there was no significant increase in blood pressure after epinephrine injection.

As shown in Figure 8, the control group (CON, n = 5, blue triangles) received fixed-depth compression, while the experimental group (EXP, n = 5, red squares) received dynamic compression guided by TI-based ETCO₂ estimate (with a target ETCO₂ of 20 mmHg and a corresponding TI peak target of 3 Ω); the dashed line indicates the target ETCO₂ level (20 mmHg). In the control group, ETCO₂ started at a lower level and mostly showed a downward trend, with high variability at the end (3–24 mmHg); only a few individuals briefly approached the target ETCO₂ level; In contrast, the experimental group started at a higher level and showed an overall upward trend, with endpoints concentrated between 20 and 30 mmHg; most individuals remained consistently close to or above the target ETCO₂ level. ROSC occurred in 4 of 5 rats in each group. Survival rates were comparable: 3/5 versus 4/5 at 24 h, 3/5 versus 3/5 at 48 h, and 3/5 versus 3/5 at 72 h.

Table 4. Blood gas analysis results for the control and experimental groups.

Parameter	Time Point	Control Group	Experimental Group	U Value	p Value
pH	BL	7.345 (7.316–7.381)	7.338 (7.315–7.356)	10	0.686
	ROSC 0.5 h	7.198 (7.149–7.275)	7.198 (7.134–7.233)	9	0.886
	ROSC 2 h	7.255 (7.225–7.288)	7.252 (7.240–7.314)	8	1.000
PaO2 (mmHg)	BL	98.000 (94.750–111.000)	99.000 (95.250–105.750)	8	1.000
	ROSC 0.5 h	82.000 (79.250–84.750)	83.500 (79.250–85.500)	7.5	0.971
	ROSC 2 h	94.000 (88.750–95.500)	94.000 (88.500–96.500)	7.5	0.971
HCO3− (mmol/L)	BL	22.550 (20.050–25.125)	21.950 (19.375–25.650)	9	0.886
	ROSC 0.5 h	15.950 (14.175–17.275)	15.300 (13.650–18.150)	9	0.886
	ROSC 2 h	18.000 (16.350–20.475)	18.050 (15.950–19.550)	9	0.886
SaO2 (%)	BL	97.500 (96.250–98.750)	97.500 (96.250–98.750)	8	1.000
	ROSC 0.5 h	91.500 (90.250–93.500)	91.500 (89.250–93.750)	9	0.826
	ROSC 2 h	95.000 (94.250–95.750)	95.000 (93.250–96.000)	8.5	1.000

summarizes arterial blood gas parameters. We detected no significant between-group differences at any time point (all p > 0.05).

Figure 9 illustrates the temporal evolution of blood gas parameters. Both groups of rats exhibited a decrease in pH 0.5 h after ROSC. Two hours after ROSC, pH levels in both groups had risen, indicating a trend toward recovery of acid-base balance. PaO₂ levels in both groups decreased 0.5 h after ROSC, followed by a significant rebound at 2 h post-ROSC. HCO₃⁻ levels decreased significantly after resuscitation and showed an upward trend at 2 h after ROSC. SaO₂ decreased slightly 0.5 h after ROSC, consistent with the downward trend in PaO₂, and rebounded 2 h after ROSC.

As shown in Figure 10, both the control and experimental groups exhibited loose cellular arrangement in the CA1 region, along with neuronal loss; some cells were smaller, hyperchromatic, and irregular in shape. The CA3 region of the hippocampus exhibited lower sensitivity than the CA1 region, with mildly disorganized cell arrangement. The DG region demonstrated significantly higher tolerance than the CA1 region; in the CA model, this region remained largely intact. Pathological changes, such as nuclear condensation and irregular cell shapes, were observed in cortical regions. Histopathological severity did not differ appreciably between groups.

Figure 11 shows pulmonary histopathology at 72 h. Both groups displayed interstitial hemorrhage, thickened alveolar walls, reduced airspace, and inflammatory infiltrates. Again, no qualitative between-group differences were evident on HE staining.

4. Discussion

In this study, we established a dataset to examine the TI–ETCO₂ relationship during chest compressions in rats. Our findings indicated a strong correlation between TI and ETCO₂ in the rat population. Based on this strong correlation, we further used simple linear regression analysis to establish a regression equation for the relationship between TI and ETCO₂. The equation indicated that for every 1 Ω increase in peak TI, ETCO₂ increases by an average of 8.56 mmHg.

We also quantified the time-dependent decay of TI during 5 min compression. A dedicated 5 min compression protocol confirmed that TI decline is common in rats, although its magnitude varies individually (absolute decay 0.21–0.66 Ω).

Within this pilot study, animals were randomized to two arms: the control group employed a personalized compression strategy (compression depth set at one-third of the anterior–posterior chest diameter), which represents an advanced resuscitation approach. Notably, 4 out of 5 rats (80%) in the experimental group maintained ETCO₂ levels above 20 mmHg during the later phase of compression, compared to only 2 out of 5 (40%) in the control group. No significant differences were observed between the experimental and control groups in terms of ROSC rate, survival rate, blood gas parameters, or organ injury, with results being comparable between the two groups.

The post-resuscitation blood gas profiles in both groups matched the temporal patterns reported previously in rat asphyxial cardiac arrest models [46]. Similarly, the histopathological findings, including neuronal loss in the hippocampal CA1 region and pulmonary hemorrhage with inflammatory cell infiltration, mirror patterns documented in CA and resuscitation studies [29,46].

Without a sham control in the present study, we cannot quantify the baseline level of physiological disturbance attributable to the experimental procedures themselves. Anesthesia and surgical trauma, including thoracotomy or vascular catheterization, activates systemic inflammatory responses and stress hormone release that may independently affect tissue histology [47,48,49]. Consequently, the specificity of our blood gas derangements and organ injury patterns as direct consequences of cardiac arrest and resuscitation—rather than confounding effects of anesthesia, mechanical ventilation, and surgical manipulation—remains uncertain. Future studies should incorporate a sham surgery group to delineate the true magnitude of injury specific to CA and resuscitation.

In a pediatric resuscitation model, ETCO₂-directed compressions maintained higher MAP than AHA-optimized CPR, yet ROSC rates were similar (70% vs. 65%), suggesting that physiological feedback may optimize hemodynamics without necessarily improving short-term survival [50]. The 2023 ILCOR CoSTR review identified four animal studies assessing ETCO₂-guided chest compressions, finding no differences in ROSC or survival compared to standard CPR [51]. These findings collectively suggest that while physiological feedback systems can improve real-time perfusion metrics, their translation into survival benefits requires further investigation in larger cohorts. The absence of significant differences between TI-guided and fixed-depth strategies in this pilot study is consistent with above study that physiological feedback systems may help maintain target physiological values during CPR without necessarily improving short-term survival in small animal models. However, our findings should be regarded as preliminary rather than conclusive.

This study utilized a model of asphyxiation-induced cardiac arrest in male rats. Although rats are widely recognized as a common experimental animal in CPR research, several anatomical and physiological differences limit the direct extrapolation of findings to humans. Nevertheless, the correlation between TI and ETCO₂ observed in rats is consistent with previous findings in porcine models, suggesting that this relationship may be conserved across mammalian species. The physiological feedback principle demonstrated in this study—namely, the use of TI as a surrogate marker for ETCO₂ to guide compression depth—is not species-specific and may be applicable to human CPR, particularly given that TI can be obtained non-invasively via defibrillation pads. Future studies in large animals and ultimately in human CA scenarios are necessary.

Outlook for the future:

Preliminary observations from this pilot study indicate that a dynamic compression strategy based on TI-estimated ETCO₂ maintains physiological targets during CPR. Future research could further explore aspects such as dataset size and the selection of models describing the relationship between TI and ETCO₂. Additionally, TI monitoring can be integrated with CPR devices to form a physiologically based closed-loop resuscitation system, further improving the timeliness and accuracy of depth adjustments.

The integration of TI monitoring into closed-loop feedback systems represents a promising engineering frontier. Current closed-loop feedback in automatic chest compressors incorporates real-time monitoring of physiological parameters to assess compression effectiveness and achieve individualized control [1]. However, these systems typically rely on invasive measurements. The non-invasive nature of TI acquisition, combined with its demonstrated correlation with both ETCO₂ and compression depth, positions it as a potential candidate for next-generation feedback devices. Recent advances in impedance-based ventilation detection during continuous CPR, achieving F1-scores exceeding 89% even in challenging clinical scenarios, further support the feasibility of multi-parameter TI analysis. Future systems could simultaneously feedback compression depth, rate, and ventilation adequacy from a single TI signal, addressing the current limitation that “feedback devices lack the ability to comprehensively assess and provide guidance on these critical aspects”.

Limitations:

This study has the following limitations. First, rib fracture assessment was not included in this study; therefore, the impact of rib fractures on chest compliance and TI measurements could not be evaluated. Second, the sample size was small (n = 5 per group), resulting in insufficient statistical power; therefore, the observed numerical differences (e.g., 80% vs. 40% achieving target ETCO₂) should be interpreted as exploratory. Third, the absence of a sham surgery control precludes determination of whether the observed blood gas derangements and histopathological changes reflect ischemia–reperfusion injury per se or confounding effects of anesthesia, surgical trauma, and mechanical ventilation.

5. Conclusions

This proof-of-concept pilot study demonstrates the technical feasibility of using TI-estimated ETCO₂ to guide dynamic compression depth in a rat CPR model. The TI-guided strategy maintained ETCO₂ closer to the 20 mmHg physiological target, but the small sample size (n = 5 per group) precludes any conclusions regarding clinical effectiveness or non-inferiority compared to fixed-depth strategies. The primary value of this study lies in establishing the TI-ETCO₂ correlation and the methodological framework for closed-loop feedback, which requires validation in adequately powered studies before clinical translation can be considered.