Mouse model for sensitivity of fluid measurement with textile electrodes

Abstract

Heart failure is a leading cause of morbidity and mortality, and fluid congestion complicates its management. Conventional home-monitoring methods, such as daily weight measurements, are insufficiently sensitive, while invasive techniques are impractical for routine use. This paper introduces a novel, non-invasive approach using textile-based dry electrodes for bio-impedance spectroscopy to detect subtle changes in thoracic fluid volume in a mouse model. Twenty-three mice underwent controlled fluid infusion into the thoracic cavity following a six-stage protocol. Bio-impedance data were collected over 256 frequencies ($3-1000$ kHz) and analyzed to extract $R_0$ and $R_{\infty }$, from which total ($V_T$), extra-cellular ($V_E$), and intra-cellular ($V_I$) fluid volumes were calculated and normalized to baseline. Statistical analyses included two-way ANOVA and multiple linear regression to correlate impedance measurements with animal length and weight. No statistically significant differences in normalized fluid volumes ($V_T$, $V_E$, $V_I$) were observed across infusion stages ($p>0.05$), though there was an anecdotal increase in $V_T$ ($\mathrm{\Delta }{V}_T=1.16\pm 1.79$ mL) and $V_I$ ($\mathrm{\Delta }{V}_I=1.80\pm 2.91$ mL) after fluid infusion. Multiple linear regression revealed moderate correlations between impedance measurements and weight ($r^2=0.33$, $p=0.035$, RMSE $=19.34$ g) as well as length ($r^2=0.34$, $p=0.0066$, RMSE $=1.00$ cm). Additionally, a strong correlation was found between length and weight ($r^2=0.81$, $p=5.65\times {10}^{-8}$). The results indicate that textile-based dry electrodes can non-invasively measure thoracic fluid volume in mice, although the sensor’s sensitivity to small changes is limited. Further refinement is needed to enhance sensitivity and determine sensitivity threshold. These findings support the continued development of textile-based bio-impedance sensors as a practical tool for non-invasive heart failure monitoring.

Introduction

Heart failure (HF) is the second most common cause of death in Canada, and currently affects more than 750,000 Canadians [1, 2]. More than 100,000 people are diagnosed every year and it is estimated to cost more than $\mathrm{\$}2.8$ billion annually by 2030 [1]. According to the World Health Organization, heart disease is the number one cause of death worldwide, accounting for $13\%$ of total deaths [3]. Since 2000, death caused by heart disease has more than tripled from 2.7 million to over 9 million deaths per year [3].

A key challenge in HF management is fluid congestion in the body [4]. Pulmonary and peripheral edema are signs that medications are not optimized and may require diuretic treatment or adjustment [4]. This is particularly important when patients are at home. A multimodal strategy is used in standard clinical follow-up, which includes the evaluation of biomarkers, symptoms (such edema and dyspnea), and home-based vital signs like blood pressure (BP) [5]. Nevertheless, these techniques are not suitable for everyday self-monitoring [6, 7]. For instance, evaluation of biomarkers typically require the collection of invasive blood sampling procedure, while observable (e.g., edema and dyspnea) and changes in vital signs (e.g., BP) frequently appear only after substantial hemodynamic congestion has already occurred. Consequently, one of the frequently suggested current at-home strategy to monitor HF is to ask patients to monitor their weight, which is unfortunately not an accurate measure of body water [10]. Weight monitoring guidelines differ greatly between organizations [10–12]. Daily weight measurement has not shown a high sensitivity to fluid decompensation and is often only evident when other symptoms are also present [10, 13]. Past literature has shown that compared to internal fluid monitoring, daily weight-based monitoring is significantly less sensitive ($23\%$ vs. $76\%$ sensitivity) and resulted in more unexplained detection (4.3 vs. 1.9/patient-year) [14].

Patients suffering from HF often experience troubling symptoms that reduce their quality of life are breathlessness, fatigue, and lower limb swelling [15]. In one study (n=540), more than half of the patients experienced breathlessness and fatigue more than once per week [15]. These patients also reported that they were moderately to extremely bothered by these symptoms [15]. Symptoms can also vary greatly between men and women, with more women experiencing hypertensive HF, and men having higher prevalence of ischemic HF [8]. Women also tended to experience more symptoms of HF, according to the functional classification of the New York Heart Association ($43\%$ vs. $38\%$ of men with class III or IV) [8].

There are two commercially available devices for measuring lung fluid: CardioMEMS and Remote Dielectric Sensing (ReDS) [16, 18]. CardioMEMS measures BP in pulmonary artery and has been shown to reduce hospital admission by $37\%$ over a mean follow-up period of 15 months [17]. However, this requires an invasive surgical procedure, and daily effort by the patient and care-providers to assess the health of the patient [17]. CardioMEMS is not widely accessible in HF management and is only recommended for the highest-risk patients. The other technology is ReDS, which is bulky, inconvenient and cannot measure lung fluid in those with an internal defibrillator [18]. Whether invasive or not, monitoring of lung water can have significant benefits for patients and the healthcare system through reducing costs. However, currently there is no device to accurately measure pulmonary congestion non-invasively at home.

Bio-impedance spectroscopy (BIS) measurement offers a potential monitoring solution for non-invasive fluid measurement. BIS has been used in clinical settings to track patients with HF receiving decongestion therapy, and has shown positive results for predicting 30-day and 1-year outcomes [19]. Clinical use of BIS has also been explored in other fluid-related diseases such as renal failure requiring dialysis [20]. Fluid measurement as an endpoint for dialysis as well as daily management is an important clinical tool for patient monitoring [20].

Current BIS sensors are based on gel electrodes which are not convenient for frequent monitoring at home. Gel electrodes are susceptible to drying out, which can affect the electrode–skin interface if they are used for long periods of time [22]. Instead of gel electrodes, dry electrodes have been used for convenient measurement of other signals such as electro-encephalography and electro-cardiography [23, 24].

To address the challenges with gel electrodes, only a handful of research work has been published where the author explored the potential of textile-based electrodes for thoracic and whole body BIS measurement. However, most of the reported textile electrodes have required pre-wetting or sweat accumulation to reduce skin electrode impedance, a hindrance of continuous use [25, 26]. Furthermore, researchers reported limitations due to electrodes displacement and motion artifacts that led to increased variability in the data compared to standard gel-based electrodes [27–29].

To overcome the challenges of gel electrodes and existing textile-based dry electrodes, we have developed and validated a prototype of dry electrodes for measuring BIS in the leg and chest [30]. First, the electrode design was optimized through iterative refinement of size, placement, and material composition and finalized a full-circumference silver yarn electrode that maximizes skin–electrode contact area without adding application complexity [31]. Second, we demonstrate that stable impedance measurements can be achieved without abrasive skin preparation by controlling the distribution of contact pressure through band tension and textile backing. This approach reduced electrode–skin interface impedance while maintaining comfort for long-duration wear. Third, the use of silver yarn to fabricate a plain weave structured textile electrode provided a mechanically robust electrode capable of repeated use that addressed a common limitation in prior textile-based sensors where conductive coatings degrade over time. Preliminary data in a healthy human population measuring calf and chest fluid have demonstrated that electrodes have a high correlation ($r^2>0.95,p<0.0001$) with gold-standard gel electrodes.

To examine the sensitivity of the textile electrodes to small volume changes using mice, we will explore the feasibility of BIS with textile-based electrodes with small volumes of fluid injected into the chest cavity. The goal of this study is to provide a proof of concept in an animal model that the developed electrodes are sensitive to the small changes in the thoracic fluid volume.

Results

The specimen information is summarized below in Table 1. This small population reflects an exploratory sample size for proof-of-principle measurement of BIS in mice.

Table 1.

Specimen information summary.

N	23(8F)
Weight	$27.35\pm 10.92$ g
Length	$8.66\pm 1.22$ cm

After normalization, none of $V_T$, $V_E$, or $V_I$ demonstrate significant differences between fluid infusion stages ($p=0.14$, $p=0.45$, and $p=0.13$, respectively). These results are shown in Figs. 1, 2, 3 below.

Fig. 1.

Boxplot comparison of $V_T$ versus stages of fluid infusion ($p=0.14$). CTRL control, PRE pre-infusion, SAL 1 mL saline, REC1 5 min post-infusion, DW 1 mL distilled water, REC2 5 min post-infusion

Fig. 2.

Boxplot comparison of $V_E$ versus stages of fluid infusion ($p=0.45$). CTRL control, PRE pre-infusion, SAL 1 mL saline, REC1 5 min post-infusion, DW 1 mL distilled water, REC2 5 min post-infusion

Fig. 3.

Boxplot comparison of $V_I$ versus stages of fluid infusion ($p=0.13$). CTRL control, PRE pre-infusion, SAL 1 mL saline, REC1 5 min post-infusion, DW 1 mL distilled water, REC2 5 min post-infusion

Anecdotally, it was observed that there was a non-significant increase in $V_T$ ($\mathrm{\Delta }{V}_T=1.16\pm 1.79$ mL, $p=0.32$) and $V_I$ ($\mathrm{\Delta }{V}_I=1.80\pm 2.91$ mL, $p=0.22$) over the entire test. This could indicate an uptake of the infused fluid into the cells within the thoracic cavity. The average magnitude of total fluid change was in line with the injected fluid volume of 2 mL ($p>0.05$).

Given the proof-of-concept scope of this study, we did not perform any prior power analysis of this study. However, to understand what the studied sample size ($N=23$), capable of detecting, we estimated the minimal detectable effect sizes using a two-sided $\alpha =0.05$. For a paired comparison, we found the minimal detectable standardized effect size of d = 0.58 and 0.68 for 80 and 90 percent power, respectively. Based on the observed variability for $V_T$ and $V_I$, these corresponds to minimal detectable changes of approximately $1.04-1.22$ mL for $V_T$ (SD = 1.79 mL), and $1.69-1.98$ for $V_I$ (SD = 2.91).

Although we did not observe statistically significant increase in the $V_T$ and $V_I$, the observed effect size for volume changes were moderate ($d=0.65$ for $V_T$, and $d=0.62$ for $V_I$), and falls within the detectable ranges. However, as the primary analysis involved repeated measures across multiple infusion stages, the statistical test required to detect an overall effect is less powerful than a single paired comparison. Consequently, the study may detect moderate stage to stage differences but still not reach significance in the omnibus test.

When multiple linear regression is performed between weight and the measured resistances, a moderate correlation is found ($r^2=0.33$, $p=0.035$, RMSE = 19.34 g). Length and resistance are similarly correlated ($r^2=0.34$, $p=0.007$, RMSE = 1.00 cm). Given the moderate observed correlation observed, we further applied a Bonferroni correction for the 2 primary anatomical comparisons (Length and Weight). Setting the adjusted significance threshold at $\alpha =0.025$, the correlation with length remains statistically significant ($p=0.007<0.025$). However, the correlation with weight ($p=0.035>0.025$), does not meet the adjusted threshold.

On the other hand, we demonstrated in our population that both length and weight are highly correlated ($r^2=0.81$, $p=5.65\times {10}^{-8}$). This is in agreement with previous literature exploring length–weight relationship in male mice [39].

The results of the correlations between length and weight with the measured resistances are shown in Figs. 4 and 5, respectively.

Fig. 4.

Actual and predicted length using $R_0$ and $R_{\infty }$ demonstrating positive correlation of measured resistance and animal length ($r^2=0.34$, $p=0.007$, RMSE = 1.00 cm)

Fig. 5.

Actual and predicted weight using $R_0$ and $R_{\infty }$ demonstrates positive correlation of measured resistance and animal weight ($r^2=0.33$, $p=0.035$, RMSE = 19.34 g)

Only the frequencies that were shared by all animals were included in the analysis for correlation with resistance. Correlations of weight with resistances at 3 and 6 kHz are shown in Figs. 6 and 7, respectively. Length and resistance are similarly shown in Figs. 8 and 9. Both weight and length are similarly correlated to resistances at 3 kHz ($r^2=0.142$, $p=0.076$) and 6 kHz ($r^2=0.005$, $p=0.74$).

Fig. 6.

Correlation between weight and resistance at low frequency (3 kHz) shows no significant relationship between weight and specific frequency resistance ($r^2=0.142$, $p=0.076$)

Fig. 7.

Correlation between weight and resistance at 6 kHz shows no significant correlation between weight and specific frequency resistance ($r^2=0.005$, $p=0.74$)

Fig. 8.

Correlation length and resistance at 3 kHz shows the same lack of significance as weight at the same frequency

Fig. 9.

Correlation between length and resistance at 6 kHz shows the same lack of significance as weight at the same frequency

For the shared frequencies between 3 and 6 kHz, the $r^2$ values decrease and p-values increase. The behavior of this change is shown in Fig. 10.

Fig. 10.

Behavior of $r^2$ and p-value between 3 and 6 kHz for weight correlation. No significant correlations were observed for frequency-specific resistances, trending to decrease $r^2$ and increase p as frequency increases

Discussion

This paper outlines the design and testing of our novel textile-based electrode for fluid measurement in animals. The main objective of this study was to determine the feasibility, and sensitivity of textile-based BIS to minor fluid change induced by infusion of saline and distilled water. We also explored correlation analysis between BIS-derived resistance with anatomical metrics of animal length and weight. Both agricultural and public markets can benefit from fluid tracking with textiles—such as detecting metabolic disorders that induce extra-cellular fluid compartments [40].

The effect of infusing more water into the body has the effect of lowering the resistance in the tissue. Using BIS with multiple frequencies, we are able to extract both $V_T$ and $V_E$ values. The current findings indicate no significant differences between stages of fluid infusion. This lack of statistical significance is probably due to a combination of biological variability and experimental noise that outweighed the signal of the small fluid challenge ($1-2$ mL), even though the mean total volume increased as predicted ($V_T$ = 1.16mL). Biologically, the observed variability might have stemmed from the the euthanasia process itself. Since $C{O}_2$, inhalation can trigger pulmonary hemorrhage and edema [44], it could potentially introduced biological noise in the form of unpredictable internal fluid shift. As a result, the $C{O}_2$ euthanasia likely masked the detection of the small, controlled fluid volumes we administered in the mice. Instrumentally, while the textile sensors accurately tracked anatomical metrics (length and weight with $r^2=0.34$), the high standard deviation (SD = 1.79–2.91) in fluid volume potentially suggests that variations in the textile–skin contact impedance might have created an instrumental noise floor that further limited the sensitivity to acute, micro-scale volume changes. The lack of significance suggests that the current textile-based electrodes may lack the sensitivity required to detect the small amount of infused fluid (2 mL). Previous studies have demonstrated that BIS-measured fluid volume can accurately measure different fluid compartments, with sensitivity limitations between $0.5-1$ mL [35].

Clinically, a weight gain of 2 kg is considered a key signal of fluid decompensation needing care [36]. In a mouse ($\approx 27$ g), 1 mL infusion represents a hypervolemic challenge of $\approx 3.7$ percent of body weight. Proportionally, this is equivalent to a 2.6 L fluid gain for a 70 kg human being. Consequently, although the absolute volume of $1-2$ mL appears small for this murine model study, this scaling demonstrates that our experimental challenge falls directly within the range of clinically significant acute decompensation observed in human patients.

Anecdotally, we observed small changes post-recovery, where fluid is redistributed into the surrounding tissue. This can be seen in $V_E$ by the apparent drop in fluid from the SAL-REC1 and DW-REC2 stage. Physiologically, the observed increase in $V_T$ reaffirmed that the infused bolus was retained within the thoracic cavity and not immediately cleared by the lymphatic drainage systems of the parietal pleura [45]. Furthermore, the corresponding behavior of $V_I$ reveals the cellular response to this retention: we observed a fluid shift from the extra-cellular space ($V_E$) to the intra-cellular space ($V_I$). This is consistent with the physiological response to a hypotonic environment [46]. As distilled water lacks solutes, it creates a steep osmotic gradient across the mesothelial membranes of the thoracic cavity [47], actively driving water from the pleural space into the cytosol of the surrounding tissues and resulting in the observed increase in intra-cellular volume.

In the mouse model, a much lower range of frequencies was observed to be effective for BIS measurement ($3-10$ kHz) compared with the full range of the equipment (256 discrete frequencies between 3 and 1000 kHz). While it was still possible to measure intra- and extra-cellular resistance with these frequencies using the Cole–Cole plot, it should be noted that the capacitive effects of membranes and non-conductive gaps in the thoracic cavity are not accurately measured at these frequencies. The effect may have led to measurement error of the total fluid volume, and intra-cellular volume by association.

Further correlation analyses indicates moderate correlation between BIS-measured resistances ($R_0$ and $R_{\infty }$) and animal weight ($r^2=0.33$, $p=0.035$), and length ($r^2=0.34$, $p=0.007$). Length and weight correlations suggest anatomical characteristics can be accurately reflected with BIS measurements of $R_0$ and $R_{\infty }$, which are derived from the Cole–Cole plot generated from multiple frequencies. However, when correcting for multiple comparisons (Bonferroni adjusted, $\alpha =0.025$), only the relationship with length remained statistically significant. This suggests that the textile sensors are primarily sensitive to the axial length of the conductive path, which is consistent with bio-impedance theory. While the sensors can estimate anatomical scale, the moderate $r^2$ values indicate that determining exact morphological metrics requires further sensitivity improvements. Individual frequencies showed non-significant correlations, and much lower correlation coefficients ($r^2<0.15$, $p>0.05$), and worsens with increasing frequency from 3 to 6 kHz. It is demonstrated that resistance at the lower frequencies may be a better predictor of animal size compared to higher frequencies.

This study has used an ex vivo model in fluid measurement, which has had an effect on fluid distribution and caused some fluid loss due to the tissue damage caused by euthanasia. It should also be noted that some blood mixed with surfactant was observed escaping from the airway in some mice during infusion. The fluid loss from the blood was minimal compared to the total fluid infused (1–2 drops over period of testing). No significant difference in measured fluid was found between the mice that had this small fluid loss.

During carbon dioxide euthanasia, lungs are susceptible to hemorrhage and perivascular edema [37]. When the saline and distilled water are infused into the thoracic space, the increased pressure on the lungs may force blood mixed with the surfactant through the airway and out of the nose. This appears as a deep red liquid with very small bubbles.

In vivo measurement offers an advantage for physiological monitoring to capture dynamic fluid shift and tissue response to infusion. While the ex vivo measurements provide a more controlled environment, reducing variability, this may not fully represent the physiological state due to the lack of systemic influence on fluid distribution. Studies have highlighted the differences in impedance measurements of in vivo and ex vivo models, emphasizing the limited capabilities for extrapolation from ex vivo findings to in vivo applications [38].

Outside of clinical care for patients with HF and other fluid-related disorders, previous studies have estimated the global market for animal biosensors at close to $\mathrm{\$}1$ billion [41]. Wearable sensors are most prominent in farming applications for monitoring animals, such as on-body sweat sensors for cows and pigs [42]. Tracking the health of animals is vital to avoid spread of disease and potential transference to humans through consumption of animal products [42].

Activity sensors are also a growing market to provide wearable tracking to assess pet health [43]. Current sensors for non-agricultural use are generally limited to on-body temperature, heart rate, and accelerometers—similar to smart watches and activity trackers for humans [43]. The presence of a private market for animal wearables provides another avenue for textile electrodes that can provide clinically relevant information on animal health and well-being.

The current study is primarily limited by the small sample size, which restricts statistical findings and generalization of the findings. The small number of animals has led to a large amount of variability, making it challenging to detect the small changes in fluid caused by infusion. Additionally, standard gel electrodes used with BIS were not used in the current study, which limits our ability to determine whether the error was a factor of the fluid infusion or textile-based error. The tissue damage caused by euthanasia may also have influenced our fluid measurements and tissue uptake.

A cylindrical model was also used to calculate mouse fluid volume, which has limited applicability to the true geometry of the animals. Since impedance is measured from forepaw to tail, the total length of the animal was used, as well as estimated average circumference from the current literature. This value is likely to be unreliable for the exact calculation of fluid since there is a clear change in circumference between the thoracic and abdominal sections of the animal. In the future, an exact circumference can be measured around the thorax and abdomen and volume can be calculated as a truncated cone with these two different values for $C_1$ and $C_2$.

Future work is required to improve the measurement of small fluid changes using BIS. Due to the small limit of fluid infusion in mice, we recommend future work explore a larger animal model to evaluate the limits of BIS with textile-based electrodes in terms of sensitivity to fluid volume change. Previous studies have demonstrated the feasibility of BIS monitoring in canine models during surgical procedures [34]. Implementing BIS during veterinary procedures could aid in monitoring fluid balance, and could be explored as a method for future testing of larger animal models in vivo.

In addition, surgical procedure monitoring could provide insight into fluid dynamics under anesthesia and during post-operative stress. Veterinary practice is another industry that can also benefit from textile-based electrodes, as gel electrodes may be difficult to administer, where textiles can be form fitted and more easily applied for animals.

Future work can also explore longitudinal studies to monitor fluid over long time periods during intervention, disease progression, and recovery. Performing this future work will help provide insight on how textile-based BIS can offer benefits in clinical applications to both traditional and veterinary medicine.

Conclusion

The primary objective of this study was to validate the feasibility of a novel textile-based electrode interface for bio-impedance acquisition in a small animal model. The study successfully demonstrated the feasibility of developed textile electrodes that could mechanically conform to the murine anatomy and maintained stable signal conductivity throughout the procedure without causing visible tissue irritation. Beyond feasibility, the study further explored the sensitivity of the textile electrodes to minor fluid changes induced by infusion of saline and distilled water, and performed correlation analysis between BIS-derived resistance and anatomical metrics. The sensors exhibited moderate to strong anatomical validity, as both length and weight had a significant positive correlation ($r^2=0.34$) with $R_0$ and $R_{\infty }$ measured with BIS.

Regarding fluid sensitivity, while the study did not demonstrate statistical significance across infusion stages due to sample size and interface variability, the results revealed consistent biological trends that warrant further investigation. Specifically, the mean total volume increases ($V_T$ = 1.16mL), aligned with the infusion magnitude, and the observed intercompartmental shifts from extra-cellular to intra-cellular space followed expected osmotic principles for a hypotonic infusion.

This distinction between statistical non-significance and biological consistency suggests that the limitation likely lies in the current prototype’s signal-to-noise ratio for detecting very small changes. It does not indicate a fundamental inability of textile sensors to detect fluid, which has already been validated in human participants for passive fluid movements within a body segment on the order of $200-300$ mL. The "biological noise" introduced by the ex vivo model, combined with the contact impedance variability of the dry textile interface, likely masked the micro-scale fluid changes ($<2$ mL) investigated in this murine model study. Consequently, these findings support the continued development of textile-based bio-impedance sensors. Future work should focus on reducing contact impedance noise and validating the technology in larger in vivo animal models, where the physiological signal exceeds the current noise floor, to fully realize the potential for convenient, at-home HF monitoring.

Finally, textile-based BIS for monitoring fluid and validation in murine model lays the foundation for a practical home-monitoring solution as the current HF management lacks convenient, non-invasive reliable techniques for detecting early congestion at home. In this study, we have demonstrated the fluid detection limits of textile electrodes which will support the refinement of these electrodes for clinical application and eventually contributein the life-saving impact of early intervention.

Methods

All animal experiments were performed considering the Animal Welfare Act and Animal Utilization Protocol approved by the University of Toronto. All study activities were performed at the Department of Comparative Medicine at the University of Toronto. The mice were humanely killed with carbon dioxide based on SOP 12.2.0 on Rodent Euthanasia.

Electrode design

Based on this teams previous work developing textile-based electrodes for fluid measurements in the chest and leg, we have created miniature textile electrodes to use on mice. The electrode material is a conductive silver/silver-chloride thread that is woven into a sheet of material. The conductive fabric was attached to a backing material with a stainless steel button connector to attach to the spectroscopy device for fluid sensing. A picture of the sensors is shown in Fig. 11.

Fig. 11.

Silver/silver-chloride sensors for bio-impedance spectroscopy of mice

All BIS measurements are taken using the ImpediMED SFB7 BIS device [32].

Specimen preparation

Mice used were not from a specific genetic strain, and a variety were used during collection of data. Length, weight, and sex were recorded for each mouse and euthanasia was confirmed individually. The mouse were then be placed on their back with nose pointing to the top of the work station.

The textile electrodes were placed on each mouse with at least 2-cm spacing on the tail and distal forelimbs. These anatomical areas were chosen because they have minimal skin–electrode impedance without the requirement for fur removal because they are naturally glabrous or sparsely haired.Electrodes were attached to the mice using strips of non-conductive plastic medical tape for the duration of data collection. The four electrodes are placed according to Fig. 12, below.

Fig. 12.

Ventral view of mouse with electrode locations for bio-impedance spectroscopy

Before the catheter was placed and fluid infusion began, a control measurement was taken of the mouse. Three measurements were taken sequentially over three seconds and the average was taken for each stage.

To place the catheter, the top of the rib cage was palpated on the right side to find a suitable point of injection. The catheter was then inserted at a 45 degree angle between the first and second rib until resistance was lost, indicating the tissue had been fully punctured. The angle of the catheter was then lowered to position it anterior to the lungs and inserted further, and the needle was removed. The syringe with saline was connected and secured rostral to the mouse, ensuring that it did not interfere with any of the electrodes.

Fluid infusion

Two rounds of fluid infusion were performed, allowing for 5 min of rest between for absorption and redistribution within the cavity. To infuse 1 mL saline distilled water, the syringe was manually depressed over 15 s. BIS measurement was performed at 6 stages; the 6 stages of fluid infusion are defined as:

• CTRL: Taken prior to fluid or catheter insertion

• PRE: Catheter placed in thorax with syringe attached, no fluid infusion

• SAL: Immediately after 1 mL saline is infused into thorax

• REC1: First recovery period, 5 min post-infusion of saline

• DW: Immediately after 1 mL distilled water is infused into thorax

• REC2: Final recovery period, 5 min post-infusion of distilled water

Post-infusion, the mice were dissected by making a transverse incision below the diaphragm to ensure that the fluid was contained within the thoracic cavity and the diaphragm was not punctured by the catheter or increased pressure.

Analysis

Prior to statistical analysis, calculated $R_0$ and $R_{\infty }$ were checked for outliers, with a cut-off of an absolute z-score = 3. No outliers were reported in this sample of specimen.

Total ($V_T$) and extra-cellular ($V_E$) fluid volume were calculated from the mouse size, and $R_{\infty }$ or $R_0$, respectively, determined by the BioIMP Software provided with the ImpediMED Device [32]. The ImpediMED SFB7 gathers resistance and reactance data at 256 frequency from $3-1000$ kHz [32]. These points are then subjected to analysis within their software to remove outliers outside of the automatic rejection limit of $1\%$ and find $R_0$ and $R_{\infty }$ using a Cole–Cole plot analysis to calculate the two y-intercepts of the fitted data [32, 33]. Fluid volumes were calculated from the following equations: (1)–(3) [21]:

$$\begin{array}{c}\hfill V_E=\frac{{\rho }^{\frac{2}{3}}}{3{(4\pi )}^{\frac{1}{3}}}\times {\left(\frac{L}{C_1{C}_2{R}_e}\right)}^{\frac{2}{3}}\times L\left(C_1^2,+,C_2^2,+,C_1,C_2\right),\end{array}$$ 1

$$\begin{array}{c}\hfill V_I=\frac{{\rho }^{\frac{2}{3}}}{3{(4\pi )}^{\frac{1}{3}}}\times {\left(\frac{L}{C_1{C}_2{R}_i}\right)}^{\frac{2}{3}}\times L\left(C_1^2,+,C_2^2,+,C_1,C_2\right),\end{array}$$ 2

$$\begin{array}{c}\hfill V_T=V_E+V_I,\end{array}$$ 3

where $C_1$ and $C_2$ are the circumference at the top and bottom of the animal, L is the length, $R_{0,\infty }$ is the measured resistance, and $\rho$ is the resistivity of the tissue [21]. Volume data were then normalized by subtracting the baseline volume at the CTRL stage of each specimen.

Differences between stages of fluid infusion were analyzed using Python-based statistics libraries to perform a 2-way analysis of variance (ANOVA) and Tukey post hoc analysis when required. ANOVA was conducted separately for each fluid volume measurement, $V_T$, $V_E$, and $V_I$. The independent variables chosen for the ANOVA are stage, sex, and weight, as well as the sex–weight interaction.

A multiple linear regression was performed for both length and weight, using $R_0$ and $R_{\infty }$ as dependent variables. The predicted length/weight was then correlated with actual value to determine the sensitivity to small differences between animals from the measured resistance values and Pearson R-values are reported for each.

The predictive value of the valid frequencies was also explored by comparing weight and length with a simple linear regression model. The Pearson R-squared value and corresponding p-values were calculated to determine which frequencies were most predictive for length and weight of the animals.

Author contributions

Piper: manuscript draft, protocol development, data collection, data processing, statistical analysis Hoque: aid in manuscript preparation Montazeri G.: aid in protocol development, data processing, statistical analysis Simmons: aid in protocol development Yadollahi: aid in manuscript preparation, protocol development, data processing, statistical analysis.

Funding

This work has been supported by funding provided by Mitacs.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

All experiments were performed under the approval of AUP 20012764, overseen by the University of Toronto.

Competing interests

The authors declare no competing interests.