Non-invasive acquisition of vital data in anesthetized rats using laser and radar application

Introduction

To implement the ‘3Rs’ principle and expedite drug discovery research, rapid acquisition and highly accurate vital data are required in evaluation of the effects of new vaccines and in drug testing and development. Selecting the appropriate anesthesia protocol for each animal and monitoring vital signs during anesthesia are also very important for better anesthesia management and reproducible experimental data acquisition in laboratory animal experiments. Furthermore, the large amount of obtained experimental data must be managed consistently after digitization.^1,2 Specifically, it is important to non-invasively acquire the behavior and vital signs of target animals and analyze data in real time.^2,3 In addition, the use of invasive methods such as implantable sensors for drug evaluation and general pharmacological studies is not less burdensome for both the operator and the animals.

In recent years, significant advances have been made in the miniaturization of laser and radar sensors, particularly those used in self-driving automobile technology. Integration of these two kinds of sensors has enabled real-time detection of moving vehicles, pedestrians and obstacles. ⁴ The benefits of miniaturization are not limited to this domain alone, and extend to various other sectors. Of note, laser and radar sensors can detect subtle movements on the body’s surface, such as those caused by heartbeats and respiration, even from a distance, in fish, various animals and human.^{5 –9} This non-contact measurement capability has led to the widespread use of radar in medical devices, with applications including health monitoring for the elderly and mental stress monitoring, among others.^10,11 As a working hypothesis, there is a high possibility that a wide variety of information could be obtained non-invasively using laser and radar sensor selected frequency bands that have little effect on living organisms.

The main purpose of our research group is to verify the possibility of obtaining vital sign information using a laser and radar sensor in a manner that is non-invasive and painless for test animals. As a result of verification, we constructed a laser and radar system that non-invasively obtained the respiration rate (RR) and heart rate (HR) of anesthetized rats.

Material and methods

Measurement protocol

All measurements were obtained under anesthesia by intraperitoneal (i.p.) injection of three mixed anesthetics, medetomidine (Domitor, NIPPON ZENYAKU KOGYO CO., LTD, Fukushima, Japan) 0.15 mg/kg body weight, midazolam (Midazolam Injection Sandoz, Sandoz K. K., Tokyo, Japan) 2 mg/kg body weight, and butorphanol (Vetorphale, Meiji Animal Health Co., Ltd, Kumamoto, Japan) 2.5 mg/kg body weight. ¹² The animals were evaluated using sedation scores reported in previous studies with a score of 4 or higher out of 5 being adequate.^12,13 After confirming deep anesthesia, all rats were sequentially placed in prone and supine positions on an integrated multiple-channel intraoperative monitor (Rodent Surgical Monitor+® (RSM), INDUS Instruments Engineering, TX, USA). Body temperature of each rat was monitored using a rectal temperature probe connected to the RSM. During vital measurements, body temperature was maintained at 37 ± 1°C by a heating system integrated to the RSM. Furthermore, a series of sensors was used for vital data acquisition (Figure 1(a) and (b)). The following items were measured from each device. RSM: electrocardiogram, respiratory waveform, body temperature, digital HR and RR. Laser sensor (IL-S065, Keyence, Osaka, Japan): waveform including RR. Radar sensor (NJR4262J: 24GHz, Nisshinbo Microdevice Inc., Tokyo, Japan): respiration and HR. Data from RSM, the laser sensor and radar sensor were recorded on a laptop computer (OS: Windows10, Panasonic, Osaka, Japan). In all experiments, the interval time for collection and comparison was set to 15 s per epoch. The laser sensor was fixed at distance of 65 mm and vertical direction from the rat’s measurement point using a magnetic stand. The radar sensor was fixed at distance of 100mm and 45° angle from the rat’s chest using an arm stand. In addition, video recording of the abdomen and chest area of each rat and the RSM screen monitor was obtained during measurements (GoPro9 and iPad, Figure 1(a) and (b)). The reason for this monitoring was to visually confirm the location where the laser measures respiratory motion, and the number of respiratory movements. First, the amplitude of respiratory movement recorded by the laser sensor was collected three times for 30 s each from six positions – at three locations of the prone and supine positions in the first rat – and compared with the amplitudes of respiratory movements in the video. The recording positions in the prone position were marked and set by ruler above the last rib(P2) and 2 cm cranially (P1) and caudally (P3) from the top of the last rib. In the supine position, it was marked and set in a similar manner at the tip of the sternum(P5), and 2 cm cranially (P4) and caudally (P6) from the sternum (Figure 1(a) and Figure 2(b)). Next, HR and RR were simultaneously measured using the laser and radar sensors. All rats were sequentially placed in prone and supine positions and measured three times for 30 s each by the radar sensor and RSM, for a total of six measurements. The RR by laser sensor was measured once from each of the six positions mentioned previously, and video recording was acquired in all measured animals. To compare with the classical method by stethoscope, the HRs were recorded with a digital stethoscope (Nextetho, SHAREMEDICAL, Tokyo, Japan). However, the recording of auscultatory can affect the measurement of other sensors owing to contact with the animal. Therefore, HRs were recorded with the digital stethoscope once for 30 s in the supine position and then measured with the other sensors. Finally, to determine response to drug administration, HR and RR data from each sensor were measured in all supine positions by administering noradrenaline (2 mg/kg, i.p., Alfresa Pharma Corporation, Osaka, Japan) only once and comparing it respectively. The measurement time was 400 s, including 90 s before i.p. administration, 10 s of i.p. administration, which was excluded from data comparison, and 300 s after i.p. administration. Injection of noradrenaline was performed by holding the abdominal skin using a hooked forceps to avoid displacing the rat position from the measurement source as much as possible. After all measurements, the anesthetized rats were injected i.p. with atipamezole hydrochloride (0.15 mg/kg body weight, Atipame Injection, Kyoritsu Seiyaku Co., Tokyo, Japan) for anesthesia recovery.

OPEN IN VIEWER

Figure 1.

Schematic representation of the measurement equipment and data acquisition flow. (a) Overview of the measuring equipment. All data were recorded simultaneously on a computer via an A/D converter. (b) Snapshot of digital video during data recording shows the arrangement of the heating plate for maintaining body temperature, the rat and placement of the laser device. The red marker point (arrow) on the body of the rat is emitted from the laser sensor and indicates the measurement position. (c) Time-series variation of the relationship between the laser sensor data (black line) and the maximal expansion points of abdominal up-and-down motion during breathing by visual observation through the video recording (red circles). ECG: electrocardiogram

OPEN IN VIEWER

Figure 2.

Waveforms and amplitudes from laser data during respiration at the six measurement positions. (a) Waveform, respiration rate (RR) and standard deviation from laser data at each measurement position from an anesthetized rat during 30 s. The RR from the fast Fourier transform (FFT) analysis of the laser data and visual counts from video recordings were compared. All waveforms from laser data show only 10 s of data. Detailed information about positions P1 to P6 is shown in the text. (b) Mean values and standard deviations are shown for the measurement positions indicated as red circles. ***Analysis of variance, p < 0.001.

Data analysis and statistics

Data analysis and plotting software were performed using Excel (Office 2019, Microsoft, WA, USA). The RSM, the laser and radar sensor data were recorded simultaneously and input to a PC via an A/D converter (USB-6003, National Instruments, TX, USA) at a sampling rate of 1 kHz (LabVIEW V2021, National Instruments, TX, USA). The number of respiratory movements at each recording position collected from the laser sensor was converted to breaths per minute, and compared with the number of visual counts from the video recording (ref-RR). The maximum amplitude of the six respiratory motion measurement positions obtained from the laser data was evaluated as normally distributed and compared using one-way analysis of variance. HR reference data (ref-HR), which were obtained by counting the number of R-wave peaks using electrocardiogram (ECG) data from the RSM (Figure 3) were converted into beats per minute. To calculate RR and HR from the laser and radar data, those data from each epoch were analyzed digitally using the fast Fourier transform (FFT) method (GUN Octave V6.2). Digital bandpass filter was not used intentionally. The frequency of the peak position was read and converted to beats per minute and breaths per minute. The I and Q channels of the time-series data were obtained as Doppler signals from the radar data. Both the I and Q channel data were frequency-analyzed, and channels in which a peak with high intensity appeared were selected as the RR and HR data. In the comparison of visual counts, the laser and radar data, the average values and standard deviation (SD) of RR and HR were obtained from each test rat, and the p value was calculated by paired sample t-test. p < 0.05 was denoted as statistically significant. Regarding the equivalence between the laser sensor and the ref-RR data, and between the radar sensor data and the ref-RR and ref-HR data, if the results from each sensor were within ±5% of the reference data at each epoch, this value was considered the same as the reference data. The fluctuation before and after noradrenaline administration between the RR values from the radar sensor and the laser sensor, and between the HR values from the radar sensor and the ECG was considered the same if the results from each sensor at each epoch were within ±10% in the same time-series section. Measured animals were divided into Group 1 (>80%), Group 2 (60–80%), and Group 3 (<60%) based on percent agreement at each epoch in the total measurement time.

OPEN IN VIEWER

Figure 3.

Data recordings obtained by the Rodent Surgical Monitor+® (RSM), the laser and laser sensor from an anesthetized rat. (a) The I-signal of the radar sensor during 15 s. (b) Q-signal of radar sensor during 15 s. (c) The fast Fourier transform spectrum distribution of the I- and Q-signals shows the peak respiration and heart rate frequencies as well as other harmonic peaks during the 15-s measurement period. (d) Simultaneous time-series of the I- and Q-signal radar data, electrocardiogram (ECG) from RSM, and respiratory movement data from the laser sensor.

Results

The laser sensor can measure the distance from beam emission to the measurement site with high accuracy. It was suitable for identifying the position of maximum amplitude during respiration (Figure 1(c)). The RR values obtained by the laser sensor in the six measurement positions during breathing in anesthetized rats were not significantly different from the number of ref-RR at all measurement positions (ref-RR from P1 to P6, 61 ± 2.2, 64 ± 2.8, 67 ± 5.4, 62 ± 2.8, 63 ± 3.0 and 65 ± 2.2, respectively; laser from P1 to P6, 60 ± 0.0, 63 ± 0.5, 66 ± 1.2, 62 ± 0.5, 62 ± 0.0 and 65 ± 0.5, respectively (p = 0.516–0.863)) (Figure 2(a)). Amplitude during respiratory movement was significantly greater at laser position P3 than at any other laser position (Figure 2(b)) (p < 0.001). The mean RR values obtained from FFT analysis of the laser and radar data (68 ± 5.9 and 68 ± 5.7, respectively) corresponded to the ref-RR (69 ± 6.2) (Table 1). There was no statistically significant difference in RR between either the mean FFT laser data or the mean FFT radar data with the ref-RR data for any rat (p = 0.233–1.000, and 0.213–1.000, respectively) except for Rat #7 (p = 0.038–0.048). Mean equivalence between the reference and laser RR data was 83%, and between the reference and radar RR data was 86%. Table 2 summarizes the comparison results of HR obtained by FFT-radar data as well as the ref-HR. There was high average equivalence (80%) between the ref-HR data and the HR radar data calculated by the FFT method. There was a slight difference among all rats between the reference HR values (336 ± 19.8) and the radar HR values (348 ± 34.1) calculated by the FFT method, but there was no statistically significant difference between the values for any individual rat (p = 0.102–0.952). The HR values by digital stethoscope (340 ± 23.0) were also measured only once in each rat and corresponded to other sensor data. Figure 3 shows simultaneous time-series of the I- and Q-signal radar data, the ECG data from RSM, and respiratory movement data from the laser sensor. The equivalence of the RR and HR values from each sensor confirms that the respiratory and ECG waveforms are synchronized with the radar waveform (Tables 1 and 2). The fluctuation in HR and RR before and after noradrenaline administration in anesthetized rats was compared for six epochs, 90 s and 20 epochs, 300 s, respectively. Rats were divided into three groups, with 3/11 animals (Group 1) having >80% concordance between reference and radar data within ±10% error, 6/11 animals (Group 2) showing variable but consistent changes in the 60–80% concordance range, and 2/11 animals (Group 3) having <60% concordance for both HR and RR data (Figure 4).

OPEN IN VIEWER

Table 1.

Frequency analysis of the respiratory rate data.

	Respiration rate (breaths/min)			p value		Equivalence (%)
Rat no.	Ref-RR a (Mean ± SD)	Laser-FFT b (Mean ± SD)	Radar-FFT c (Mean ± SD)	Ref-RR vs. laser-FFT	Ref-RR vs. radar-FFT	Ref-RR vs. laser-FFT	Ref-RR vs. radar-FFT
1	74 ± 1.4	75 ± 1.9	75 ± 1.5	0.363	0.096	100	100
2	63 ± 1.9	63 ± 1.9	63 ± 1.5	1	0.551	100	83
3	65 ± 2.5	65 ± 1.9	65 ± 1.9	1	1	100	83
4	68 ± 2.6	69 ± 1.5	69 ± 1.5	0.631	0.631	83	83
5	66 ± 1.2	67 ± 3.0	66 ± 2.0	0.657	1	67	83
6	85 ± 5.7	82 ± 6.0	82 ± 6.0	0.438	0.438	100	100
7	72 ± 1.6	69 ± 2.7	69 ± 1.9	0.048	0.038	67	83
8	65 ± 1.9	64 ± 2.3	63 ± 2.7	0.341	0.213	100	67
9	67 ± 3.2	66 ± 5.5	67 ± 5.0	0.736	0.903	50	100
10	61 ± 2.7	59 ± 3.8	62 ± 2.0	0.233	0.703	50	75
11	68 ± 4.5	66 ± 3.8	65 ± 4.3	0.544	0.307	100	83
Average	69 ± 6.2	68 ± 5.9	68 ± 5.7	0.545	0.534	83	86

p values were calculated by t-test. Equivalence values (ratio of laser or radar data within ±5% of the visual count data) are shown for individual rats.

a

Value of visual count data from video recordings.

b

Value from the fast Fourier transform (FFT) analysis of the laser data.

c

Value from the FFT analysis of the radar data.

OPEN IN VIEWER

Table 2.

Frequency analysis of heart rate data.

	Heart rate (beats/min)			Ref-HR vs. radar-FFT
Rat no.	Ref-HRa(Mean ± SD)	Radar-FFTb(Mean ± SD)	Stethoscope	p value	Equivalence (%)
1	301 ± 2.7	325 ± 54.9	318	0.377	83
2	321 ± 5.5	319 ± 3.8	330	0.102	100
3	319 ± 3.6	317 ± 4.9	300	0.363	83
4	345 ± 7.8	347 ± 12.7	324	0.444	83
5	372 ± 2.3	373 ± 3.6	366	0.611	100
6	360 ± 3.3	359 ± 22.1	372	0.952	50
7	346 ± 3.1	368 ± 42.0	348	0.274	66
8	315 ± 5.4	315 ± 6.0	318	0.363	100
9	343 ± 3.0	340 ± 12.6	366	0.625	83
10	331 ± 5.4	346 ± 40.2	333	0.428	50
11	341 ± 3.6	333 ± 13.6	360	0.315	83
Average	336 ± 19.8	348 ± 34.1	340 ± 23.0	0.441	80

p values were calculated by t-test. Equivalence values (ratio of radar data within ±5% of the reference data) are shown for individual rats.

a

Value of the R-wave peaks using electrocardiogram data from the Rodent Surgical Monitor+®.

b

Value from fast Fourier transform (FFT) analysis of the radar data.

RR: respiration rate; Ref-RR: respiration rate reference data.

Ref-HR: heart rate reference data.

OPEN IN VIEWER

Figure 4.

The fluctuation in heart rate (HR) and respiration rate (RR) obtained by radar relative to the reference values according to pattern of change before and after noradrenaline administration (intraperitoneally). Group 1: corresponding fluctuations (>80%) in the reference and radar data before and after noradrenaline administration. Group 2: slight difference (60–80%) between the reference and radar data. Group 3: large fluctuations (<60%) in the radar data for HR and RR relative to the reference data. The black arrows indicate the timing of noradrenaline administration. ECG: electrocardiogram

Discussion

A data acquisition method that reduces the stress experienced by experimental animals is highly likely to enable accurate evaluation of drug effects and their effects on the body and also improve animal welfare.^{3,14 –16} There is increasing demand to get results rapidly from in vivo animal experiments conducted to develop treatments for acute conditions caused by unknown diseases and viruses.^1,2 Implantable sensors are commonly used to acquire behavioral activity data and vital signs in experimental animals. A recovery period is utilized to eliminate post-operative physiological changes associated with a general anesthesia and a surgical procedure. However, it has been difficult to minimize its impact on the data. There are also several non-invasive rodent monitoring devices available (physiosuite®, Kent Scientific Corporation, Torrington, USA; MouseOx® PLUS, STARR Life Sciences Corp.; MARTA, Vigilitech AG, Heiden, Switzerland). These products are used for a variety of studies, but there is always a need for more user-friendly products for different purposes. In general, it requires an enormous amount of time and specialized staff for researchers to visually classify and quantify behavioral changes from videos that record detailed behavioral changes before and after administration of test drugs. Human observation tends to cause problems in data accuracy due to subjective contamination. In addition, the use of implantable devices requires time and specialized staff. This applies not only for data interpretation but also to mastering the surgical procedure required for the implantation, which is invasive and complex. To solve this problem, it is necessary to quantify biological information and behavior using non-invasive methods suitable for each experimental animal. Ultimately, it will be important in the future to utilize digital technology to analyze large amounts of experimental data in a short period of time and obtain useful information from it.

In this study, we investigated the possibility of acquiring vital sign data using a non-contact laser and radar sensor that reduces unnecessary stimulation and stress in live animals, which has been a challenge in the field of research and development using experimental animals. Our results aim to non-invasively quantify the vital signs of free-ranging experimental animals and may further lead to the integration of digitized experimental data into data servers. This aim has not been achieved in the field of research and development using experimental animals yet, despite the availability of excellent off-the-shelf non-invasive rodent monitoring devices. To clarify the equivalence of the laser and radar data with reference data, we tabulated and analyzed the data obtained from each animal in this study. Anesthetized rats were used, because the movement of the animals might affect the measurement results. Although previous studies have reported methods for non-invasive and non-contact HR and respiration measurements in animals, including in experimental animals, these methods have not been evaluated relative to definitive reference data.^17,18 In other words, the HR and RR data obtained by radar in these previous reports have not been thoroughly evaluated in comparison with the vital sign data obtained from other devices. In addition, rats have significantly faster heart and respiratory rates than humans (rat HR: 250–450 beats/min, RR: 70–115 breaths/min). Therefore, we considered that it was necessary to use a sampling rate >1 kHz in order to detect the peaks of the ECG and respiratory waveforms for use as the reference data.

In quadrupeds such as rodents and dogs, thoracic respiration strongly involves the rectus abdominis, external oblique, internal oblique, and transversus abdominis muscles that move the ribs covering the lungs.^14,19,20 Thoracic respiration is maintained by these muscles; in contrast, movement of the diaphragm is termed abdominal respiration. Respirations of the present anesthetized rats were easily recognized visually as vertical movements on the skin surface near the dorsal last rib and the ventral diaphragm (Figure 1(c) and Figure 2(a)). Accordingly, the obtained respiratory waveform included the combined movement of the ribs in thoracic respiration and movement of the diaphragm in abdominal respiration, and the waveform with maximum amplitude was obtained at position P3 (Figure 2(b)). However, the vertical movement observed at the laser peak timing was not statistically significantly different from that observed in the video recording at all measurement positions from P1 to P6 in the rat. Therefore, a respiration waveform of sufficient amplitude could also be recognized in data obtained at any position (Figure 2(a)). This finding suggests that respiration could be easily measured by laser sensor on either the dorsal or the ventral side.

The mean values of the ref-HR and ref-RR data were almost equivalent to those obtained by frequency distribution of the laser and radar waveforms in the same individuals (Tables 1 and 2). Comparison of respiratory rates among all rats showed a statistically significant difference only for rat #7, which is likely due to the very low variance in the data. The percentage of laser and radar data within ±5% of the RR visual counts data ranged from 50% to 100% among individuals. The correspondence of the RR data obtained from the laser and radar waveforms to the ref-RR data demonstrates that the non-invasive laser and radar sensor measured not significantly differently from the respiratory rate in anesthetized rats.

The correlation between the reference and the radar data tended to be lower for HR than for RR (Tables 1 and 2). The rate of equivalence within ±5% of the HR data from ECG by RSM ranged from 50% to 100% for radar data, with an average value of 80%. Among the 11 rats, equivalence was ≤80% in three rats, and the equivalence for HR data showed a lower tendency than that in the RR data. As is evident from the ECG waveform, heart activity is more complex than respiration, consisting of P, Q, R, S and T waves that are aligned with cardiac excitation. Therefore, there are about 10 peaks in the radar data frequency band from 0 to 12 Hz after FFT processing (Figure 3(c)). Rats with larger discrepancies between the ref-HR and HR radar data tended to have higher heart rates than the ECG data from RSM (Table 2). If the R-wave amplitude in the radar data (as the source of HR detection) was too small, the frequency analysis might not have identified the HR peaks. Numerous previous studies that measured human HR by radar have used the FFT method for analysis;^14,17,18 however, we recommend that a different method should be considered for animals with a fast HR, such as rodents. It is also possible that measurement accuracy might be improved by adjusting the position of the radar sensor.

The fluctuation in HR and RR after noradrenaline administration was >80% concordance between the reference data and radar data in three animals (Group 1), showed consistent changes with 60–80% concordance in six animals (Group 2) and was severely compromised with concordance less than 60% in both the HR and RR data in two animals (Group 3). It is not clear why this phenomenon occurred, but we speculate that i.p. administration of noradrenaline under anesthesia caused noradrenaline uptake from the mesentery and was accompanied by destabilization of the reaction in some individuals. Alternatively, we cannot rule out the possibility that the i.p. administration of noradrenaline caused a shift in body position or measurement position, or missed abdominal cavity. In addition, compared with the other groups, the HR and RR data from the radar sensor in Group 3 showed greater variability, respectively, and the FFT analysis used in this study might not have detected a clear peak owing to the effects of noradrenaline administration on cardiac activity and respiratory body movement.

The present findings indicated that RR and HR measurements obtained by non-invasive laser and radar sensor in anesthetized rats were comparable to the reference data. In addition, although the distance from the sensor to the test animal was measured and set using a rough scale, there was no problem between the measured data and the reference data. For future application for the RR and the HR measurements by each sensor in anesthetized rats during surgery, it is necessary to investigate the positional relationship between the surgical field and the sensor that does not interfere with data reception. Alternatively, it may be better to allow time during surgery to measure only the HR and the RR as appropriate. In addition, only male rats were used in this study; it will be necessary to evaluate by using females and other animal species in the future. In future work, we aim to investigate how irregularities in measurement after drug administration can be minimized, and to quantify non-invasive vital signs in freely moving experimental animals.