After selecting the data to be analyzed, the code 1 (see Appendix, taken from http://noc.ac.uk/using-science/crosswavelet-wavelet-coherence) was applied, which identified the maximum amplitude peaks for each wave. The next step was to create an artificial sine signal to obtain an expression of conversion of periods into frequency. Thus, a sine equation with known frequencies was created in code 2 (line b). Later a trend line was created and translated by the following equation y= 49.416e-0.057x; R² = 0.9999.

After the creation of the artificial sine signal, the Wavelet transform for the PPG signal was calculated. Using code 3, the Wavelet transform decomposes the signal into a periodogram resuming all different frequency components (Figure 2). With the minimum and maximum period of each wave obtained in the periodogram, we were able to obtain, minimum and maximum frequencies for each wave using the described equation. Replacing the value of x in the equation with the period value and then multiplying by 60, we obtained the activity value per minute of each wave in the periodogram.

The protocols mentioned above were applied to each animal in all records obtained before and after administration of medication.

Results

Figure 3 shows an original PPG signal and the respective decomposition by the Wavelet transform analysis.

The maximum and minimum frequencies of each wave of the different animals analyzed before medication were, respectively: wave 1 - 3.11 and 9.43 Hz; wave 2 - 1.18 and 3.00 Hz; wave 3 - 0.57 and 0.97 Hz; wave 4 - 0.23 and 0.44 Hz; wave 5 - 0.12 and 0.21 Hz; wave 6 (registered only in three animals) - 0.09 and 0.14 Hz and wave 7 (registered only in two animals) - 0.04 and 0.09 Hz. The PPG wave-frequencies obtained after administration of medication were: wave 1 -between 1 - 2.04 and 7.55 Hz; wave 2 - 0.86 and 2.04 Hz; wave 3 - 0.28 and 0.71 Hz; wave 4 - 0.12 and 0.26 Hz; wave 5 - 0.07 and 0.12 Hz and wave 6 (only one animal) between -0.08 and 0.10 Hz.

Discussion

This study illustrated, for the first time, the use of PPG as a monitoring technology for heart and respiratory rates in dogs in a surgical setting in veterinary medical practice, as well as the characterization of the different frequency waveforms present in the PPG signal.

Seven frequency waves were identified in these signals. Based on previous reports conducted in humans, rats and mice (6,9), taking also into account the normal heart and respiratory rates in dogs, we assumed that the different frequency waveforms were related to (in descending order of frequency averages): (i) the harmonic of cardiac activity (1st wave, between 3.11 and 9.43 Hz), (ii) cardiac activity (2nd wave, between 1.18 and 3.00 Hz), (iii) respiratory activity (3rd and 4th waves, between 0.57 and 0.97 Hz and 0.23 and 0.44 Hz, respectively), (iv) afferent vagal activity (5th and 6th waves, between 0.12 and 0.21 Hz and 0.09 and 0.14 Hz, respectively) and (v) to sympathetic tone groundings (7th wave, between 0.04 and 0.09 Hz).

After the administration of the sedative medication, a decrease in the values for cardiac and respiratory activity and a decrease in the mean frequency of each wave were clear within the PPG signal. In medical practice, the utility of PPG in determining cardiac and respiratory rates are widely recognized (7,10). In veterinary medicine, as far as our current knowledge goes, only one study has associated PPG with respiratory rate (11). Differences found regarding respiratory activity before and after administration of medication were not statistically significant, although a slight decrease in maximum respiratory activity and a (larger) decrease in minimum respiratory activity were detected. However, the purpose of this exploratory study was to test the applicability of this technology in the present setting, and in our opinion this goal was achieved.

We recognize the limitations involved in our study, including the small dimension of the sample used, the inter-individual variability, and the difficulty in selecting the ideal anatomical site for placing the sensor. Clearly, additional prospective animal studies are needed to validate the PPG technique in the present setting. Nevertheless, taking into account these results and our experience, for future studies we foresee (i) to include to a larger and more homogeneous population in terms of breed, age or live weight of the animals, in a way to reduce experimental bias, (ii) to use different anatomical sites, such as the ear or the extremities of the thoracic and pelvic limbs, (iii) to increase the duration of data collection to limit associated with animal movement and (iv) to conduct a comparative study of heart rate and respiratory rate obtained using other referenced techniques / methods in addition to PPG.

This study demonstrated the potential utility of such a wearable instrument in the clinical environment. The possibility of obtain biometrical data rapidly and easily, including cardiac and respiratory rates, from the photoplethysmography signal can help, in a simple way, to manage several conditions where the monitoring of these parameters gains a crucial importance to manage animals with chronic cardiopulmonary conditions, geriatric patients and convalescence animals upon surgery.

Conclusion

This pilot study demonstrated the relevance of reflection PPG, a non-invasive technique, to measure the heart and respiratory rates in dogs, and its utility in the clinical and ambulatory environment in veterinary medicine. It was also possible to confirm that the tail seems to be an appropriate anatomical area in dogs to collect these parameters.

Authors Contributions Statement 

JR, LMR, conceptualization and study design; RA, LL experimental; RA, CR, data curation; RA, JR, LMR, drafting, editing and reviewing; JR, LMR, supervision and final writing.

Acknowledgements

This work was funded by national funds through FCT - Foundation for Science and Technology, I.P., under the Scientific Employment Stimulus - Institutional Call - CEECINS/00127/2018 and supported by the project UIDB/CVT/00772/2020 and associated laboratory AL4AnimalS.

Conflict of Interests 

Editors involved in this manuscripts’ authorship had no participation in the review or decision process. All authors have stated that there are no financial and/or personal relationships that could represent a potential conflict of interest..

Appendix 

Code 1

(http://noc.ac.uk/using-science/crosswavelet-wavelet-coherence)

a. t=1:length(ppg);

b. figure(1);

c. plot(t,ppg)

d. title('Raw Signal')

e. xlabel('Samples');

f. ylabel('Voltage(mV)')

g. legend('Noisy PPG Signal')

h. grid on

i. %Find Peaks

j. [~,locs_Peak] = findpeaks(ppg,'MinPeakHeight',a,...'MinPeakDistance',b);

k. figure(3)

l. hold on

m. plot(t,ppg);

n. plot(locs_Peak,ppg(locs_Peak2),'rv','MarkerFaceColor','r');

o. axis([0 5000 440 640]); grid on;

p. legend('PPG Signal','Peak-waves');

q. xlabel('Samples'); ylabel('Voltage(mV)')

r. title('Peak in PPG Signal')

Note: At this point the letter 'a' has been replaced by the minimum amplitude of the waves of the signal 'ppg' and the letter 'b' by the minimum distance between the maximum amplitude peaks of each wave.

Code 2

a. for i=1:d, x(i)=i;

b.seno(i)=sin(2*3.14*4/100*i)+sin(2*3.14*2/100*i)+sin(2*3.14*1/100*i)+sin(2*3.14*0.5/100*i ) + sin(2*3.14*0.25/100*i)+sin(2*3.14*0.125/100*i)+sin(2*3.14*0.0625/100*i)+sin(2*3.14*0.03 125/100*i);

c. end;

d. [waveseno,periodseno,scaleseno,coiseno,sig95seno]=wt(seno);

e. amplitudeseno=abs(waveseno);

f. for i=1:c mediaondasseno(i)=mean(amplitudeseno(i,1:d)); end

g. plot(mediaondasseno);

Note: At this point, the letter “d” represented the total number of points for analysis that were chosen in the original sign (on the abscissa axis of the original sign) and the letter “c” represented the number obtained by the equation.

Code 3

a. [waveppg,periodppg,scaleppg,coippg,sig95ppg]=wt(ppg);

b. [wave,period,scale,coi,sig95]=wt(ppg,'MakeFigure',1);

c. amplitudeppg=abs(waveppg);

d. for i=1:c mediaondasppg(i)=mean(amplitudeppg(i,1:d)); end

e. figure(2), plot(mediaondasppg).

References 

1. Griffies, J. D., Zutty, J., Sarzen, M., & Soorholtz, S. (2018). Wearable sensor shown to specifically quantify pruritic behaviors in dogs. BMC Veterinary Research, 14(1), 1–10. https://doi.org/10.1186/s12917-018-1428-x.

2. Dorlas, J. C., & Nijboer, J. A. (1985). Photo-electric plethysmography as a monitoring device in anaesthesia. British Journal of Anaesthesia, 57, 524–530. https://doi.org/10.1093/bja/57.5.524.

3. Allen, J. (2007). Photoplethysmography and its application in clinical physiological measurement. Physiological Measurement, 28(3), R1–R39. https://doi.org/10.1088/0967-3334/28/3/R01.

4. Sahni, R. (2012). Noninvasive monitoring by photoplethysmography. Clinics in Perinatology, 39(3), 573–583. https://doi.org/10.1016/j.clp.2012.06.012.

5. Schultz-Ehrenburg, U., & Blazek, V. (2001). Value of quantitative photoplethysmography for functional vascular diagnostics: current status and prospects. Skin Pharmacology and Applied Skin Physiology, 14(5), 316–323. https://doi.org/10.1159/000056362.

6. Nilsson, L. M. (2013). Respiration signals from photoplethysmography. Anesthesia and Analgesia, 117(4), 859–865. https://doi.org/10.1213/ANE.0b013e31828098b2.

7. Leonard, P., Beattie, T., Addison, P. S., & Watson, J. N. (2003). Standard pulse oximeters can be used to monitor respiratory rate. Emergency Medicine Journal, 20, 524–525. https://doi.org/10.1136/emj.20.6.524.

8. Houston D. M. (2000). Clinical exam in dogs and cats. In O. M. Radostits, D. M. Houston, & I. G. Mayhew (Eds.), Veterinary Clinical Examination & Diagnosis (1st ed., pp. 98-101). W. B. Saunders Co. Ltd.

9. Silva, H., Roux, É., Gadeau, A. P., & Rodrigues, L. M. (2020). Wavelet analysis of microcirculatory flowmotion reveals cardiovascular regulatory mechanisms-data from a beta-blocker. Applied Sciences, 10(11), 4000. https://doi.org/10.3390/app10114000

10. Karlen, W., Garde, A., Myers, D., Scheffer, C., Ansermino, J. M., & Dumont, G. A. (2015). Estimation of respiratory rate from photoplethysmographic imaging videos compared to pulse oximetry. IEEE Journal of Biomedical and Health Informatics, 19(4), 1331–1338. https://doi.org/ 10.1109/JBHI.2015.2429746.

11. Nabutovsky, Y., Pavek, T., & Turcott, R. (2012). Chronic performance of a subcutaneous hemodynamic sensor. Pacing Clinical Electrophysiology, 35(8), 919–26. https://doi.org/10.1111/j.1540-8159.2012.034