Chapter 6. Algorithm validation 39
0 1 2 4
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Noise level
Euclidean distance [a.u.]
Euclidean distance
0 1 2 4
30 31 32 33 34 35 36 37 38
Noise level
SNR [dB]
SNR
0 50 100 150 200 250
−0.8
−0.6
−0.4
−0.2 0 0.2 0.4 0.6
Normalized time of one period [n.u.]
Normalized amplitude [n.u.]
Waveform
Correlation based Inverse variance weighted Reference
Figure 6.6: The metrics after processing a synthetic signal with different levels of additive noise and interference. Waveforms are shown at noise level 4
the Euclidean distance metric. Where our method produced good results over all levels, the original method shows bad results at level 2 and 4. In the lower left plot we show the extracted pulse waveforms from both methods and also the reference at interference level 4. By visual inspection we can see that by using our spatial processing the ex-tracted pulse waveform looks like the reference. By using the original spatial processing the extracted pulse waveform is clearly not similar to the reference and therefore not useful for further analysis or applications.
We only show the results up till a level of 4 because above this level both methods give results with errors. The original spatial processing gives a pulse waveform that does not look like the input anymore. Our proposed spatial processing selects the interfering signal as dominant frequency for level 8 and higher so the extracted waveform doesn’t represent the actual BVP signal anymore. Including the levels 8 and higher in the plot could potential lead to wrong interpretations of the results.
40 Chapter 6. Algorithm validation
1 2 3 4 5 6 7 8 9
0 5 10 15 20 25 30 35
Volunteer
SNR [dB]
Inverse variance weighted Trimmed−mean Correlation based
Figure 6.7: The SNR results of the different spatial processing methods on all volun-teers.
Measurement Method I Method II Method III ∆I−III ∆II−III
Volunteer 1 31.27 31.09 32.01 0.74 0.92
Volunteer 2 24.87 25.70 27.08 2.21 1.37
Volunteer 3 27.33 26.20 31.22 3.88 5.02
Volunteer 4 29.87 28.43 31.02 1.15 2.60
Volunteer 5 30.49 31.35 30.82 0.33 -0.52
Volunteer 6 29.93 30.09 30.70 0.77 0.61
Volunteer 7 30.06 30.02 31.18 1.11 1.16
Volunteer 8 26.41 26.82 28.63 2.22 1.81
Volunteer 9 30.70 31.16 32.19 1.49 1.03
Mean (µ) 28.99 28.98 30.54 1.55 1.55
Standard deviation (σ) 2.22 2.25 1.65 1.09 1.55
Table 6.1: The SNR results of the different spatial processing methods. Method I refers to the inverse variance weighted method, II refers to the trimmed-mean method and III refers to our new proposed method based on correlation. All values are in
decibel (dB).
favor of our proposed method. We also calculated the SNR for the 9 recording from our volunteers. This shows an average increase of µ = 1.55 dB with a standard deviation of σ = 1.55 dB compared to the original methods. Figure 6.7 shows the SNR of each of the methods of all volunteers. In 8 out of 9 cases a higher SNR is achieved with our new proposed method, in 1 case the trimmed-mean method performs best. Table 6.1 gives the results of this experiment.
To test whether this improvement is significant we used a paired student’s t-test. This
Chapter 6. Algorithm validation 41
test shows whether the two methods are really different or that the observed differences can be caused by random variation. According to the 95% confidence interval of the difference, our proposed method shows an improvement between [0.71, 2.38] compared to the inverse variance weighted method with a p = 0.003 and an improvement between [0.36, 2.75] compared to the trimmed-mean method with a p = 0.017. So based on these experiments we can state that our proposed method, with a 95% confidence, improves the signal quality in terms of SNR over the other methods. We also looked at the difference between method I and II and performed the paired student’s t-test on these two methods. As p = 0.975 for this test we can say that these two methods perform similar in terms of SNR.
Chapter 7
Conclusions and future work
Whereas photoplethysmography (PPG) is used for many years in the medical area, re-mote PPG (rPPG) is relative new. Recently, the research on rPPG increased in popular-ity, where most research is about improving motion robustness. The pulse waveform of a single period of the PPG signal is used in various medical applications. We think that it would be meaningful to be able to extract the pulse waveform with a camera. That is why we proposed a method to extract the pulse waveform from a rPPG signal. To the best of our knowledge we are the first who try to extract this pulse waveform from rPPG and so we have no golden standard reference to compare our extracted waveforms with.
We tested our method with generated data and its robustness to noise and interfering signals. These experiments show that our method is capable of extracting the pulse waveform under various conditions. Next to extracting the pulse waveform we also used this extracted waveform to improve the rPPG signal. We proposed a new method for the spatial processing of the rPPG signal where the pulse waveform is used as a selection criterion. This is done with a feedback mechanism where the pulse waveform is updated during the processing. This new proposed spatial processing showed good results on the generated data with interfering signals and noise. When we run the new proposed method on recordings of the hands of 9 volunteers we see an improvement in 8 out of 9 cases in terms of SNR. The average improvement in terms of SNR is µ = 1.55/1.55 dB compared to inverse variance weighted / trimmed-mean method. The spread of the improvement is σ = 1.09/1.55 dB and the results of a paired student’s t-test show a statistical significant improvement with p = 0.003/0.017.
43
44 Chapter 7. Conclusions
7.1 Future work
In the experiments for comparing the SNR we used a peak detector in the spectral do-main to determine the heart-rate and used this peak as signal in the SNR calculation.
A good approach would be to simultaneously use an ECG and rPPG signal. Whereby you could use the detected heart-rate from the ECG to calculate the SNR of the rPPG signal. Also several parameters in both the temporal and spatial processing are chosen arbitrary (nevertheless performing well) and can probably be optimized by using some sort of cross validation. Furthermore extra research should be performed on the actual meaning and possible use of the extracted pulse waveform from rPPG. A good exper-iment would be to test whether the features that are contained in the pulse waveform from PPG are also available in the pulse waveform of a rPPG signal.
Bibliography
[1] A. B. Hertzman, “Photoelectric plethysmography of the fingers and toes in man,” in Proceedings of the Society for Experimental Biology and Medicine.
Society for Experimental Biology and Medicine (New York, NY), vol. 37, no. 3. Royal Society of Medicine, 1937, pp. 529–534. [Online]. Available:
http://ebm.sagepub.com/content/37/3/529.short
[2] P. Sahindrakar, G. de Haan, and I. Kirenko, “Improving motion robustness of contact-less monitoring of heart rate using video analysis,” Master’s thesis, 2011. [Online]. Available: http://alexandria.tue.nl/extra1/afstversl/wsk-i/
sahindrakar2011.pdf
[3] J. Allen, “Photoplethysmography and its application in clinical physiological measurement,” Physiological measurement, vol. 28, no. 3, p. R1, 2007. [Online].
Available: http://iopscience.iop.org/0967-3334/28/3/R01
[4] Y. S. of Medicine Patient Care. (2013, July) Anatomy of the skin. [Online].
Available: http://www.yalemedicalgroup.org/stw/Page.asp?PageID=STW023939 [5] B. Medical. (2013, July) Bluetooth pulse oximeter with software. [Online].
Avail-able: http://bio-medical.com/products/bluetooth-pulse-oximeter-with-software.
html
[6] L. Gailite, J. Spigulis, and A. Lihachev, “Multilaser photoplethysmography technique,” Lasers in Medical Science, vol. 23, no. 2, pp. 189–193, 2008. [Online].
Available: http://link.springer.com/article/10.1007/s10103-007-0471-9
[7] J. A. Crowe and D. Damianou, “The wavelength dependence of the photoplethysmogram and its implication to pulse oximetry,” in Engineering in Medicine and Biology Society, 1992 14th Annual International Conference of the IEEE, vol. 6. IEEE, 1992, pp. 2423–2424. [Online]. Available:
http://ieeexplore.ieee.org/xpls/abs all.jsp?arnumber=5761522
45
46 Bibliography
[8] A. Jonsson, “Pressure sore etiology-highlighted with optical measurements of the blood flow,” Ph.D. dissertation, University of G¨avle, 2006. [Online]. Available:
http://hig.diva-portal.org/smash/record.jsf?pid=diva2:120505
[9] P. Y. Cheang and P. R. Smith, “An overview of non-contact photoplethysmogra-phy,” Electronic systems and control division research, pp. 57–59, 2003.
[10] N. S. Trivedi, A. F. Ghouri, N. K. Shah, E. Lai, and S. J. Barker, “Effects of motion, ambient light, and hypoperfusion on pulse oximeter function,” Journal of clinical anesthesia, vol. 9, no. 3, pp. 179–183, 1997. [Online]. Available:
http://www.sciencedirect.com/science/article/pii/S0952818097000391
[11] S. Hu, J. Zheng, V. Chouliaras, and R. Summers, “Feasibility of imaging photoplethysmography,” in BioMedical Engineering and Informatics, 2008. BMEI 2008. International Conference on, vol. 2. IEEE, 2008, pp. 72–75. [Online].
Available: http://ieeexplore.ieee.org/xpls/abs all.jsp?arnumber=4549138
[12] F. Wieringa, F. Mastik, and A. Van der Steen, “Contactless multiple wavelength photoplethysmographic imaging: a first step toward spo2 camera technology,”
Annals of biomedical engineering, vol. 33, no. 8, pp. 1034–1041, 2005. [Online].
Available: http://link.springer.com/article/10.1007/s10439-005-5763-2
[13] W. Verkruysse, L. O. Svaasand, and J. S. Nelson, “Remote plethysmographic imaging using ambient light,” Optics express, vol. 16, no. 26, pp. 21 434–
21 445, 2008. [Online]. Available: http://www.opticsinfobase.org/abstract.cfm?
URI=oe-16-26-21434-1
[14] M.-Z. Poh, D. J. McDuff, and R. W. Picard, “Non-contact, automated cardiac pulse measurements using video imaging and blind source separation,” 2010.
[Online]. Available: http://dspace.mit.edu/handle/1721.1/66243
[15] M. Lewandowska, J. Ruminski, T. Kocejko, and J. Nowak, “Measuring pulse rate with a webcama non-contact method for evaluating cardiac activity,” in Computer Science and Information Systems (FedCSIS), 2011 Federated Conference on. IEEE, 2011, pp. 405–410. [Online]. Available:
http://ieeexplore.ieee.org/xpls/abs all.jsp?arnumber=6078233
[16] Y. Sun, S. Hu, V. Azorin-Peris, S. Greenwald, J. Chambers, and Y. Zhu, “Motion-compensated noncontact imaging photoplethysmography to monitor cardiorespi-ratory status during exercise,” Journal of Biomedical Optics, vol. 16, no. 7, pp.
077 010–077 010, 2011.