Investigation of Optimal Light Source Wavelength for Cuffless Blood Pressure Estimation Using a Single Photoplethysmography Sensor

Abstract

Routine blood pressure measurement is important for the early detection of various diseases. Recently, cuffless blood pressure estimation methods that do not require cuff pressurization have attracted attention. In this study, we investigated the effect of the light source wavelength on the accuracy of blood pressure estimation using only two physiological indices that can be calculated with photoplethysmography alone, namely, heart rate and modified normalized pulse volume. Using a newly developed photoplethysmography sensor that can simultaneously measure photoplethysmograms at four wavelengths, we evaluated its estimation accuracy for systolic blood pressure, diastolic blood pressure, and mean arterial pressure against a standard cuff sphygmomanometer. Mental stress tasks were used to alter the blood pressure of 14 participants, and multiple linear regression analysis showed the best light sources to be near-infrared for systolic blood pressure and blue for both diastolic blood pressure and mean arterial pressure. The importance of the light source wavelength for the photoplethysmogram in cuffless blood pressure estimation was clarified.

Keywords: blood pressure; heart rate; modified normalized pulse volume; optical wavelength; photoplethysmography.

Figure 1

Developed multiwavelength photoplethysmogram (PPG) sensor: (a) layout of light-emitting diodes (LEDs) and photodiode (PD). B LED = blue LED (470 nm), G LED = green LED (528 nm), R LED = red LED (620 nm), and NIR LED = near-infrared LED (870 nm); (b) actual PPG sensor.

Figure 2

Example of simultaneous recordings of blue, green, red, and near-infrared light photoplethysmograms (PPGs) measured using a DC amp: B = blue LED PPG, G = green LED PPG, R = red LED PPG, and NIR = near-infrared LED PPG.

Figure 3

Schematic of photoplethysmogram (PPG) waveform feature values: AC is the amplitude of PPG, DC is the mean value of PPG, T is the peak interval time, HR = heart rate, and mNPV = modified normalized pulse volume [31].

Figure 4

Experimental protocol. The experiment consisted of 2 periods: baseline (BL) and mental arithmetic (MA). Each period lasted 180 s. A total of 2 BP measurements and a PPG measurement (28 s) were taken for each period. The BL period measured the resting condition, and the MA period measured the stress condition induced via mental arithmetic.

Figure 5

Flowchart of BP estimation. BP = blood pressure (SBP/DBP/MAP), HR = heart rate, and mNPV = modified normalized pulse volume.

Figure 6

Accuracy of systolic blood pressure (SBP) estimation at each wavelength. Scatterplots of SBP estimated using proposed method and SBP measured with a cuff sphygmomanometer (N = 42). From left to right, the light source wavelengths are blue (B), green (G), red (R), and near-infrared (NIR). The solid line on each scatterplot is the regression line with geometric mean regression.

Figure 7

Accuracy of diastolic blood pressure (DBP) estimation at each wavelength. Scatterplots of DBP estimated using proposed method and DBP measured with a cuff sphygmomanometer (N = 42). From left to right, the light source wavelengths are blue (B), green (G), red (R), and near-infrared (NIR). The solid line on each scatterplot is the regression line with geometric mean regression.

Figure 8

Accuracy of mean arterial pressure (MAP) estimation at each wavelength. Scatterplots of MAP estimated using proposed method and MAP measured with a cuff sphygmomanometer (N = 42). From left to right, the light source wavelengths are blue (B), green (G), red (R), and near-infrared (NIR). The solid line on each scatterplot is the regression line with geometric mean regression.

Figure 8

Accuracy of mean arterial pressure (MAP) estimation at each wavelength. Scatterplots of MAP estimated using proposed method and MAP measured with a cuff sphygmomanometer (N = 42). From left to right, the light source wavelengths are blue (B), green (G), red (R), and near-infrared (NIR). The solid line on each scatterplot is the regression line with geometric mean regression.

Figure 9

Bland–Altman plots of estimated systolic blood pressure (SBP) against those measured with a cuff sphygmomanometer (N = 42). From left to right, the light source is blue (B), green (G), red (R), and near-infrared (NIR). The solid line and the dashed lines on each plot represent fixed bias (M) and M ± 1.96 standard deviation range, respectively.

Figure 10

Bland–Altman plots of estimated diastolic blood pressure (DBP) against those measured with a cuff sphygmomanometer (N = 42). From left to right, the light source is blue (B), green (G), red (R), and near-infrared (NIR). The solid line and the dashed lines on each plot represent fixed bias (M) and M ± 1.96 standard deviation range, respectively.

Figure 11

Bland–Altman plots of estimated mean arterial pressure (MAP) against those measured with a cuff sphygmomanometer (N = 42). From left to right, the light source is blue (B), green (G), red (R), and near-infrared (NIR). The solid line and the dashed lines on each plot represent fixed bias (M) and M ± 1.96 standard deviation range, respectively.

Figure 11

Bland–Altman plots of estimated mean arterial pressure (MAP) against those measured with a cuff sphygmomanometer (N = 42). From left to right, the light source is blue (B), green (G), red (R), and near-infrared (NIR). The solid line and the dashed lines on each plot represent fixed bias (M) and M ± 1.96 standard deviation range, respectively.