Exploration of the Conditions for Occurrence of Photoplethysmographic Signal Inversion above the Dorsalis Pedis Artery

Figures

• 

  Figure A1. Images of measurement setup and 3D printed strap. The PPG sensor is located within the rectangular house and can be moved laterally along the mm scale.

• 

  Figure A2. Flexiforce Tekscan A201 force sensor and calibration curve (red).

• 

  Figure A3. Example of Triangulation algorithm for angle estimation for an inverted signal. Angle estimates are acute and agree with the located inverted pulse candidates from the symmetry algorithm. The inversion estimate was −0.9487 for this signal. The Y-axis is post SG/HP filtering and is negated. Crosses are systolic end points, while circles are adjacent troughs located by the algorithm.

• 

  Figure A4. Example of Triangulation algorithm for angle estimation for a non-inverted signal. Angle estimates are acute and agree with the located non-inverted pulse candidates from the symmetry algorithm, Inversion estimate was 0.96 for this signal. The Y-axis is post SG/HP filtering and is negated. Crosses are systolic end points, while circles are adjacent peaks located by the algorithm.

• 

  Figure A5. A 90 s recording example where the triangulation algorithm disagrees with the symmetry algorithm. X-marks refers to pulse candidates located by the symmetry algorithm (small peaks). Triangulation was stopped for easier interpretation. The symmetry algorithm will consider this a non-inverted signal; however, this sample is distorted, causing the initial angle estimates to be obtuse and the sample is thus flagged for investigation. The inversion estimate is 0.80 for this signal. The Y-axis is post SG/HP filtering and is negated. Crosses are systolic end points, while circles are adjacent peaks located by the algorithm.

• 

  Figure A6. Histogram of all recordings mean degree of inversion µInversion in logarithmic scale. Noisy or corrupted PPG signals are excluded. With a threshold of µInversion < 0.5 (equal to DOI > 0.25), around 12.65% of all recordings were considered inverted.

• 

  Figure 1. Illustration of the PPG-based measurement system used in the study. Colors on the PC (right) represents the 4 wavelengths (ir, red, green and blue) emitted during acquisition.

• 

  Figure 2. (A) Inversion algorithm flowchart and logic representation. (B) Systole/diastole asym-metry check. Blue and red shaded areas indicate the systolic and part of the diastolic phase in a PPG signal. In this case, both the non-inverted and the inverted signals fulfill the systole/diastole symmetry condition since their deflections are steepest in the previous 240 samples. (C) Triangulation with adjacent local peaks for pulse candidate sharpness evaluation. Sharp candidates are thought to represent systolic deflections. The angle theta of the constructed triangle is calculated by the law of cosine [13].

• 

  Figure 3. Examples (a–d) of 10 s segments of PPG recordings from different participants demonstrating signal inversion at different wavelengths. The line colors correspond to the different LEDS, where IR = black, and the vertical axes represent the negative Z score standardization of the sensor analog to digital converted value. (e,f) displays the contrast in signal shape between PPG from different LEDs measured simultaneously, caused by inversion.

• 

  Figure 4. Percentwise occurrence of PPG signal inversion versus sensor position relative to the DPA for the four LEDs dependent on contact pressure (a–d), averaged over pressure levels to visualize LED dependency (e), averaged over LEDs to visualize pressure dependency (f). Averaged occurrence over all pressures and LEDs for each subject is shown in (g), and the grand mean (averaged over LEDs, pressures, and all participants) with the confidence interval of the mean is shown in (h), with the addition of the median (grey) for comparison. The positive position on the x-axis corresponds to sensor positioning in the medial direction. Color between subject (g).