Your smartphone could act as a pulse-oximeter and as a single-lead ECG

Figures

• 

  Figure 1. A quick graphical summary of this work.

• 

  Figure 2. Highlights of the pre-processing done during the training and testing phases of: (1) vitals estimation, and (2) vPPG-to-ECG translation.

• 

  Figure 3. Detrending and denoising: two key steps in the preprocessing of a PPG signal.

• 

  Figure 4. The CLIP neural network model for one-shot estimation of Vitals.

• 

  Figure 5. The model architecture details of encoders in Vitals-CLIP. The inner architecture of the PPG encoder’s embedding layer and linear projection layer is given in Table 2.

• 

  Figure 6. Training and validation loss of fine-tunned (a) CNN-Net, (b) MT-Net, (c) ViT-Net, (d) ResNet50 models using K20-vPPG and BIDMC datasets.

• 

  Figure 7. Ground truth caption and top three matching captions predicted by Vitals-CLIP for the PPG signals from (a) PulseDB dataset, (b) BIDMC dataset, (c) K20-vPPG dataset. Here GT stands for ground truth, while C1, C2, C3 are top three matching captions for which the dot product of text and PPG embedding are highest, second highest and third highest respectively. HR is measured in beats/min, RR is measured in breaths/min, SpO2 is measured in percentage, SBP and DBP are measured in mmHg.

• 

  Figure 8. Screenshot of results panel of our custom Android app developed using the proposed Vitals-Net framework.

• 

  Figure 9. Impact of window size on vitals estimation (x-axis represents window size, the y-axis represents MAE and SAE).

• 

  Figure 10. DCT coefficients of the PPG signal and the single-lead ECG signal (one cardiac cycle only). The proposed P2E-Net framework learns the regression between the two sets of DCT coefficients.

• 

  Figure 11. P2E-Net model architecture in which each fully connected (FC) layer uses ’tanh’ activation and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$C_P = 150$$\end{document}CP=150 and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$C_E = 150$$\end{document}CE=150.

• 

  Figure 12. Training and validation loss of P2E-Net.