Explainable Machine Learning-Based SpO2 Estimation Using Photoplethysmograms Measured from a Neck Wearable

11 Pages Posted: 14 Feb 2023

See all articles by Yuhao Zhong

Yuhao Zhong

Texas A&M University

Ashish Jatav

Texas A&M University

Kahkashan Afrin

Texas A&M University

Tejaswini Shivaram

Texas A&M University

Satish Bukkapatnam

Texas A&M University

Abstract

Conventional transmittance-based photoplethysmogram (PPG) sensors for measuring blood oxygen saturation (SpO2) are limited by standard measuring sites like fingers. Reflectance-based sensors provide more flexible options of measuring sites, but they are mostly limited by accuracy, especially when applied to different subjects, due to the diverse human characteristics (skin colors, hair density, etc.) and usage conditions of different sensor settings. This study addresses the estimation of SpO2 at non-standard measuring sites employing reflectance-based sensors. It proposes an automated construction of subject inclusion-exclusion criteria for SpO2 measuring devices, using a combination of unsupervised clustering, supervised regression, and model explanations. This is perhaps the first adaptation of SHAP to explain the clusters gleaned from unsupervised learning methods. As a case study of wellness application, a patent pending pillow-based wearable device was developed to collect reflectance PPGs from both brachiocephalic and carotid arteries around the neck. The experiment was conducted on 33 subjects, each under totally 80 different sensor settings. The proposed approach tackled the human and the device variations as well as the heterogeneous mapping between signals and SpO2 values. It identified the effective device settings and the characteristics of their applicable subject groups (i.e., subject inclusion-exclusion criteria) that resulted in a maximum absolute error (MaAE) [[EQUATION]] 1.7% and a root mean squared error (RMSE) [[EQUATION]] 0.9% for estimating SpO2 over the range of 90% to 100%. Thus, the approach works to detect the propensity of SpO2 falling into an unhealthy level (< 95%).

Note:
Funding declaration: This work was supported in part by National Science Foundation under Grant NSF-PFI-AIR-TT 1543226.

Conflict of Interests: No competing interests in the research conducted or reported in the manuscript.

Ethical Approval: IRB approved experiments at Texas A&M University (protocol IRB2015- 0623F, approved in 2017 by Texas A&M Human Research Protection Program).

Keywords: Explainable machine learning, neck reflectance Photoplethysmogram (PPG), subject heterogeneity, subject inclusion-exclusion criteria, SpO2 estimation

Suggested Citation

Zhong, Yuhao and Jatav, Ashish and Afrin, Kahkashan and Shivaram, Tejaswini and Bukkapatnam, Satish, Explainable Machine Learning-Based SpO2 Estimation Using Photoplethysmograms Measured from a Neck Wearable. Available at SSRN: https://ssrn.com/abstract=4333613 or http://dx.doi.org/10.2139/ssrn.4333613

Yuhao Zhong

Texas A&M University ( email )

Ashish Jatav

Texas A&M University ( email )

Kahkashan Afrin

Texas A&M University ( email )

Tejaswini Shivaram

Texas A&M University ( email )

Satish Bukkapatnam (Contact Author)

Texas A&M University ( email )

Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Downloads
61
Abstract Views
205
Rank
652,764
PlumX Metrics