VITAL SIGN EXTRACTION

Vital Sign Extraction

[Name of the Institutiton]

Table of Contents

6.1 Introduction3

6.1.1 Background3

6.1.2 Problem Statement5

6.2 Experiment Data Collection6

6.3 Heart Rate Extraction System8

6.3.1 Overview8

6.3.2 Heart Rate Extraction11

6.3.3 Weight/onset map estimation14

Chapter 6: Vital Sign Extraction

6.1 Introduction

Inside the hospital, medical devices that require attachment to the patient are in widespread use for continuous monitoring. Examples of such devices include ECG electrodes, pulse oximeter finger clips and blood pressure cuffs. Since lives often depend on this medical care, patients routinely accept some level of burden and discomfort in using these devices. There are relatively few instances of noncompliance when it comes to their use within the hospital system.

Outside of the hospital, it seems to become a different scenario. Despite the understanding that routine monitoring of medical parameters is important for longer-term assessment of medical conditions, noncompliance is a prevalent issue. Measurement of physiological parameters can indicate the worsening of disease, the need for re-titration of medication or, hopefully, indicate the management of a chronic condition is under control. However, the typical patient is often reluctant to take the time to attach a device for measurement or leave the device connected for the duration necessary for proper monitoring.

The reasons for this can be attributed to the fact that conventional medical devices can be cumbersome to use and uncomfortable to attach. Or people may either be embarrassed by using an obtrusive device, in denial of their health status, or facing enough day-to-day interruption of their life due to an illness that they avoid any additional interference. There is a clear need for new, unobtrusive medical devices that can overcome these issues while providing clinically relevant information for medical care.

As Eulerian video magnification can reveal color changes caused by blood circulation, it shows the potential that medically interesting information can be extracted from video contactlessly. Given the earlier success on measuring heart rate and respiration rate from videos [12,13], we are motivated to explore how to use Eulerian-based spatio-temporal processing to achieve more accurate heart rate estimation under noise and motions.

In this chapter, we firstly provide physiological background about feasibility of vital sign extraction from videos. Secondly, we describe what kind of data we have acquired for evaluating the accuracy of vital sign estimations and how we collect them. Lastly, we present an early attempt on using Eulerian-based spatio-temporal processing for heart rate extraction. With the localized time series of color values produced by Eulerian method, we can combine them according to their local characteristics to achieve more accurate heart rate estimation.

6.1.1 Background

Vital sign monitoring system nowadays usually requires physical contact with patients while performing the vital sign measurement. For example, the state-of-art technique for measuring cardiac pulse is electrocardiogram (ECG), which requires patients to wear adhesive patches on their chest. Physical contact is usually undesirable because it may cause discomfort and inconvenience, and it is sometimes even infeasible for some patients such as burned patients and pre-mature babies. Therefore, contact-free vital sign extraction system with comparable accuracy is not only preferred but also ...