A Broader Look: Camera-Based Vital Sign Estimation across the Spectrum

Area of Science:

• Physiological monitoring
• Medical imaging
• Biomedical engineering

Background:

• Camera-based vital sign estimation offers contactless assessment of physiological parameters.
• Significant advancements occurred in the 1930s, 1980s, and 2000s, with accelerating progress.
• This survey focuses on recent developments, commonalities, limitations, and novel approaches.

Purpose of the Study:

• To provide an overview of recent research in camera-based vital sign estimation.
• To identify common features and shortcomings in the existing literature.
• To highlight unique or outlier studies in the field.

Main Methods:

• Comprehensive literature research and quantitative analysis of papers published between 2016 and 2018.
• Extracted data included subject numbers, study populations (healthy vs. pathological), datasets, settings (lab vs. real-world), camera types, machine learning usage, and spectral properties.
• Qualitative analysis covered illumination conditions and algorithmic advancements.

Main Results:

• 116 papers published since 2016; 59% studied 20 or fewer subjects.
• Most studies (n=100) focused on healthy subjects, with limited real-world applications (27 papers) and few using public datasets (4 datasets in 10 papers).
• Red-green-blue (RGB) data was most common (n=61), with limited use of consumer-grade equipment (n=39) and machine learning (n=24).

Conclusions:

• The field of camera-based vital sign estimation is broad, with diverse approaches but common shortcomings.
• Lack of standardization hinders comparability of techniques and performance evaluation.
• Sharing algorithms and datasets is crucial for advancing the field, particularly for adopting newer methods like deep learning.