ABSTRACT
This chapter discusses the potential use of machine learning-assisted assessment of arterial stiffness (AS), particularly carotid-to-femoral pulse wave velocity (cf-PWV), which is considered the gold-standard measurement of AS. The current method of measuring cf-PWV is considered challenging for clinicians and patients due to its operator dependency and potential inaccuracies. To overcome these limitations, different machine-learning pipelines were trained and tested using features extracted from peripheral pulse waveforms. Three modalities were investigated, including time domain-based features, frequency domain-based features, and semi-classical signal analysis-based features. Results show that these proposed features and algorithms have the potential to estimate cf-PWV and assess AS non-invasively, indicating the feasibility of using machine learning approaches as smart surrogate measures of vascular indicators and potential predictors for cardiovascular diseases.
