Cardiovascular Disease Risk Assessment using Carotid Ultrasound Image Phenotypes with Machine Learning

Abstract

Abstract newlineCardiovascular disease (CVD) leads to the annual deaths of approximately 17.9 million newlinepeople in the world. Besides this, the death rate is also high in low and middle-income newlinecountries such as India, China, and other developing nations, which contributes 75% of newlinethe total deaths caused due to CVD. In South Asian countries especially in India, CVD is newlinean important cause of mortality. According to a recent study, Indian people are affected newlineearlier than European countries. It is estimated that the total deaths due to coronary heart newlinedisease by 2030, per 100,000 will reach up to 3,070 in India, which is higher than in newlineChina and Brazil. The estimated global expenditure for CVD in 2010 was $863 billion, newlinewhich is expected to reach up to $1044 billion by 2030. It is imperative that mortalitywise newlineand economically, CVD is a big global challenge. Therefore, there is an alarming newlineneed to prevent CVD by developing tools that are accurate and affordable for both newlinephysicians and patients. newlineAt present, conventional risk prediction models are being used for CVD risk newlineassessment. However, often conventional risk prediction models cannot explain the newlineelevated CVD risk in patients. This is because such models are based on only traditional newlinerisk factors that do not capture the variations in atherosclerotic plaque. To provide better newlineCVD risk assessment, it is important to use low-cost imaging modalities like carotid newlineultrasound (CUS), which can accurately capture variations in the atherosclerotic plaque. newlineSome popular carotid ultrasound image-based phenotypes (CUSIP) such as carotid newlineintima-media thickness (cIMT) and total carotid plaque area are considered the markers newlineof several cardiovascular events such as coronary artery disease, acute coronary newlinesyndrome, and myocardial infarction. Therefore, for accurate CVD risk assessment, there newlineis a need to utilize the effectiveness of CUSIP in the risk prediction models. newlineWith the above background, the main objective of the presented work is to develop newlinethe CVD risk assessment tools or systems that