Artificial intelligence (AI) fields, including machine learning (ML) and deep learning (DL), have demonstrated promising outcomes in the risk stratification of cardiovascular disease (CVD) or coronary artery disease. Due to its high performance and suboptimal clinical results, bias in ML systems is of tremendous interest. Understanding the risk of bias in AI research for CVD risk prediction is the primary goal. A total of 23 AI experiments on ultrasound technology were chosen as the top ones using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) methodology. With the use of ultrasonic methods, the ML and DL structures are thoroughly examined. By grading the solution together with an assessment of three bias bins and categorizing the 23 AI research into these three categories (low-bias, moderate-bias, and high-bias), the study showed how effective a technique for estimating bias is. By proving that DL-based research is less biased than that conducted using other AI systems, we validated the hypothesis in this way. In addition to the aforementioned, the study shed light on the bias in AI that is based on ranking methods. Furthermore, we showed how the AI framework incorporated variables like office-based biomarkers, laboratory-based biomarkers, and carotid ultrasound imaging phenotype.