Recent advances in artificial intelligence for cardiac imaging

Abstract

In recent years, major advancements have been made in Artificial Intelligence (AI), which are rising in sophistication, complexity and autonomy. A continually veritable and explosive data growth with a rapid iteration of the innovation of computer hardware provides a turbo boost for AI development. AI is an overarching term in computer science and an umbrella concept that provides means to imbue machines with human-like “general” intelligence with minimal human intervention. It encompasses a wide variety of research studies, from computer vision, natural language processing, and robotics to medical data analysis, including both theoretical and practical development of machine learning and newly rebranded and prosperous deep learning. Cardiovascular disorders are the leading cause of death and morbidity worldwide. AI approaches, in particular, deep learning, are especially suited to solving the problems of scalability and high data dimensionality and are showing great potential in the research of cardiac imaging. Recent advances include automated coronary artery calcium score analysis from non-contrast cardiac-gated CT scans (Zhang et al., 2021a, b), multitask learning for estimating multitype cardiac indices in MRI and CT (Yu et al., 2021), diagnosis of chronic myocardial infarction on non-enhanced cardiac Cine MRI (Zhang et al., 2019), direct quantification of coronary artery stenosis (Zhang et al., 2020), diagnosis of coronary cardiac disease using intravascular ultrasound (Cao et al., 2020), segmentation and quantification of scars of atrial fibrillation (Li et al., 2020a; Yang et al., 2020), post-processing and segmentation for in vivo cardiac diffusion tensor MR (Ferreira et al., 2020), echocardiographic sequences segmentation (Li et al., 2020b), and multimodal whole heart segmentation (Shi et al., 2018; Zhuang et al., 2019). In addition to computer-aided diagnosis, anatomical and lesion segmentation, recent studies also include investigations in fast cardiac imaging (Schlemper et al., 2018; Seitzer et al., 2018). In this Special Issue, we have carefully gleaned the state-of-the-art AI powered cardiac imaging research studies, which have addressed outstanding methodological issues as the area transitions from pilot studies to widespread clinical deployment, from one-dimensional global descriptors to high-resolution patient-specific representations of both whole heart and regional structural and functional analysis. A total of twenty-two papers submitted underwent two to three rounds of rigorous peer review. In this Special Issue, thirteen papers were eventually chosen for publication. Each paper was closely reviewed by 3–4 experts and went through a thorough phase of revision, usually consisting of at least two rounds of revision. There were a few excellent papers that, unfortunately, due to space constraints and reviewers’ feedback, could not be included in the Special Issue. 2. Special issue papers