Representative images of the segmentation of the aortic lumen (in red) and the intraluminal thrombus (in green). (
) CT scan cross-sectional views of patients with infrarenal AAA. (
) Manual segmentation. (
) Automatic segmentation. Reprinted with permission under open access from Lareyre et al.
CT pulmonary angiography (CTPA) is a standard imaging modality in diagnosing pulmonary embolisms (PEs) with high accuracy. However, the interpretation of CTPA images is time-consuming and requires radiologist expertise. Further, the use of CTPA in emergency departments has increased significantly over the last decades, and increased workloads and fatigue may lead to more diagnostic errors in emergency radiology
[167]. Therefore, the use of an automatic PE detection method could assist radiologists’ decisions to ensure the rapid diagnosis of positive PE cases while avoiding mistakes.
Studies have reported promising results regarding the application of DL models for the automatic detection of PEs based on CTPA images
[168][169][170][171]. A recent systematic review and meta-analysis of using DL in the detection of PEs showed a pooled sensitivity of 88% and a specificity of 86% based on an analysis of five studies
[168]. This indicates that further studies are required to validate the DL models to detect PEs based on large datasets of CTPA images. The Radiological Society of North America (RSNA) chose PE as its AI challenge in 2020, later publishing a public dataset of 12,195 annotated CTPA studies to encourage the development of DL models for PE detection
[172].
6. Summary
Cardiovascular CT is playing an important role in the diagnosis of cardiovascular disease, and its role will continue to grow with further advancements in CT technology. The traditional reliance on the standard CT imaging approach has been significantly augmented with the use of recent technologies such as photon-counting CT, 3D printing, FFRCT, VR, AR, and MR, and AI. This recent progress has created great opportunities for incorporating these advanced technologies into education and clinical practice to achieve better outcomes.
The clinical value of CT has been further advanced with the recent emergence of photon-counting CT, which can be used to obtain images with superior spatial and contrast resolution. Despite being introduced into clinical practice very recently (in 2021), photon-counting CT represents the future of cardiovascular CT imaging and is set to revolutionize the current cardiovascular CT imaging approach, especially in the diagnostic assessment of cardiovascular disease.
The current applications of 3D printing technology in cardiovascular research are maturing, with more evidence available from multi-center or randomized controlled trials being developed. Three-dimensionally printed models are highly accurate and reliable when it comes to replicating both cardiovascular anatomy and pathology, thus serving as a useful tool for medical education, surgical planning, and the simulation of challenging cardiovascular procedures, guiding intraoperative surgeries to improve patient outcomes. Three-dimensionally printed models improve communication between clinicians and patients, as well as communication between clinical colleagues.
The clinical value of FFRCT has been validated by several multi-center randomized controlled studies and many single-site studies, and its role will continue to grow with the increasing use of DL algorithms in the medical domain. The main barrier to implementing FFRCT in daily cardiology practice lies in the fact that most of the data analyses have been performed at off-site workstations, although on-site image processing and analyses are available, as evidenced by the TARGET trial. With the increasing prevalence of DL models and widespread use of AI in clinical practice, FFRCT will be implemented into diagnostic approaches to guide the revascularization of patients with coronary artery disease, leading to improvements in the utilization of healthcare resources.
VR, AR, and, more recently, MR are showing great promise, and they are set to complement traditional visualizations and assist healthcare providers and patients with cardiovascular disease. However, their applications in current practice are still at an early stage of development due to several limitations. First, the real-time integration of cardiovascular CT imaging in a VR/AR environment is challenging. Second, a critical aspect regarding the use of AR and MR in surgical planning or guiding surgical procedures is to balance AR/MR with the real-world environment and add digital elements to the field of view to achieve the harmonization of data flow and interfaces. Third, ethical considerations need to be considered, as the main goal of using VR/AR/MR should focus on enhancing the patient–provider relationship.
The widespread use of AI in medicine and the application of AI in cardiovascular disease is inevitable, and clinicians must be aware of pitfalls when applying this rapidly evolving technology to their practice.
Figure 16 is a summary of AI applications in cardiology practice
[173]. Given the wide range of AI algorithms available, the input data used for training purposes must be examined to ensure high data quality. AI model performance must be examined to guarantee that findings are robust, and external validation is also an important consideration. Medical graduates and clinicians’ skills and confidence in managing AI applications need to be improved, as this will have a direct impact on using AI in clinical practice. Ethical issues related to the sharing of healthcare data and legal challenges should be addressed, and AI should be included in the medical curriculum and professional education. Collaboration among a multi-disciplinary team consisting of computer scientists, clinicians, clinical investigators, academic researchers, and other users is essential for identifying the best approach and data sources to achieve the goal of delivering personalized treatment in cardiovascular disease cases.
Figure 16. The applications of artificial intelligence in clinical cardiology practice. CAC—coronary calcium score, CAD—coronary artery disease, EAT—epicardial adipose tissue, PVAT—perivascular adipose tissue, LV—left ventricle. Reprinted with permission under open access from Jiang et al.
[173].