Integration of artificial intelligence into clinical patient management: focus on cardiac imaging

Abstract

Cardiac imaging is a crucial component in the management of cardiac patients, and as such it influences multiple, inter-related parts of the clinical workflow: physician-patient contact, image acquisition, image pre-and post-processing, study reporting, diagnostics and outcome predictions, medical interventions, and, finally, knowledge-building through clinical research. With the gradual and ubiquitous infiltration of artificial intelligence into cardiology, it has become clear that, when used appropriately, it will influence and potentially improve -through automation, standardization and data integration─ all components of the clinical workflow. The aim of this review is to present a comprehensive vision of a full integration of artificial intelligence into the standard clinical patient management ─with a focus on cardiac imaging, but applicable to all information handling-and discuss current barriers that remain to be overcome before the wide-spread implementation and integration.

Functional Coronary Artery Assessment: a Systematic Literature Review

“…In the scope of Machine Learning-based (ML) techniques, the major part of recent studies use CCTA characteristics as reference [3,5,21,28,34,39,45,52], but most of them also were evaluated with a reduced number of patients and arteries [5,21,28,33,34,42,50]. These researches use geometrical lesion data [3,21,28,39,42,45,52], clinical risk scores [39] and anatomical descriptors [34].…”

“…In the scope of Machine Learning-based (ML) techniques, the major part of recent studies use CCTA characteristics as reference [3,5,21,28,34,39,45,52], but most of them also were evaluated with a reduced number of patients and arteries [5,21,28,33,34,42,50]. These researches use geometrical lesion data [3,21,28,39,42,45,52], clinical risk scores [39] and anatomical descriptors [34]. In general, the most common recommendation is the use of combined model with anatomic-physiologic analysis to determine the real severity of the lesion [1,36,59] as well as the use of angiography-based methods when FFR is not available [59].…”

“…The major part of the ML-based techniques proposed in recent studies follow the image-based approach [16,21,28,31,33,42,45,54] using (deep) neural networks, which typically require large databases to be properly trained. This need for a huge amount of data associated with the lack of openly available datasets has also led to synthetically generated databases [21].…”

“…The plaque characteristics are mainly used in ML-based techniques and are challenging to reproduce [3,28,71]. For this reason, we noticed the increasing use of transfer learning (TL) approaches to leverage the networks performance [33,45] and the creation of a complete sequence of steps to stenosis detection [54]. TL is the process of improving the target predictive function by using the related information from an already trained function, improving a small part of the network parameters and reusing the initial layers, which represent the most generic features [75].…”

“…The importance of non-image-based characteristics to determine the lesion severity, such as patient conditions, can provide a better view of the patient. Current evidences suggest the need of improvements in these methods to reach better correlation with FFR in terms of outcome [8,33,42,45], even showing good performance in low contrast exam regions [43].…”