Deep learning in spatiotemporal cardiac imaging: A review of methodologies and clinical usability

https://doi.org/10.1016/j.compbiomed.2020.104200Get rights and content
Under a Creative Commons license
open access

Highlights

  • First review of deep learning methods in dynamic cardiac imaging using the full spatiotemporal image information.

  • Overview of deep learning models, architectures and configurations relevant for spatiotemporal image data.

  • Evaluation of clinical usability w.r.t. applications, datasets, learning and training methods, and test performance.

  • Clinical usability of deep learning applications solely achieved a robust candidate or proof of concept status.

Abstract

The use of different cardiac imaging modalities such as MRI, CT or ultrasound enables the visualization and interpretation of altered morphological structures and function of the heart. In recent years, there has been an increasing interest in AI and deep learning that take into account spatial and temporal information in medical image analysis. In particular, deep learning tools using temporal information in image processing have not yet found their way into daily clinical practice, despite its presumed high diagnostic and prognostic value. This review aims to synthesize the most relevant deep learning methods and discuss their clinical usability in dynamic cardiac imaging using for example the complete spatiotemporal image information of the heart cycle. Selected articles were categorized according to the following indicators: clinical applications, quality of datasets, preprocessing and annotation, learning methods and training strategy, and test performance. Clinical usability was evaluated based on these criteria by classifying the selected papers into (i) clinical level, (ii) robust candidate and (iii) proof of concept applications. Interestingly, not a single one of the reviewed papers was classified as a “clinical level” study. Almost 39% of the articles achieved a “robust candidate” and as many as 61% a “proof of concept” status. In summary, deep learning in spatiotemporal cardiac imaging is still strongly research-oriented and its implementation in clinical application still requires considerable efforts. Challenges that need to be addressed are the quality of datasets together with clinical verification and validation of the performance achieved by the used method.

Keywords

Deep learning
Cardiovascular imaging
Spatiotemporal image data
Clinical usability

Cited by (0)

Karen Andrea Lara Hernandez (f), MSc is a PhD student in biomedical engineering at the Institute of Health Care Engineering with European Testing Center of Medical Devices, Graz University of Technology, Austria. She is a director of the study program in biomedical engineering at Galileo University, Guatemala City. Her research interests include biomedical image and signal processing, AI and deep learning.

Theresa Rienmüller (f), PhD is an Assistant Professor at the Institute of Health Care Engineering with European Testing Center of Medical Devices, Graz University of Technology, Austria since 2018. She received her PhD at UMIT, Austria in 2013. Multiple research grants in the field of biomedical engineering. Her research interests include cardiac physiology and medical image processing, modeling and simulation, control and systems theory, biomedical sensors and signal processing.

Daniela Baumgartner (f), MD is a Professor of Pediatric Cardiology at Medical University of Graz, Austria since 2015. Deputy Director of the Clinic of Pediatric Cardiology and Head of the Outpatient Clinic for Pediatric Cardiology. She received her MD at Medical University of Innsbruck, Austria in 1991, Associate Professor in 2006, Jus practicandi in General Medicine, Pediatrics, Pediatric Cardiology, Pediatric Intensive Care Medicine and Neonatology. Her research interests include cardiac imaging, connective tissue disorders like Marfan syndrome, aortic elasticity, aortic coarctation, aortitis and cardiac genetics.

Christian Baumgartner (m), PhD is Professor and Head of the Institute of Health Care Engineering with European Testing Center of Medical Devices at Graz University of Technology, Austria since 2015. PhD degree (1998) from Graz University of Technology, Austria, diploma in organ and conducting (1998). Professor of Biomedical Engineering at UMIT, Austria (2009–2014). Dr. Baumgartner is the author of more than 180 publications in refereed journals, books and conference proceedings, and patents and is a reviewer for more than 40 scientific journals. He served as a deputy editor of the Journal of Clinical Bioinformatics, is section editor of Sensors (biomedical sensors), co-section editor of the IMIA Yearbook of Medical Informatics, and is an editorial board member of Cell Biology and Toxicology and Methods of Information in Medicine. His main research interests include cellular electrophysiology, biomedical sensors, signals and imaging informatics, machine learning and computational biology as well as medical device development, safety and regulatory affairs.