A review of the top 3 AI in cardiology papers of the past year.

 

1. Zimmerman A et al. Usefulness of Machine Learning in COVID-19 for the Detection and Prognosis of Cardiovascular  Complications.Review in Cardiovascular Medicine 2020.

Machine learning can be used to assess massive quantities of data for applications in diagnosis, prognosis, and therapy in COVID-19 infection. While there are conventional trials and case series/prospective and retrospective cohort studies, there is a paucity of studies utilizing machine learning. Common algorithms used in machine learning include support vector machine and decision trees as well as artificial neural network. Ongoing clinical trials on COVID-19 and the cardiovascular diseases using artificial intelligence include QT-Logs (safety of hydroxychloroquine), CARDICoVRISK, and Coviva (survival). There is a dire need for real-time artificial intelligence on an international cardiovascular disease data repository.

2. Ghorbani a et al. Deep Learning in Interpretation of Echocardiograms. Nature NPJ Digital Medicine, 2020.

Deep learning with a 75-layer convolutional neural networks (CNN) (EchoNet) and 2.6 million echocardiogram images (from close to 3,000 patients) with TensorFlow model training can identify cardiac structures and features (such as enlarge left atrium and left ventricular hypertrophy), estimate cardiac function and volumetric measurements, and predict systemic phenotypes (age, sex, etc) that modify cardiovascular risk. This can be useful for preliminary interpretations and improving workflow as well as remove repetitive tasks.
This study can lead to future investigations using combined CNN and recurrent neural network (RNN) as well as comparing humans and humans + artificial intelligence. Future applications also include improved image acquisition and automated interpretations that include comparisons with past studies.

3. Yabushita H et al. Development of Novel Artificial Intelligence to Detect the Presence of Clinically Meaningful Coronary Atherosclerotic Stenosis in Major Branch from Coronary Angiography Video. Journal of Atherosclerosis and Thrombosis, 2020.

A total of close to 2,000 videos from 200 patients were collected for a three-dimensional convolutional neural network to train. A training set of 146 patients data was used and split into additional data sets for validation. The predictive accuracy of the test set showed a c-statistic of 0.61 in both validation and test datasets for detection stenosis of 75% or greater. The modest predictive value may be of use as an augmentation tool.

Moderator: Dr Anthony Chang, Founder, AIMed and Chief Intelligence and Innovation Officer, Children’s Hospital of Orange County (CHOC)

Dr Rob Brisk, DevRel and Alliance Manager, NVIDIA and Clinical Researcher in AI and Cardiology, Craigavon Area Hospital

Louise Sun, MD SM FRCPC FAHA, Cardiac Anesthesiologist and Clinician Scientist, Director of Big Data and Bioinformatics Research, University of Ottawa Heart Institute

Animesh Tandon, MD, MS; FAAP, FACC, FAHA, Assistant Professor of Pediatrics; Division of Cardiology, Joint appointment in Radiology; Faculty in Biomedical Engineering, University of Texas Southwestern Medical School and Children’s Medical Center Dallas

Thank you faculty and attendees for your knowledge and expertise as we all learned a great deal today at AIMed Cardiology!