Learn how imaging leaders are benefiting from early AI adoption

AI's role in CV care

While some stakeholders view AI as a potential threat to the workforce, the evolution of AI in cardiovascular (CV) care is not so much supplanting experienced clinicians as it is supporting them. There is a significant opportunity for AI to support faster identification of risk and disease in patients. Not only would this advance have an impact on care quality—enabling earlier intervention to slow disease progression—but it would also help save on cost for the service line by increasing clinician efficiency.

Get best practices to build a high-value CV care team

Some of the largest tech companies in the world, such as Apple and Google, are already investing in AI technologies they hope will innovate CV care, and this trend is expected to keep increasing over time. To help you start thinking about how your service line could benefit from AI, here are some of the most recent ways AI technologies are improving CV patient identification and care planning:

1. Forecast CV disease risk. Google AI published a study in the journal Nature earlier this year showing that the organization was able to use machine learning to predict patients' risk of CV disease from retinal images. Using data from 284,335 patients, the model was able to accurately identify CV risk factors such as age, gender, smoking history, and blood pressure from pictures of the patients' retinas. From these identified risk factors, the algorithm was then able to directly predict the patients' CV disease risk with a relatively high degree of accuracy, creating a surprising new risk prediction model. Google AI stated that this algorithm was not developed with the intention of imitating or replacing existing diagnostic techniques, but rather to complement those techniques and give new insights into the underlying causes of CV disease.

2. Detect symptoms in real-time. Apple's announcement that its new Apple Watch Series 4 would feature an FDA-approved electrocardiogram (EKG) on the back of the device prompted a flurry of discussion throughout the CV world. The watchmaker says the device is capable of detecting atrial fibrillation (AF) in its wearers in real-time.

Read our take on the impact of the Apple Watch's new EKG feature

This feature, however, is not the first time developers have looked to the Apple Watch as a means of detecting arrhythmia. In 2017, a team at the University of California-San Francisco (UCSF) developed an AI algorithm capable of detecting AF in Apple Watch wearers with 97% accuracy using data from an app called Cardiogram. Stanford University is also currently engaged in a study with Apple to see if the Apple Watch effectively detects AF in more than 400,000 subjects, but the results of this trial are not expected to be released until early 2019.

3. Automate echocardiogram interpretation. A  study published in the journal Circulation in September 2018 showcased the first example of a fully automated echo interpretation model. Using 14,035 echoes spanning a 10-year period, the algorithm was trained to identify image viewpoint and segmentation; quantify chamber volumes, ventricular mass, ejection fraction, and longitudinal strain; and detect hypertrophic cardiomyopathy, cardiac amyloid, and pulmonary arterial hypertension from the images. According to the study, the algorithm's reads were better than manual measurements based on 11 consistency metrics, but the study authors insist that the purpose of this algorithm is not to replace techs or cardiologists. Rather, the authors hope that the algorithm will augment existing clinical practices by helping to identify diseases in earlier stages, shift echo evaluation to primary care settings in rural areas, and drive down imaging costs for providers and patients.

Download the playbook for reducing CV care variation

4. Identify coronary artery disease. Using AI and a technology called Cardiac Space Phase Tomography Analysis (cPTSA), researchers at Analytics 4 Life were able to develop a machine learning algorithm to detect the presence of coronary artery disease (CAD) in patients. The algorithm was tested on 94 patients, and the success of the results suggest that this algorithm can accurately detect CAD in patients without the need for a stress test or exposure to radioactivity.

5. Accelerate procedure planning for TAVR patients. A team in South Korea developed a machine learning algorithm to automate the localization of aortic valve landmarks in the CT images of pre-operative transcatheter aortic valve replacement (TAVR) patients. This is a critical process in pre-operative TAVR workflows, and doing this manually is arduous and time-consuming. This novel algorithm requires only 12 milliseconds to identify all eight aortic valve landmarks, so this technology has the potential to save physicians time, accelerate the pre-operative surgical planning process, and standardize TAVR care.

6. Predict HF readmission risk. Partners Health and Hitachi, Ltd. (Japan) collaborated to develop a machine learning algorithm to predict heart failure (HF) readmission risk. The algorithm was trained using data from 11,510 HF patients with 27,334 admissions and 6,369 30-day readmissions, and from these data the computer created a 3,512-variable HF readmission risk prediction model. This model performed better than traditional EHR-based risk prediction techniques for HF readmissions, and the research team hopes that this algorithm will be used to further minimize HF readmissions, reducing hospital costs and improving outcomes.

With all of these opportunities coming down the pike, there's understandably a lot of excitement surrounding the progression of AI in the CV space. That said, there are still hurdles to widespread adoption. Among these barriers are the cost of the technologies, staff resistance, workflow implications, and legal and ethical considerations.

Read the full list of challenges

While some health care providers are justifiably apprehensive about the future of AI, embracing the potential of these technologies could help alleviate the burden of future staffing shortages and increased margin pressure that institutions are inevitably going to face.