Read more

August 25, 2020
3 min read
Save

AI can benefit care of patients with CVD, diabetes

You've successfully added to your alerts. You will receive an email when new content is published.

Click Here to Manage Email Alerts

We were unable to process your request. Please try again later. If you continue to have this issue please contact customerservice@slackinc.com.

Artificial intelligence has the ability to affect CV and diabetes care for both patients and clinicians, a presenter said.

Irving K. Loh

“The promise of artificial intelligence in health care is intriguing for controlling costs, increasing convenience, improving quality and improving outcomes,” Irving K. Loh, MD, FACC, FAHA, FCCP, FACP, chief medical officer and co-founder of Infermedica, internist and cardiac specialist at the Ventura Heart Institute in Thousand Oaks, California, chair of technology and innovation for the California chapter of the American College of Cardiology and adjunct lecturer in the department of medicine at Stanford University, said during a presentation at the virtual Heart in Diabetes Conference. “We do need to resolve the issues of data privacy, interoperability and regulatory concerns.”

Heart matrix_Adobe Stock
Source: Adobe Stock.

AI provides the ability to take general principles and apply them to new environments. The most common use of AI is narrow AI, which is used for a restricted range of predefined functions to accomplish narrowly defined tasks, Loh said during his presentation.

“Artificial intelligence health care will actually revolutionize the way physicians, other health care providers and patients interact with health data,” Loh said.

AI can help transform large amounts of data into meaningful and actionable information, according to the presentation. Loh said providers must be vigilant on what to train their systems on to avoid potential learning bias in training sets.

Subsets of AI include machine learning, which uses statistical methods to improve experiences, and deep learning, which allows computation of multilayer neural networks, according to the presentation. A neural network is the architecture of a machine learning strategy, which teaches a computer to perform a specific task with training data sets. This network can learn from unstructured and unlabeled data.

Artificial intelligence will not replace human health care providers because computers cannot, as of yet, reason, but they do ... better statistical predictions when they’re trained on huge data sets,” Loh said. “I want to point out that this is not human intelligence. Computers do not have common sense.”

Instead, AI can provide tools to augment and extend effectiveness, especially in the presence of lots of data like whole-genome sequencing, new clinical trials and data streams from wearables, according to the presentation.

AI can play three major roles in health care: classification, prediction and facilitation. These tasks could reduce the burden of electronic health records by creating more intuitive interfaces and automating routine processes. In the future, EHRs may feature virtual assistants with embedded AI for reminders, scheduling and medication refills, among other tasks, according to the presentation.

AI and machine learning can offer several benefits to cardiology. Predictive analytics can help inform cardiologists of problems before they occur such as sudden cardiac death and MI. It can also help with precision phenotyping through unsupervised learning analysis of complex data sets. This can lead to improved patient care because it provides cardiologists with more information in addition to providing streamlined care.

Medical grade and FDA-certified wearables can be linked to the EHR and track vital signs, including heart rate, BP and temperate. Wearables can also track diseases such as atrial fibrillation, HF and diabetes.

AI and machine learning can also offer several benefits to the approximately 425 million people with diabetes globally, according to the presentation. This accounts for approximately 12% of the world’s health care expenditures. In addition, half of these people often go undiagnosed and untreated.

The areas in which AI can focus on related to diabetes include automated retinal screening, predictive analytics, self-management tools such as glucose sensors and clinical decision support.

“AI-driven predictive modeling identifies diabetics with highest risk of avoidable complications, hopefully reducing hospitalizations and readmissions,” Loh said.

AI can also potentially help patients with diabetes to achieve better blood control and reduce hypoglycemic episodes, in addition to detecting disease earlier and provide greater accuracy for patients, their families and physicians, according to the presentation.

Clinical decision support is another area that can benefit from AI, as it can help sift through information to suggest the next steps related to diagnosis, testing and treatment options, according to the presentation. It can also warn of potential problems and drug interactions through prompts, alerts and alarms.

Loh said AI can help reconcile the complexities of evidence-based guidelines from Europe and the United States. When the technology is given specific patient information, clinicians can navigate the different concepts and complexities from these guidelines, which can lead to the best assessment and treatment options for their patients. 

Clinicians still must learn more about how to integrate AI into the clinical workflow so that it is seamless and painlessly adopted, according to the presentation. 

“If we partnered with AI technologies and unburdened ourselves from the clerical drudgery, we can refocus ourselves on clinical tasks that actually matter and restore meaning and purpose in being a physician with new levels of efficiency and accuracy,” Loh said.