There have been two major inflection points in the development of artificial intelligence (AI) in cardiac ultrasound that will lead to major changes in echocardiography workflows in the near future. These include the introduction of deep learning and generative pre-training transformer (GPT) algorithms.
These two key AI technologies will fundamentally change how cardiologists and sonographers do their jobs, explained cardiology AI expert Partho Sengupta, MD, DM, Henry Rutgers Professor of Cardiology and chief of cardiovascular medicine at Robert Wood Johnson Medical School and chief of the cardiovascular service line at Robert Wood Johnson University Hospital. He spoke on these topics in AI sessions at the American Society of Echocardiography (ASE) 2023 meeting in June, and shared his thoughts in an interview with Cardiovascular Business.
Deep learning is automating measurements and helping read exams
"Deep learning is going to completely automate all the measurements and these will auto populate on the report. This is real and it is coming. Some of this is going to be built into the ultrasound machines, and it is really going to offload the technologists and doctors who feel so burned out doing all these measurements," Sengupta explained. "In the next two or three years, everything in echocardiography will be automated with AI in terms of measurements."
AI automation will save significant amounts of exam and post-processing time for sonographers and the cardiologists. If the tedious and time-consuming work of drawing lines and making measurements can be removed, he said the humans can spend more time interacting with their patients rather than with the machine.
Most of the FDA-cleared AI algorithms in echo are for the automation of measurements. Many of these are already available on current ultrasound systems to help automate ejection fractions, assessing all the measurements needed for structural heart assessments, strain, and others. Other deep learning algorithms can help act as a second set of eyes to help diagnose patients with specific conditions, including aortic stenosis, heart failure and amyloidosis.
"These applications are very useful in echocardiography to automate the process and create more time for us to do more higher-caliber activity," he said.
By training AI to always take measurements from ideal planes and anatomic landmarks, it can make measurements much more reproducible and help eliminate variability between sonographers.
GPT AI may significantly change cardiac imaging
"The second inflection point that I am really excited about is the development of GPT," Sangupta said.
He explained these algorithms basically connect the language in the body of the reports and in the electronic medical records for new ways of communication between physicians. GPT in the future might also begin connecting the physicians and the patients to the images.
Sangupta said the new frontier is going to be the ability to train GPT models to create effortless reports and communicate clinical information to referring physicians. GPT is also being developed to look at clinical reports and other patient information to instantly create a layman's language version of exam reports that are more patient friendly and easier to understand.
"It will give them the ability to understand what is in their reports and their records. There is no doubt that communication piece is lacking right now. This will give them the ability to digest and understand the report so they can have a decision-making capability," Sangupta explained.
GPT AI also might be able to provide additional relief for clinicians who are feeling burned out with the amount of work that is required. Sangupta said it is still in the early days of GPT development, so the technology is not ready yet for regular use in patient care. There are also many questions about where the algorithms are pulling their data and how they reach conclusions.
"There is a great deal of uncertainty, but there is also a great deal of opportunity," he explained. "But we need to embrace this because this change is imminent."
This change is also necessary to enable moving patient care forward and to address burnout.
"I believe a lot of physicians experience burnout because of the additional work they have to do with the electronic medical record, which is not really rewarding. So the whole idea of being able to create more free time so doctors can do what they like and do best, which is talk to the patients. Everything else can be put into the EMR using ChatGPT, and that is absolutely needed," he said.
Why human clinicians will not be replaced by AI or GPT technology
Sangupta is certain there is no way a clinician can be replaced by GPT technology anytime soon because the technology is just not as capable as human brain or understanding context in the human world.
"The human level competency of GPT is still very underdeveloped. Even if these models are able to become more generalizable and precise, they lack the human element of intelligence, which has several layers. There is contextual information that is clearer and easier for the human mind to digest, versus a ChatGPT," Sangupta explained.
Often the diagnosis or treatment of a patient is based on a doctor's intuition and clinical judgement. And this is hard for programs to replicate and that is where clinical expertise, for now, is still a human trait required for patient care.
"It also comes down to empathy and why we took the Hippocratic Oath, to make patients feel better. ChatGPT does not have feelings, so it does not have these sets of information to drive what is good for patient care, and that is where human doctors need to focus on," Sangupta said.
What he said will happen is that clinicians will have new tools for exploring the unknown components of the images with the ability to see far better than the human eye. He also said this new AI will be able to parse out large amounts of data and distill it for easier human consumption.
Dave Fornell has covered healthcare for more than 17 years, with a focus in cardiology and radiology. Fornell is a 5-time winner of a Jesse H. Neal Award, the most prestigious editorial honors in the field of specialized journalism. The wins included best technical content, best use of social media and best COVID-19 coverage. Fornell was also a three-time Neal finalist for best range of work by a single author. He produces more than 100 editorial videos each year, most of them interviews with key opinion leaders in medicine. He also writes technical articles, covers key trends, conducts video hospital site visits, and is very involved with social media. E-mail: dfornell@innovatehealthcare.com