Artificial intelligence is evolving healthcare in innumerable ways.
Though the public at large currently pictures consumer-focused tools like ChatGPT and the Dall-E image generator when they think of AI, a myriad of AI-based applications are being implemented in the background of healthcare systems around the country, improving patient outcomes and enhancing the workflows of providers everywhere.
One of our partners, Echo IQ, recently sought FDA approval for their AI-based cardiology workflow, EchoSolv. This is the world’s only AI platform built on the world’s largest database of echocardiograms linked to mortality.
Invisible to the patient because it supplements the workflows of cardiologists and cardiology imaging teams, EchoSolv automatically identifies patients at risk of structural heart disease with unparalleled speed and accuracy. Namely, it helps cardiologists detect aortic stenosis, particularly in patients who might otherwise go undiagnosed.
With multiple additional diagnostic pathways in development, including heart failure, pulmonary hypertension and more, Echo IQ is reshaping cardiac ultrasound workflows and aiding cardiologists in their ability to help patients.
Here's how their groundbreaking AI tool is making a difference for patient lives around the world.
Removing Bias from the Equation
Echo IQ began EchoSolv’s development by focusing on aortic stenosis because it's a condition that’s radically underdiagnosed, particularly among women. Their early research determined women were actually 66% less likely to be accurately diagnosed than their male counterparts.
There are many different reasons for that, but one key issue is unconscious bias, when the possibility of aortic stenosis is dismissed because the patient’s body type or background doesn’t immediately “fit the picture” of what heart disease looks like.
First, there's the biases that we tend to think of, socially and culturally, as well as things like gender or age or race. But there's also symptom bias. For instance, if someone presents without symptoms, they're outside of the general phenotype of what you would expect and thus might not be properly diagnosed. If you had a young female who had no symptoms, those are not the typical characteristics you would expect for a disease like aortic stenosis, which tends to impact elderly patients with comorbidities. If you don't have someone with symptoms and you don't have any other reason to look for it, you might not catch it.
Echo IQ’s researchers made intentional decisions to make sure that those biases were removed from the training data, ensuring that EchoSolv helps clinicians minimize their in-field bias and achieve consistency across their patient population.
LVOT or Leave It
There’s one particular area of the cardiac system that is somewhat difficult to image and, therefore, can contribute to an inaccurate diagnosis: the left ventricular outflow tract, or LVOT.
Sonographers who are able to properly gain an image of the LVOT and do so quickly without rescans are truly artists, because it’s a very hard thing to do correctly. The left ventricular outflow tract is difficult because it’s not a circle, it's more of an oval. You need absolute precision in order to be able to measure it.
That precision is key, because once you input this measurement, it affects the resulting continuity equation used to determine the size of the aorta and thus the possibility of stenosis. Whatever error is there, even a small one, will be amplified exponentially.
If a patient doesn't present symptomatically and has other characteristics that seem atypical for aortic stenosis, and that LVOT measurement had a slight error, the cardiologist might have no reason to initially suspect aortic stenosis is present, or they might otherwise dismiss evidence of stenosis because it’s not consistent with the results of the LVOT ultrasound.
This is known as hemodynamic bias. When combined with outdated stenosis guidelines based on a 1960s, primarily male sample size, it creates a clinical process that tends to underestimate the at-risk population.
Where AI and EchoSolv come in is in the enormous amount of data and the ability to apply different mathematical processes and permutations to make sure the continuity equation is properly assessed for accuracy. In addition, EchoSolv’s entire neural network was trained to avoid the LVOT entirely because of the outsize possibility for error. So if other evidence suggests stenosis, an inaccurate LVOT scan won’t contradict that finding.
Integration Is at The Heart of Adoption
One of the biggest challenges for some AI applications is integrating with current workflows, and EchoSolv was built to solve this dilemma.
Cardiologists and their teams have an intense job, and they're constantly looking at different sets of data to make a diagnosis. This can create fatigue from going from one screen to another, taking in information, and then processing it.
To this end, EchoSolv doesn’t require EMR integration or another screen. It automatically pulls information from the cloud and updates accordingly. The cardiologist can take the measurements that are of interest and receive a notification of the AI's findings, with the system trained to surface data that requires further human attention. All in a package that is compliant and secure. And then, if an organization does want to integrate EchoSolv into their EMR, they can do so easily.
Wearing Your Halo
One question administrators will want to know is the additional revenue potential of any AI system. EchoSolv creates halo revenue because of the increase in aortic stenosis detection, which necessarily creates referrals for additional tests. This is something that benefits the patient, obviously, as they’ll be the recipient of further diagnostics that can help confirm a diagnosis or suggest treatment plans, but it also means further revenue for the healthcare facility in question.
Because more people could be properly diagnosed with severe aortic stenosis, that means more patients are likely to be sent to a cath lab for further evaluation. They’re probably going to get calcium scoring done, where previously they might not have even been diagnosed. They might be scheduled for follow-up surgeries or other interventions.
The financial benefits of these additional interventions are likely to be dispersed throughout the healthcare ecosystem, meaning the revenue is spread around between, say, the imaging center, the hospital, the cath lab, etc. This makes EchoSolv an attractive option for larger health systems, because they capture all of that revenue.
Evolving to Serve Patients
Perhaps the most exciting part about EchoSolv is that its potential isn’t theoretical. Aortic stenosis is a solvable disease so long as it can be properly diagnosed as early as possible. The interventions work. If the right patients are diagnosed at the right time, and the evidence supports intervention to achieve a beneficial outcome, then those patients’ lives could be saved.
Obviously, intervention is not without risk. But underdiagnosed patients can die or face other adverse outcomes at an even more alarming rate, often because they’re asymptomatic right up until they have a cardiac event.
There’s a lot of concern about AI right now, but tools like EchoSolv truly bring democratization to healthcare, making sure people are provided equal treatment regardless of their background and situation.
To learn more about Echo IQ and EchoSolv, click here, and contact us for more details about how to start improving diagnoses of aortic stenosis at your facility.
Comments