I recently published an article on ‘intelligent stethoscope’ sales and the predicted growth for the future. While researching the details for that, I came across another very interesting development worthy of discussion.
The fact that smart stethoscopes will eventually incorporate AI is inevitable. And one recent clinical study has looked into the practicalities of utilising Artificial Intelligence for heart screening and diagnosis.
The results are clear to read in the title of this article. I for one, was shocked by the outcome.
Artificial Intelligence Detects Pediatric Heart Murmurs With Cardiologist-Level Accuracy
The clinical study abstract was presented by Eko, (a well known stethoscope manufacturer, I have covered here) earlier this month.
The presentation took place at the American Heart Association (AHA) Scientific Sessions in Chicago, and caused quite the discussion.
The team at Eko have been working on next generation cardiac monitoring using non-invasive sensors combined with machine learning, artificial intelligence.
Eko announced that during the clinical study, the heart murmur detection algorithm was able to outperform 4 out of 5 cardiologists. The test was successfully detecting a heart murmur in a series of screening sounds.
In short, the computer outperformed the cardiologists 80% of the time. To me that is staggering.
The algorithm has been designed with the aim of giving clinicians more advanced diagnostic tools in order to provide superior cardiovascular care.
And while this is definitely a positive development, the fact an algorithm was able to outperform the human diagnosis so comfortablly, does give pause for thought.
The AI Algorithm – Development & Tests
Eko, a San Francisco-based startup founded in 2013 by Connor Landgraf, Jason Bellet, and Tyler Crouch, is at the forefront of AI, heart sound and EKG sensor development for heart disease monitoring.
The company has created a wide range of FDA approved software to help facilitate in-clinic heart disease screening as well as methods of telemedicine healthcare.
The team, ‘trained’ the neural network AI algorithm using thousands of heart sound recordings.
The AI was then tested using a huge amount of recorded heart sounds. This was all configured alongside echocardiogram imagery. The system, literally learned via the digital consumption of a wide cardiovascular screening dataset.
For the study, the AI and 5 pediatric cardiologists listened to a series of heart sound recordings and individually, each participant had to determine whether a murmur was present or not.
The results, as we have already mentioned were certainly affirmative.
The Future for Artificial Intelligence Heart Screening
It is clear that such technology will be beneficial for healthcare professionals.
As was stated in the presentation, use of AI in this scenario will help narrow the clinical skill gap between the 27,000 cardiologists currently practicing in the U.S.
Those that are considered experts at murmur detection will benefit from a 2nd opinion so to speak; other clinicians that do not have the necessary experience, could essentially be made aware of the condition using while using this technology in a standard auscultation check up, (assuming they have access to the diagnostic tool in question).
That all being said, it could be quite some time before we see this AI being used in practice. Eko is pursuing FDA clearance for the algorithm, however one assumes there will be a significant amount of ‘red tape’ before sufficient regulatory clearance is provided.
The future does look promising however, after Eko has received clearance, their murmur screening algorithm will naturally be linked to their already FDA-cleared Eko Core and Eko DUO devices.
The result will take ‘intelligent stethoscopes’ to a whole new level, enabling clinicians immediate access to a databank of sounds, and an AI system able to assist in complex diagnosis.
- Eko Blog post: AI Algorithm Outperformed Majority of Cardiologists in Detection of Heart Murmurs in Clinical Study
- Becker Hospital Review: Mayo Clinic partners with AI firm to screen patients
- What This Computer Needs Is a Physician Humanism and Artificial Intelligence; Authors: Abraham Verghese, MD1; Nigam H. Shah, MBBS, PhD1; Robert A. Harrington, MD1
- Artificial intelligence powers digital medicine; Authors: Alexander L. Fogel & Joseph C. Kvedar