Researchers at Mount Sinai in New York City say they’ve developed a new artificial intelligence model designed for electrocardiogram analysis.
WHY IT MATTERS
The innovative AI approach could vastly improve the efficacy and accuracy of ECG assessment, according to the health system, as the model enables interpretation of cardiac readings as language.
This approach can enhance both the accuracy and efficacy of ECG-related diagnoses, said Mount Sinai clinicians, especially for rarer cardiac conditions where there’s not as much data on which to train machine learning algorithms.
Their new report, A foundational vision transformer improves diagnostic performance for electrocardiograms, is published this week in npj Digital Medicine.
In it, researchers describe how the deep learning model – called HeartBEiT – enabled models that surpassed current methods for ECG analysis, and could eventually be a foundation for other specialized approaches to diagnosis and assessment.
Mount Sinai says its research builds on “the intense interest in so-called generative AI systems such as ChatGPT, which are built on transformers.”
Those transformers, algorithms trained on huge text datasets to generate human-like responses to user prompts, are informing this current research, which uses an image-generating model to create discrete representations of small parts of the electrocardiogram – enabling researchers, according to Mount Sinai, to analyze raw ECG data as “language.”
HeartBEiT was pre-trained on 8.5 million ECGs from more than 2 million patients, collected over the past 40 or so years from four Mount Sinai hospitals. Its performance was then tested against standard convolutional neural network architectures on three specific cardiac diagnostic areas.
“The three tasks we tested the model on were learning if a patient is having a heart attack, if they have a genetic disorder called hypertrophic cardiomyopathy, and how effectively their heart is functioning,” said Dr. Akhil Vaid, instructor of data-driven and digital medicine at the Icahn School of Medicine at Mount Sinai. “In each case, our model performed better than all other tested baselines.”
The study found that HeartBEiT had notably more accurate and explainable performance at smaller sample sizes.
“These representations may be considered individual words, and the whole ECG a single document,” said Vaid. “HeartBEiT understands the relationships between these representations and uses this understanding to perform downstream diagnostic tasks more effectively.”
THE LARGER TREND
Mount Sinai notes that more 100 million electrocardiograms are performed each year in the United States – but that ECGs’ usefulness is limited since it’s difficult for physicians to identify by sight certain patterns representative of disease.
Like in so many other clinical use cases, AI could be transformative in this area.
“We want to be clear that artificial intelligence is by no means replacing diagnosis by professionals from ECGs,” said Dr. Girish Nadkarni, director of the Charles Bronfman Institute of Personalized Medicine at Mount Sinai, “but rather augmenting the ability of that medium in an exciting and compelling new way to detect heart problems and monitor the heart’s health.”
ON THE RECORD
“Neural networks are considered black boxes, but our model was much more specific in highlighting the region of the ECG responsible for a diagnosis, such as a heart attack, which helps clinicians to better understand the underlying pathology,” said Nadkarni. “By comparison, the CNN explanations were vague even when they correctly identified a diagnosis.”
“Our model consistently outperformed CNNs, which are commonly used machine learning algorithms for computer vision tasks,” said Vaid.
“Such CNNs are often pretrained on publicly available images of real-world objects,” he added. “Because HeartBEiT is specialized to ECGs, it can perform as well as, if not better than, these methods using a tenth of the data. This makes ECG-based diagnosis considerably more viable, especially for rare conditions which affect fewer patients and therefore have limited data available.”
Mike Miliard is executive editor of Healthcare IT News
Email the writer: [email protected]
Healthcare IT News is a HIMSS publication.
Source: Read Full Article