Trends in Cardiovascular AI Device Development

This is wild. A simple ECG, transformed by AI, could become a powerful tool for early detection of structural heart disease. A recent study in Nature Portfolio introduces Echonext, an AI tool that analyzes ECGs to identify patients who may need an echocardiogram. By using a common, low-cost test to guide imaging decisions, it helps surface high-risk patients who might otherwise be missed. This technology improves performance and imaging stewardship, moving towards higher value for patients and systems. Appropriate use of echocardiograms has long been a challenge. Some patients don’t get the imaging they need, while others undergo unnecessary exams. Compared with standard of care, the AI achieved higher accuracy, and when applied retrospectively, it flagged thousands of high-risk patients who had been overlooked. This once again shows that human plus machine can be better than human alone. As a radiologist, I see the power of AI to help guide appropriate evaluation and use of imaging tools as truly transformational. That represents a new diagnostic opportunity at scale. Smarter referrals. Earlier detection. Improved outcomes. Clinical judgment will always be essential. But AI can sharpen that judgment when the signals are subtle or easy to miss. If we can turn 400 million ECGs each year into 400 million opportunities to detect disease earlier, we are improving care and changing lives. We are following this work closely at Ainsley Advisory Group. 🔍 https://lnkd.in/eXjw7Gvq

⚕️ AI in Cardiology: From Code to Clinic I am excited to share a groundbreaking study recently published in Nature Magazine, led by our colleague at Columbia University Irving Medical Center, Dr. Pierre Elias, MD. This research introduces "EchoNext," a deep learning model designed to detect structural heart disease (SHD) from electrocardiograms (ECGs). SHD encompasses pathologies that affect the heart's valves, walls, or chambers , and it is a growing epidemic that remains substantially underdiagnosed. Early detection is key, but widespread screening is limited by the cost and accessibility of definitive diagnostic tools like echocardiography. 🔬 The research team developed EchoNext by training it on an extensive and diverse dataset of over 1.2 million ECG-echocardiogram pairs from more than 230,000 patients. The model demonstrated high diagnostic accuracy in both internal and external validation (AUROC was 85.2% in the internal test set) 🧠 In a direct comparison based on 150 ECGs, the EchoNext model alone was more accurate at detecting SHD from an ECG than the 13 participating cardiologists. The AI model alone achieved 77.3% accuracy. This outperformed both cardiologists working without AI assistance (64.0% accuracy) and cardiologists who were assisted by the AI's prediction (69.2% accuracy). I think this demonstrates the immense potential for AI to augment our diagnostic capabilities, using a test that is already ubiquitous in medicine. 🩺 What I find particularly compelling is that this research didn't stop at retrospective validation. The team conducted a 100-patient prospective clinical trial, called DISCOVERY, to test an AI-ECG model's ability to find previously undiagnosed heart disease in a real-world setting. The trial successfully identified patients (n=37) with previously unknown, clinically significant SHD. 🌐 To foster further innovation and transparency, the authors have publicly released model weights and a large, annotated dataset of 100,000 ECGs from 36,286 unique patients, which is a commendable move to advance the entire field. Link to the paper: https://lnkd.in/ed7KT52g

"Advanced artificial intelligence (AI) models can evaluate cardiovascular risk in routine chest computed tomography (CT) scans without contrast, according to new research published in Nature Communications.[1] In fact, the authors noted, the AI approach may be more effective at identifying issues than relying on guidance from radiologists. A team of cardiac imaging specialists with Cedars-Sinai Medical Center used two separate AI algorithms—one that measures coronary artery calcium (CAC) scores and another that segments cardiac chamber volumes—to examine low-dose chest CT results from nearly 30,000 patients with a median age of 61 years old. All CT exams were originally performed for lung cancer screening. The AI models delivered CAC scores and chamber volumes in just a matter of seconds, and they only failed to segment 0.1% of cases. While 19% of patients had a CAC score of zero, 37% had a CAC score of 1-100, 19.8% a score of 101-400 and 24.3% had a score of more than 400." https://lnkd.in/gT-xkxBr

We Have the 🛠️ Tools. The Potential 💡 Is Clear. Let’s Rethink ❤️🩹Cardiovascular Care ❤️🩹at Scale. A compelling review by Aline Pedroso, PhD and Rohan Khera in Nature Portfolio’s Cardiovascular Health. Great outline on how AI-powered wearables, PPG/ECG sensors, point-of-care ultrasound, and edge-AI models can and are transforming cardiovascular care—extending reach, reducing friction, and bringing precision to the front lines. 👉 Article: https://lnkd.in/eCNVj8_F Why this matters: ✅Community-based detection of arrhythmias and structural heart disease is feasible now. ✅Multimodal sensor + AI fusion improves prediction, risk stratification, and monitoring. ✅Cloud and edge tech enable privacy-preserving integration into clinical workflows. ✅Tools like AI-guided echocardiograms with GE HealthCare’s Caption Guidance (FDA-cleared for use by any medical professional) allow earlier, scalable echo screenings—no sonographer required. ✅These shifts are especially powerful in under-resourced or preventive care settings. Call to action for Health Systems, Payers, MedTech and Innovators: 1️⃣ Advance interoperability—connect consumer and bedside data with clinician workflows. 2️⃣ Fund pragmatic RCTs to validate outcomes, not just signal accuracy. 3️⃣ Build reimbursement models that reward early detection and smarter triage. 4️⃣ Design inclusively—this must close gaps, not widen them. 💡 We’re past proof of concept and evolve the platform. Time to implement boldly, equitably, and at scale. #DigitalHealth #AIinHealthcare #CardiovascularCare #HealthEquity #Wearables

Earlier this year, I witnessed how AI and machine learning can enhance patient care in cardiology in practical, impactful ways.   A speaker at the AI Cures conference at MIT shared how ML can be applied to data from minimally invasive home monitoring devices like ECGs.   A patient’s hemodynamic measures are incredibly useful in monitoring a patient, however given the equipment involved, can only be done in the hospital. With this new algorithm that was presented, the model can actually infer a patient's hemodynamic measurements, like pressures, fairly accurately from the ECG waveform data alone. I found that rather amazing. And useful!   This means patients could be monitored closely at home, with the ML model providing cardiologists with clinical indicators like pressure risks they wouldn't otherwise have without bringing the patient in.   Examples like this, where ML provides incremental advantages and empowers clinicians, excite me most about AI in healthcare.   The technology is maturing to the point where we can apply it to increase access to care, fill in gaps, and connect disparate data sources - rather than pursue AI applications for their own sake.   What other opportunities exist where AI/ML could provide an extra layer of insight to improve clinicians' abilities? I'd love to hear your ideas! #AI #artificialintelligence #codetocare

🔉 Now in Nature Portfolio's Nature Cardiovascular Research, our study reports the development of a novel #DeepLearning #AI model that detects Hypertrophic Cardiomyopathy (HCM) from photos of ECGs. HCM : 🏥 Is a leading cause of sudden cardiac death in young adults 🪢 Has a long asymptomatic course 📽️ Is Infeasible to screen as requires cardiac imaging for dx Our approach detects HCM: 📸 Using photos/images of electrocardiograms ⭐ Across layouts of images 🇺🇸 🇬🇧 🇳🇱 Validation across 3 multinational sites, with AUC >0.9 at all 👁️🗨️ Identifies patterns of LV thickness in test POSITIVES vs NEGATIVES Read more: https://lnkd.in/eY8qGaR5 Research tool at: https://lnkd.in/e7ytvrVN We are working to improve targeted screening approaches to drive up positive predictive value further Led by star members of the Cardiovascular Data Science (CarDS) Lab, Veer Sangha & Lovedeep Dhingra, along with Evangelos K. Oikonomou, Philip Croon & Arya Aminorroaya MD, MPH, and our collaborators, Harlan Krumholz, Folkert Asselbergs, Martin S. Maron, MD, & Matthew Martinez

We dove into the FDA's healthcare AI data in the latest Cardiac Wire issue, and found some info folks don't seem to be talking about. >>> Everyone keeps talking about the massive number of AI clearances, and how it grew from 692 in July 2023 to 882 in March 2024... However, much this growth due to the fact that products must get re-cleared as their algorithms change. *** There aren't anywhere close to 882 unique/active FDA cleared AI products, many are just different generations of the same products *** >>> Cardiovascular AI maintains a (distant) second largest share of FDA-clearances (10%, 90 clearances), well below radiology’s 76% share (671). >>> Cardiovascular AI actually makes up a larger 17.4% share of total clearances (154) if you also count cardiovascular imaging AI products that the FDA technically categorized within its “Radiology” segment (e.g. FFRCT, coronary plaque, etc.). >>> Even with this broader definition, cardiovascular AI’s total share of AI clearances is declining, falling from roughly 25% of clearances in 2018-2019, to 16.5% in 2020-2022, and 13.5% since the start of 2023. >>> Cardiovascular AI applications also appear to be getting more diverse. Between 2020 and 2022, an overwhelming 86% of all cardiovascular AI clearances were for products that either analyzed imaging or ECG, but imaging and ECG AI’s share of cardiovascular AI clearances fell to 66% in 2023-2024. Check out the rest (and sign up to get the next Cardiac Wire) here :: https://lnkd.in/gB8B_CM2

New from me - the FDA has cleared an AI algorithm that, when used with a digital stethoscope from startup Eko Health, can detect a key risk factor for heart failure. The algorithm was originally developed by the Mayo Clinic for use with electrocardiograms, but as adapted by Eko, it can be used by primary care physicians during routine checkups. Which means that early signs of heart disease might be caught before any symptoms emerge. “These tools are incredibly powerful — they help us screen for conditions for which we have treatments," Mayo Clinic cardiovascular head Paul Friedman told me.