AI Echo in Clinical Workflow
Well, AI will not replace a physician, but a physician using AI will replace a physician not using AI!
Wojciech Mazur, MD
The Christ Hospital
In this interview, Dr. Wojciech Mazur, MD, Director of Advanced Cardiac Imaging at The Christ Hospital Health Network, and Dave Ladd, Business Development at Us2.ai, discuss the integration of AI-driven echocardiography into clinical practice. They explore how AI enhances efficiency, reduces variability, and improves workflow automation in cardiac imaging. Dr. Mazur shares insights from his personal experience working with the Us2.ai software and how AI is changing echocardiography.
Read the key highlights from the interview below.
Q: What are the biggest challenges in echocardiography today, and how can AI help address them?
Dr. Mazur: This is a very loaded question because there is a list of challenges:
- Increasing Complexity of Studies – Echocardiography has evolved significantly over the past 10-15 years, with additional assessments like LV strain and advanced valvular evaluations extending study durations. A full echocardiogram can now take up to 90 minutes, leading to longer patient wait times (4-6 weeks).
- Sonographer Shortage – A worldwide shortage of trained sonographers forces institutions to rely on temporary staff (“traveling sonographers”) unfamiliar with institution’s standardized protocols, potentially affecting imaging quality.
- Sonographer Fatigue and Diagnostic Variability – Sonographers face both physical and mental fatigue, balancing image acquisition and measurement accuracy under time pressure. Physicians reviewing 50-70 echocardiograms daily also experience declining qualitative interpretation as the day progresses, increasing the risk of errors.
AI has the potential to address these challenges by streamlining workflows, reducing workload burden, and improving diagnostic consistency.
Q: What are your experiences with AI in cardiac imaging and how has that been evolving lately?
Dr. Mazur: There are a few ways which AI has helped to enhance echo workflows.
- Reducing Measurement Burden for Sonographers – Sonographers typically spend 15-20 minutes per study performing manual measurements, often prioritizing speed over image quality to complete protocols within time constraints. With Us2.ai, AI automates echo measurements, allowing sonographers to focus on acquiring higher-quality images.
- Improving Reproducibility & Reducing Variability – The ongoing sonographer shortage has increased reliance on traveling sonographers, each trained under different protocols. This often results in inconsistent measurements for the same echocardiogram. AI standardizes and automates measurements, ensuring high reproducibility and reducing intra- and inter-observer variability. Us2.ai ensures that Physicians still have the ability to review and adjust measurements accordingly.
AI is an invaluable support tool for both sonographers and cardiologists. It enhances workflow efficiency, reduces fatigue, and improves diagnostic accuracy. By streamlining imaging and ensuring measurement consistency, AI ultimately supports better patient care and clinical decision-making.
Q: What led you to consider incorporating AI into your workflow?
Dr. Mazur: The biggest drivers were patient complaints about long wait times and administrative pressure to complete echocardiograms within an hour. The challenge was how to improve efficiency without compromising quality. As a physician, my priority is always outcomes and accuracy, and I didn’t see a way to significantly reduce study time without affecting quality—until we introduced AI. With AI, we were able to save 15-20 minutes of sonographer time per study without sacrificing image acquisition quality, whilst improving measurement quality.
AI doesn’t just enhance image acquisition and interpretation—it also helps improve the overall productivity of our imaging section. By streamlining processes and reducing variability, AI ultimately supports better patient care and operational efficiency.
Q: What does it take to successfully implement AI into your clinical workflow?
Dave: Implementing AI is a team effort. It’s crucial to have buy-in from cardiologists, sonographers, and service line directors—everyone needs to be aligned on how AI can enhance workflow and solve key challenges. Openness to AI is also important, as it’s a hot topic in healthcare. The willingness to integrate AI into daily workflow and use it effectively is key. Us2.ai provides training sessions to ensure that all users are comfortable with the software and can maximize its benefits.
Dr. Mazur: Initially, we had a technical challenge because our institution uses Merge, which at the time did not fully support Us2.ai. This meant that while AI automated the measurements, we had to manually enter them. However, Merge has since integrated Us2.ai seamlessly, resolving this issue.
Another key challenge was trust in AI—both from physicians and sonographers. Some sonographers were concerned that AI might replace their roles, but I see that as a misconception—AI is not here to replace sonographers but to make their jobs easier by reducing stress and improving workflow. AI allows sonographers to focus on acquiring high-quality images, making their work both physically and mentally less demanding, and more enjoyable!
Q: At the recent AHA meeting, a research study—the AI-ECHO RCT—was presented. Can you share some insights on its findings and how it highlights the benefits of AI in echocardiography? Do you see similar benefits at your institution?
Dr. Mazur: Absolutely! The AI-ECHO RCT was a well-designed prospective study aimed at assessing the non-inferiority of AI-generated measurements compared to human technologists. The study evaluated several key factors, including:
- Echo performance time with and without AI
- Sonographer satisfaction
- Image quality
- Number of measurements analyzed
One of the most important findings was the significant improvement in image quality—which is something we’ve also observed in our own institution. Without the pressure to rush through a study and manually complete measurements, sonographers had more time to focus on acquiring optimal images.
Another major takeaway was the number of measurements analyzed. AI was able to generate 3.3 times more measurements than human technologists, and in a fraction of the time. On average, AI provided nearly 80 measurements per echocardiogram—a truly remarkable result that we’ve also seen replicated in our own clinical practice.
Q: Can you elaborate on how AI is improving workflow at your institution? It sounds like AI is automating tedious manual tasks, allowing staff to focus on areas where they can add more value to the exam.
Dr. Mazur: That’s absolutely correct. AI helps eliminate up to 20 minutes of repetitive, tedious work, reducing mental fatigue and increasing sonographer job satisfaction. At the same time, it improves diagnostic accuracy in echocardiography.
A great example is Aortic Stenosis grading, which Us2.ai is approved for. One of the biggest challenges we’ve faced has been LVOT (left ventricular outflow tract) measurements—there’s often significant variability between sonographers, and this inconsistency tends to increase as sonographers become fatigued. AI, on the other hand, doesn’t get tired. It ensures consistent, reproducible measurements, allowing us to track a patient’s progression with confidence.
Us2.ai also has a pattern recognition feature for Cardiac Amyloidosis detection. The study presented at EuroEcho last year demonstrated that using a single-frame four-chamber view, AI can detect Cardiac Amyloidosis with decent sensitivity and specificity. So, in the event that I don’t completely agree with the AI’s diagnosis, it at least alerts me to the possibility, making me less likely to miss it. In the case of Cardiac Amyloidosis, a missed diagnosis could have fatal consequences for the patient.
Q: Where do you see AI making the most impact in echocardiography in the near future?
Dr. Mazur: Well, AI will not replace a physician, but a physician using AI will replace a physician not using AI!
Looking ahead, I see continuous machine learning and pattern recognition playing a key role in several areas for disease detection:
- Heart Failure with Preserved Ejection Fraction (HFpEF): This condition is reaching epidemic levels in the U.S. and worldwide. AI can help with early diagnosis, leading to earlier treatment and better patient outcomes.
- Cardiac Amyloidosis Detection: Us2.ai is already making strides here, using single-frame four-chamber views to detect early signs of Amyloidosis.
- Valvular Heart Disease: Us2.ai has done a phenomenal job with Aortic Stenosis and Mitral Regurgitation. AI should also focus more on Tricuspid Valve Regurgitation, especially now that new treatment options are available.
AI is also revolutionizing retrospective research. I’m currently involved in a study analyzing 3,000 echocardiograms to assess right ventricular (RV) function and survival rates in cancer patients undergoing chemotherapy. If we were to do this manually, it would require a large team or take years to complete. With AI, we can process the entire dataset within days, allowing us to quickly generate insights and start writing manuscripts. AI opens entirely new opportunities for research, enabling us to reanalyze decades of stored data with high reproducibility and unprecedented speed!
Q: What advice would you give to sites looking to implement AI in echocardiography?
Dr. Mazur:
- Ensure Compatibility with Your Platform: Before rolling out AI, confirm that your existing system integrates well with Us2.ai or any AI solution you choose. Early on, we faced frustration because our system (Merge) didn’t initially communicate with Us2.ai. This led to manual data entry of AI-generated measurements, which was inefficient. Now, with full integration, this is no longer an issue.
- Build Trust in the AI System: In the first few weeks or the first month, compare AI-generated measurements with your own. Validate its accuracy by double-checking the measurements manually. This will build confidence in the tool’s reliability.
- Communicate with Sonographers: Address concerns early—sonographers may fear that AI is replacing them. Emphasize that AI is a support tool, not a replacement—it improves image quality, workflow efficiency, and measurement accuracy while reducing stress and fatigue.
Engage sonographers in training sessions and discussions to help them understand AI’s role in enhancing their work, not taking it away.
By considering these points, I think sites can ensure a smooth AI adoption process, reduce resistance, and maximize the benefits of AI-assisted echocardiography.
