New AI Tool Revolutionizes Diagnosis of Severe Aortic Stenosis
Researchers at the Cardiovascular Data Science (CarDS) Lab have made a groundbreaking discovery that could transform the diagnosis and treatment of a common heart condition called severe aortic stenosis (AS). By harnessing the power of artificial intelligence (AI), the team has developed an innovative approach that can detect AS from regular ultrasound scans of the heart. This breakthrough, published in the prestigious European Heart Journal, has significant implications for routine clinical care.
Severe aortic stenosis is a health condition that primarily affects older adults. It is caused by the narrowing of the aortic valve, which leads to restricted blood flow from the heart to the rest of the body. Early diagnosis is crucial as it allows for timely interventions to alleviate symptoms and reduce the risk of hospitalization and premature death.
Currently, specialized ultrasound imaging known as doppler echocardiography is the main method used to detect AS. However, the team at CarDS Lab has developed a deep learning model capable of automatically identifying severe AS using simpler heart ultrasound scans. This technology represents a significant breakthrough in the field of cardiovascular medicine.
Led by Rohan Khera, MD, MS, an accomplished assistant professor of cardiovascular medicine and health informatics and director of the CarDS Lab, the team collaborated with experts from UT Austin’s Chandra Family Department of Electrical and Computer Engineering to develop this cutting-edge AI model. They trained the model using 5,257 studies that comprised a staggering 17,570 videos recorded between 2016 and 2020 at Yale New Haven Hospital. The model’s accuracy was further validated using an additional 2,040 studies from diverse cohorts in New England and California.
Dr. Khera explained the motivation behind their work, stating, Our challenge is to accurately evaluate AS for effective patient management and risk reduction. While specialized testing remains the gold standard, relying solely on patients who visit our echocardiographic laboratories may miss individuals in the early stages of the disease. To address this, the team aimed to create a machine learning approach suitable for point-of-care ultrasound screening.
One of the study’s co-first authors, Evangelos Oikonomou, MD, DPhil, a cardiology fellow and postdoctoral researcher in the CarDS Lab, emphasized the significance of their work. Our research enables early detection of aortic stenosis, ensuring patients receive timely care. Our approach facilitates broader community screening by utilizing handheld ultrasounds, which are increasingly being used in emergency departments and various other care settings, he explained.
What sets this achievement apart is the close collaboration between clinicians and computer scientists. Greg Holste, a Ph.D. student at UT Austin and co-advised by Dr. Khera, played a vital role in developing the innovative methodology that made this technology possible. Dr. Khera stressed the importance of multidisciplinary collaboration in leveraging emerging technology to improve clinical care.
This discovery not only revolutionizes the diagnosis of aortic stenosis but also highlights the immense potential of AI in healthcare. By utilizing machine learning and advanced algorithms, medical professionals can make more accurate and timely diagnoses, leading to better patient outcomes.
As the field of AI continues to advance, it is clear that the integration of technology into healthcare practices has the potential to positively impact the lives of millions. The development of AI tools like this one brings us one step closer to a future where cutting-edge technology works hand in hand with medical expertise to deliver the best care possible.
