I N T E L L I G E N T T E C H

The study, conducted by Tufts Medical Centre and published in European Heart Journal – Imaging Methods and Practice, shows that an ML model finetuned for use on Butterfly iQ + devices can achieve high accuracy in identifying AS. The findings support the value of ML model development and represent a positive step toward portable screening for earlier detection of this life-threatening condition.

Butterfly Network pitches itself as driving a digital revolution in medical imaging with its proprietary Ultrasoundon-Chip semiconductor technology and ultrasound software solutions.

Aortic stenosis, a narrowing of the aortic valve, affects more than 13 % of Americans over 75-years-old and is often missed until it is advanced and symptomatic. Studies suggest that a significant number of AS cases remain undiagnosed, particularly among underserved and ageing populations. This delay in diagnosis is associated with higher-risk procedures, worse outcomes and increased healthcare costs. As earlier identification and treatments of AS are increasingly linked with better patient outcomes, the need for upstream accessible screening is critical.

“ This research shows a promising path forward where lower-level providers, not just cardiologists or trained sonographers, could screen for aortic stenosis using AI-assisted handheld ultrasound,” said Dr John Martin, Coauthor of the study and Butterfly’ s Chief Medical Officer Emeritus.“ This opens the door to early detection in a wide variety of care venues including primary care offices, long-term care facilities, urgent care facilities and even in the home.”

The study validated that an ML model trained on hospital-grade ultrasound images performed well when adapted to Butterfly’ s handheld ultrasound device. After fine-tuning the final layer of a neural network, researchers achieved an area under the receiver operator characteristic curve( AUROC) of 0.94 for differentiating between no aortic stenosis and any degree of aortic stenosis on handheld ultrasound images – a level of accuracy that supports potential clinical use in screening workflows.

In a related development, Butterfly has released the latest expansion of its Butterfly ScanLab AI-powered ultrasound education app: an Aorta Exam Protocol.

The app uses animations, anatomical labelling and a quality indicator to help teach users how to scan the abdominal aorta and recognise normal anatomy.

With that training, users can then use POCUS( Point-of-Care Ultrasound) to detect life-threatening conditions such as abdominal aortic aneurysms and dissections.

Designed for use on an iPad, the module is included in every Butterfly membership and compatible with existing Butterfly probes.

“ Early detection saves lives but access to imaging remains a major barrier. This work is about democratising diagnostics. With AI and education working hand-inhand, we’ re empowering more providers to deliver proactive care in places that were never possible before,” said Joseph DeVivo, President, CEO and Chairman of Butterfly Network. �

www. intelligenthealth. tech 29