GE HealthCare Develops Healthcare-Specific Foundation Models with NVIDIA Tech

GE HealthCare recently used NVIDIA technology to develop its recent research model, SonoSAMTrack.
 
SonoSAMTrack combines a promptable foundation model to segment objects on ultrasounds called SonoSAM. The technology segments anatomies, lesions, and other essential places in ultrasound images. There’s also a streamlined version called SonoSAMLite.
 
Foundation models have recently risen in prominence because they can operate as human-in-the-loop AI systems. Foundation and generative AI models could help enable quick adaptation to many diseases, facilitating screening, early detection, tracking progression, and identifying non-invasive biomarkers with minimal training requirements, such as zero-shot or few-shot settings.
 
GE HealthCare conducted a recent study showing SonoSAMTrack had high performance in seven ultrasound datasets. It covered a wide range of anatomies (adult heart, fetal head) and pathologies (breast lesions and musculoskeletal pathologies), as well as other scanning devices. It notable outperformed competing methods by a significant margin.
 
Further, SonoSAMTrack required only two to six clicks for precise segmentation. This was enabled through distillation and quantization techniques, using the NVIDIA TensorRT software development kit and other capabilities for quantization-aware training.
 
“Combining NVIDIA’s accelerated computing and AI technology stack with GE HealthCare’s medical imaging expertise will help enhance patient care by making ultrasound diagnostics quicker and more accurate,” said David Niewolny, NVIDIA’s director of business development for healthcare and medical. “This collaboration underscores the importance of using AI for life-saving advancements and setting new standards in healthcare.”
 
“GE HealthCare is committed to investing in innovative technologies that help tackle some of the industry’s biggest challenges.  Our vision is to accelerate advancements in medical imaging by introducing foundational AI technologies, thereby empowering data scientists to expedite AI application development and eventually help clinicians and enhance patient care. By utilizing these versatile, generalist models, we aim to adapt more efficiently to new tasks and medical imaging modalities, often requiring far less labeled data compared to the traditional model retraining approach. This is particularly significant in the healthcare domain, for which data is especially time-consuming and costly to obtain,” added Parminder Bhatia, chief AI officer of GE HealthCare.