Getting poked with a needle multiple times at the doctor’s office could become a thing of the past, thanks to an AI-enabled ultrasound imaging system that’s being developed by Fujifilm SonoSite and the Allen Institute for Artificial Intelligence.
The collaboration, announced today, aims to provide better interpretations of ultrasound images with AI, opening the way for new ultrasound applications and enhanced accuracy.
One of the first applications would be to identify blood vessels under the skin, said Vu Ha, technical director at the AI2 Incubator.
“We train deep learning models on ultrasound images, where veins and arteries have been carefully labeled by sonographers,” Ha explained in an email to GeekWire. “When deployed, these trained models would detect vessels and visualize them on the ultrasound screen in real time. Clinicians would then use these visualizations to locate veins with much higher confidence, minimizing the need to poke patients multiple times.”
Ha said visualizing the arteries would show clinicians which spots on the body to avoid when they’re sticking in a hypodermic needle.
Finding the sweet spots for intravenous procedures isn’t a trivial achievement.
“Annually, 20% of the population in the U.S. goes through the painful experience of getting poked three or more times during IV procedures, due to the fact that they have hard-to-find veins,” Ha said. “With an AI-assisted vein detection ultrasound device, this problem can potentially be solved for many millions of patients.”
The collaboration demonstrates how Pacific Northwest connections can pay off: Fujifilm SonoSite, a subsidiary of Japan’s Fujifilm that’s headquartered in Bothell, Wash., reached out to the Seattle-based AI2 Incubator for advice on improving their compact ultrasound imaging systems.
“The AI2 Incubator was a perfect place to look for help in creating breakthrough technology,” Rich Fabian, SonoSite’s president and chief operating officer, said in a news release. “They have the type of talent that is hard to recruit, combined with the hunger of a startup. We look forward to collaborating more.”
AI2 plans to apply deep learning and computer vision to a wider range of ultrasound scenarios. “The plan is to incorporate this technology in multiple products at SonoSite,” Ha said.
Ultrasound imaging is significantly more affordable and portable than X-ray imaging, CT scans or PET scans, with none of the downside associated with radiation exposure. “Ultrasound’s comparative disadvantage is its lower image quality, which we aim to address with the use of deep learning,” Ha said.
Artificial intelligence has already been applied to image interpretation for the diagnosis of maladies ranging from malaria to early-stage lung cancer, breast cancer and cervical cancer. AI2 and SonoSite aim to make similar advances in the ultrasound realm.
“The combination of deep learning and medical imaging is very exciting for the future of detection,” said Diku Mandavia, senior vice president and chief medical officer of Fujifilm SonoSite. “Better care and catching anomalies earlier and faster is a core mission.”