AINEWS
An AR Headset Approach to Ultrasound: Promise and Open Questions
Trained sonographers may not need it, but a new MIT AR ultrasound system could help close the gap for novices learning the craft.
Trained sonographers may not need it, but a new MIT AR ultrasound system could help close the gap for novices learning the craft.
Ultrasound is one of the cheapest and most portable imaging tools in medicine, and it is also one of the hardest to read. A probe pressed against a patient returns a flat, two-dimensional slice that takes years of training to interpret with confidence. A team at MIT thinks the solution is to make the image three-dimensional and put it right in front of the operator's eyes.
The researchers built a system called AR-VIU, or augmented real-time volumetric imaging in ultrasound, that streams a live 3D model of scanned tissue into a VR headset, superimposed over the patient's body. Tilt your head and you see the scan from a different angle. In a study published June 10 in Nature Communications Engineering and described by MIT News, the system let novice users perform nearly as well as trained sonographers at identifying objects hidden inside opaque gelatin blocks and marking biopsy-needle paths on tissue phantoms.
That is a striking result for a field where the gap between novice and expert is the difference between a confident diagnosis and a guess. The system compresses ultrasound voxel data and renders it through the Unreal Engine, the same 3D graphics tool used in video games, which is what makes the live overlay possible on consumer headsets.
The study involved 18 participants: nine ultrasound experts, including sonographers and physicians, and nine novices. Each worked through four conditions, including traditional 2D screens, 3D screens, and the AR view. Across the board, AR-VIU improved object identification and localization. The biggest gains came from novices, who with the AR view closed most of the gap to expert performance on a 2D screen.
The result came with a complication the MIT press release does not hide. Many of the trained sonographers in the study still preferred traditional 2D imaging, not because the AR view was wrong, but because it was unfamiliar. Several said they could see real value in AR-VIU for needle placement in biopsies and for visualizing heart-wall motion during echocardiography, the ultrasound imaging of the heart. None suggested they would abandon the tools they already trust.
That tension, novices helped, experts unconvinced, is the most honest read of the research at this stage. Ultrasound training pipelines are short, sonographer shortages are widespread, and a tool that flattens the learning curve has obvious commercial appeal. The MIT probe, slightly smaller than a deck of cards, uses a chirped data acquisition (cDAQ) system and a sparse ultrasound array arranged in a square, which the team says reduces cost and power compared with conventional 3D ultrasound machines. The published paper in Nature Communications Engineering is the primary record for the technical claims about that design; the MIT press release is the source for the user-study numbers and the participant quotes.
This is a research demonstration, not a clinical product. No patient trials are reported. The objects in the study were springs, balls, and screws embedded in gelatin, not tumors in actual tissue. The team lists the next steps as improving resolution and running further validation, and the funding comes from the MIT Media Lab Consortium, the National Science Foundation, and MIT graduate fellowships. None of that should be read as a reason to dismiss the work, but it does set expectations: a controlled phantom study is the start, not the end, and the MIT release does not yet describe patient trials or independent clinical validation.
What to watch next is whether the MIT group, led by associate professor Canan Dagdeviren with grad students Jason Hou and Shrihari Viswanath as lead authors, can run a study with real patients and with sonographers who have had time to adapt to the headset. Independent clinical validation, especially in a workflow that touches biopsies, is the gap between a promising prototype and a tool that changes how a sonographer works.