From lighting up “zombie” cells to seeing through wound dressings with AI, here are 11 recent radiology innovations:
- Researchers from the University of Houston and the University of Texas MD Anderson Cancer Center developed an AI-powered software system called MedGaze which mimics radiologists’ eye movements and predicts where a radiologist would likely look next.
- Stanford (Calif.) Medicine researchers developed a non-invasive imaging method to visualize hard-to-find senescent or “zombie” cells.
Senescent cells are alive but do not divide or grow and can contribute to osteoarthritis. Targeting senescent cells through senolytic therapy has been approved by the FDA for the treatment of cancer and other conditions.
- A Stanford Medicine surgical team became the first in the U.S. to use a fluorescent imaging system that identifies leftover cancer cells after a lumpectomy.
Patients receive an injection of fluorescent dye about two hours before surgery. After tumor removal, the operating room is darkened and surgeons use the imaging system to find and remove leftover cancer cells. - Researchers from the University at Buffalo (N.Y.) developed imaging technology capable of seeing through wound dressings.
The system, called mmSkin, uses high-frequency radio waves and an AI algorithm to measure a wound’s moisture content without requiring the removal of gauze dressings.
- Cincinnati Children’s and GE HealthCare launched a research program to drive pediatric-focused projects in MRI, CT, ultrasound and molecular imaging.
The program will also gather feedback on future GE HealthCare products, including PET and single photon emission CT. - Physicians at Philadelphia-based Temple University Health System are integrating spirometry during low-dose CT scans to identify serious lung conditions earlier.
- A team at Chapel Hill, N.C.-based UNC Health developed an MRI model to improve imaging accuracy.
Researchers trained the model, called Brain MRI Enhancement foundation, on more than 13,000 images from diverse populations and scanner types. The model reportedly can perform motion correction, super resolution, noise reduction, harmonization and contrast enhancement.
- Researchers from New York City-based NYU Langone Health evaluated the accuracy and effectiveness of “opportunistic screening” — defined as when radiologists use a patient’s existing medical images for diagnoses beyond the original imaging intent.
Using an AI-enabled measurement to quantify and score aortic calcification levels from 3,662 abdominal CT scans, researchers were able to accurately predict patient risk of a future cardiovascular event.
Similarly, researchers identified signs of bone loss from 3,708 CT scans originally performed as lung cancer screenings. - Houma, La.-based Terrebonne General Health System leveraged AI in its radiology service line for the early detection and treatment of lung nodules.
With the new technology, patients identified as needing additional care will be automatically followed up with after imaging is complete. The health system sees the potential to expand the technology into other areas of disease detection and treatment.
- With funding from a $3.3 million National Cancer Institute award, researchers at Tucson-based University of Arizona are developing a breast imaging method that will not require physical compression.
The method will use CT scan technology to produce a high-resolution image while the patient lies in the prone position. The grant will be used to refine the technology and recruit 600 participants to compare the method to 3D mammography imaging.
- A Boston-based Brigham and Women’s Hospital emergency radiologist developed an AI clinical decision support tool to identify patients at risk for intimate partner violence.
The tool provides a real-time assessment by analyzing radiological data from medical imaging performed in the emergency room alongside the patient’s clinical history at about 80% accuracy.
Bharti Khurana, MD, created the tool with her team, which identifies at-risk patients four years before the patient self-reports with about 80% accuracy. The tool appears in an EMR “safe zone.” It includes conversation guides for providers as well as QR codes to share resources without requiring the patient to take any physical information home.
