New ultrasound technology developed at Johns Hopkins can distinguish fluid from solid breast masses with near perfect accuracy, an advance that could save patients, especially those with dense breast tissue, from unnecessary follow-up exams, painful procedures and anxiety.
In initial tests with real patients, doctors working with the new method accurately identified masses 96% of the time—they were right just 67% of the time analyzing the same masses with their regular tools.
"This is important because the benefits of ultrasound in breast cancer detection can be limited by the similar appearance of benign fluid masses and solid masses, which can be cancerous," said senior author Muyinatu "Bisi" Bell , a Johns Hopkins biomedical and electrical engineer who specializes in imaging technology. "Our achievement will change the landscape of how breast cancer is diagnosed. Radiologists can be immediately confident in diagnoses. And patients won't be sent for biopsies and invasive procedures when there is more confidence that a mass is nothing to be concerned about."
The federally-funded work is published today in Radiology Advances.
It's recommended every woman over 40 gets a mammogram to detect breast cancer early. But the results can be inconclusive for women with dense breast tissue. Those women are often sent next to get ultrasounds— technology that also has trouble with dense breast tissue.
Ultrasound works by sending sound waves through a probe into the breast. The sound bounces off of structures like masses and is recorded. When it works perfectly, the sound travels directly from the mass right back to the probe. But with dense breast issue, the sound scatters before it reaches the mass, causing "acoustic clutter" in the image. A benign liquid-filled cyst which should appear black in images will often look gray inside, which is how a cancerous growth would look.
The new method changes nothing with ultrasound production, but improves on how the signals are processed. Conventional ultrasound relies on the amplitude of signals, turning high and low signals into blacks, whites or grays. The new method is "coherence-based," meaning the image relies on how similar signals are to neighboring signals.
In addition to providing cleaner images, the new system makes it even easier for radiologists by providing a number score for each mass—only those above a certain threshold are considered worrisome.
"It's really exciting because what we do is take the same ultrasound data, sensed through the same process, but we change the signal processing and do a much better job at interpreting these images," Bell said. "When we combine the visual with a number score, that's when the technology really shows the greatest improvement. It takes away decision fatigue by automating something that would ordinarily require more thought and interpretation."
A study of 132 patients determined that radiologists can correctly identify masses 96% of the time using the new technology, compared to 67% of the time with traditional ultrasound.
"The results of this study are important for our specialty because they suggest that this technique can improve our ability to differentiate between solid masses and certain types of cysts that can mimic solid masses on ultrasound," said co-author Eniola Oluyemi , a diagnostic radiologist at Johns Hopkins Medicine. "This improved diagnostic certainty can lead to fewer false positive results and decrease the need for follow-up and biopsies, thus helping to give our patients increased peace of mind at the time of the initial exam."
Existing artificial intelligence can distinguish between benign and cancerous masses in ultrasound images. The team believes that their innovation, used together with AI, could allow doctors at an initial ultrasound appointment to quickly determine the makeup of a mass, and if it's cancer.
Bell also hopes that the innovation could someday become something people could use at home, as part of a breast self-examination.
"My long-term vision is that as society becomes more self-sufficient and ultrasounds becomes even less expensive than they are today, patients might not have to go to a hospital or specialized clinic—our approach could instead be performed at home," Bell said. "With an inexpensive ultrasound scan, a single number extracted from a coherence-based ultrasound image could tell whether or not a palpable breast lump is something to be concerned about."
Authors, all of Johns Hopkins, include Arunima Sharma; Madhavi Tripathi; Emily B. Ambinder; Lisa A. Mullen; Babita Panigrahi; Joanna Rossi; Nethra Venkatayogi, and Kelly S. Myers.
