One in five cases of breast cancer is missed by radiologists. Can artificial intelligence improve that percentage and save lives?
A Triangle radiology practice believes so, and it is adding AI to the firm’s arsenal to improve the odds of detecting breast cancer sooner. (AI refers to a broader idea where machines can execute tasks “intelligently.”)
Wake Radiology UNC REX Healthcare says it is the “first and only” outpatient radiology practice in the region to offer iCAD’s Profound AI, called “cutting-edge” deep-learning, breast cancer detection software which is designed for use with 3D mammography.
The stakes are high. Some 43,000 people a year (including 500 men) die from breast cancer. Further, Cancer.net says more women are diagnosed with breast cancer than any other type of cancer other than skin cancer. In 2020, an estimated 276,480 women in the US alone were expected to be diagnosed with invasive breast cancer, along with 2,620 men.
Developed by the Nashua, N.H.-based medtech firm iCAD, the software is said to improve cancer detection rates by 8 percent and reduce unnecessary patient recall rates by around 7 percent.
“That’s pretty significant,” Dr. Susan Kennedy, Wake Radiology’s director of breast imaging, tells WRAL TechWire. “We want to miss as few cancers as possible.”
Computer-aided detection (CAD) has been around in mammography for more than a decade, Kennedy adds, but up to this point, it had been limited to standard (2D) mammograms. This new platform is the first FDA-cleared AI platform designed specifically for 3D mammograms.
Unlike a 2D mammogram that produces only four images, a typical 3D screening mammogram can create 200 or more images for each patient.
Kennedy compares it to “paging through a book.”
“Instead of just seeing the outside cover of the book, now we can flip through the pages and see inside the book better.”
ProFound AI aids the radiologist in evaluating this large data set by marking areas of potential concern for a more focused review.
“It’s a safety net,” she says.
After a three-month trial, the practice is rolling out the new technology across its 14 locations starting this month.
“Artificial intelligence won’t replace the radiologists,” Kennedy adds. “However, we believe the radiologist using AI tools is able to do a better job than the radiologist alone.”
Breast cancer is the most common type of cancer in women globally, occurring in about one in eight women. Mammography is the most widely used screening tool, but diagnosing cancer from these images can be tricky.
At the other end of the spectrum, 50 percent of all women who undergo screening for a 10-year period will experience a false positive, in which cancer is wrongly suspected.
A false positive can lead to overtreatment with invasive biopsies and unnecessary stress for patients. A false negative can result in delayed detection and treatment.
“All mammograms aren’t the same,” Kennedy warns. “It’s dependent on the person, and how good they are at interpreting that exam.”
The quality of the images is also a factor.
Kennedy compares 2D mammograms to taking a picture with a camera from the 1950s, “like an old Polaroid.”
“It’s just not as good; it’s old technology. The technology has really improved, and we should use it.”
In the wake of COVID-19, Kennedy says she’s noticed a dip in women coming in for their annual mammogram.
In fact, some medical groups were encouraging women to put it off because it felt “risky.”
However, that’s no excuse anymore.
“It’s critical to come in every year for your mammogram,” she says. “Everyone has masks, and we’re social distancing; it’s very, very safe.”
Women should have the opportunity to begin annual screening between the ages of 40 to 44. Women aged 45 and up should be screened annually or every other year, according to guidelines for the American Cancer Society.
Of course, these aren’t hard and fast rules.
“Always come in if there’s any change in your breast or any kind of concern,” Kennedy says. “We don’t want anyone to get a delayed diagnosis.”
You need to be logged in to save this episode to a playlist.