By Aine Cryts
Artificial intelligence has been shown to improve cancer detection efficacy among radiologists using digital breast tomosynthesis, according to a study published in Radiology: Artificial Intelligence, a new publication launched by the Radiological Society of North America (RSNA) in early 2019.
Study authors learned that the artificial intelligence platform had a demonstrable impact on reading time (decreased by 53%), sensitivity (increased from 77% without artificial intelligence and 85% with artificial intelligence), specificity (increased from 63% to 70%), and recall rate for non-cancers (decreased from 38% to 31%), among other variables.
The artificial intelligence platform used in the study was ProFound AI for breast tomosynthesis from Nashua, N.H.-based iCAD, Inc. The study compared the performance of 24 radiologists (13 of whom were breast subspecialists) interpreting 260 digital breast tomosynthesis studies (of which 65 were cancer cases with malignant lesions and 65 biopsy-proven benign cases). Included in the study were cases from women between the ages of 26 and 85 who were imaged between June 2012 and October 2017.
According to Senthil Periaswamy, PhD, vice president of research at iCAD, Inc. and a study co-author, “This technology is revolutionizing the way radiologists read mammography and addresses an emerging challenge clinicians are facing as 3D mammography grows in adoption.”
For context, he points out that the growing embrace of digital breast tomosynthesis has meant new challenges for radiologists, due to the extensive amount of data this technology generates—and that can lead to a substantial increase in interpretation time. The resulting fatigue experienced by radiologists can be addressed through the use of artificial intelligence with digital breast tomosynthesis, adds Periaswamy.
In contrast with previous mammography CAD studies, which showed a small increase in sensitivity and a small reduction in specificity, the study published in Radiology: Artificial Intelligence revealed “significant improvements in all of these areas with the use of ProFound AI,” Periaswamy tells AXIS Imaging News. That includes an increase in sensitivity and specificity while cutting interpretation time for radiologists by more than half.
Radiologists are right to be skeptical about the use of artificial intelligence alongside digital breast tomosynthesis, but this new algorithm “is clearly superior in sensitivity and specificity to any other [digital breast tomography] mammography cancer detection program,” he says. Periaswamy adds that iCAD’s artificial intelligence solution to support breast detection in 3D tomosynthesis is the first to receive FDA clearance.
In addition to having a potential impact on overall treatment costs and improvement in outcomes, the use of artificial intelligence alongside digital breast tomosynthesis can improve patient satisfaction due to the lower chance of unnecessary recalls, according to Periaswamy. In addition, fewer false positives could reduce costs associated with unnecessary procedures.
Periaswamy notes that iCAD recently kicked off an exclusive relationship with researchers at The Karolinska Institutet in Stockholm, Sweden. The goal of the research partnership is to co-develop an artificial intelligence platform that leverages information from mammography images and age to identify a woman’s individual short-term risk of developing breast cancer.
Aine Cryts is a contributing writer for AXIS Imaging News.