Can Artificial Intelligence Improve Accuracy of Digital Breast Tomosynthesis?
Posted: Tuesday, October 8, 2019
According to a study published in Radiology: Artificial Intelligence, using digital breast tomosynthesis with an artificial intelligence (AI) system was found to improve the efficiency and accuracy of breast cancer detection and allow for shorter image reading times. The study was conducted by Emily F. Conant, MD, of the Perelman School of Medicine at the University of Pennsylvania in Philadelphia, and colleagues.
“We know that digital breast tomosynthesis imaging increases cancer detection and lowers recall rate when added to two-dimensional mammography,” said Dr. Conant in a press release from the Radiological Society of North America. “Since adding digital breast tomosynthesis to the two-dimensional mammogram approximately doubles radiologist reading time, the current use of AI with digital breast tomosynthesis increases cancer detection and may bring reading times back to the about the time it takes to read digital mammography–alone exams.”
Researchers developed a deep learning AI system and trained it on digital breast tomosynthesis images to identify suspicious soft-tissue and calcified lesions. After developing the system, researchers tested its performance with 24 radiologists, including 13 breast subspecialists, to read 260 digital breast tomosynthesis examinations with and without AI assistance.
Radiologist performance for the detection of malignant lesions measured by mean AUC increased to 0.852 with the use of AI compared with 0.795 without. Reading time decreased to 30.4 seconds with AI compared with 64.1 seconds without it. Sensitivity increased to 85% with AI compared with 77% without it. Specificity increased to 69.6% with AI compared with 62.7% without it. Recall rate for noncancers decreased to 30.9% with AI compared with 38.0% without it.
The deep learning approach is expected to improve as it is exposed to more images, according to the researchers. “The results of this study suggest that both improved efficiency and accuracy could be achieved in clinical practice using an effective AI system,” concluded Dr Conant.
Disclosure: The study authors’ disclosure information can be found at pubs.rsna.org.
