Currently Viewing:
Newsroom

AI Accurately and Efficiently Improves Breast Cancer Detecting Technology, Study Shows

Matthew Gavidia
Artificial intelligence can be used to ameliorate reading time for digital breast tomosynthesis while maintaining or improving the accuracy of the reading, according to a new study.
Artificial intelligence (AI) can improve breast cancer detection efficacy in the imaging tool digital breast tomosynthesis (DBT), according to a new study.

DBT is an advanced imaging technology that enhances cancer detection and reduces false-positive recalls found in screening solely with digital mammography (DM). While DBT has been shown to improve breast cancer detection, it can take nearly twice as long to interpret as DM. This lapse of time can prove inefficient for radiologists as it increasingly becomes the standard of mammographic imaging.

The study, published in the journal Radiology: Artificial Intelligence, evaluated the use of AI to ameliorate DBT reading time while maintaining or improving accuracy. Researchers compared the performance of 24 Mammography Quality Standards Act–qualified radiologists, 13 of whom were distinguished as breast subspecialists, on reading 260 DBT examinations both with and without AI:
  • Readings included 65 cancer cases with 66 malignant lesions and 65 biopsy-proven benign cases.
  • Readings occurred in 2 sessions separated by a memory washout period of at least 4 weeks to minimize recall bias.
  • Readers reviewed 130 cases without AI and another 130 cases with AI during one session and complimentary cases during a second session to evaluate each radiologist’s performance in reading each case both with and without AI.
A deep learning AI system was developed and trained by the researchers to identify suspicious soft tissue and calcified lesions in DBT images. This type of AI can mine vast amounts of data to find subtle patterns beyond human recognition. Researchers evaluated the area under the receiver operating characteristic curve (AUC), reading time, sensitivity, specificity, and recall rate with statistical methods for a multireader, multicase study of the DBT AI system.

AI use represented an overall statistically significant improvement in accuracy and shorter reading times. Sensitivity increased from 77% without AI to 85% with it, specificity increased from 62.7% without to 69.6% with AI, recall rate for non-cancers decreased from 38% without to 30.9% with AI, detection of malignant lesions (measured with AUC) increased from 0.795 without AI to 0.852 with, and reading time decreased from 64.1 seconds without AI to 30.4 seconds with:
  • Sensitivity improvement with AI was 8% (95% CI, 2.6%-13.4%; P <.01)
  • Specificity improvement with AI was 6.9% (95% CI, 3.0%-10.8%; noninferiority P <.01)
  • Recall rate decrease with AI was 7.2% (95% CI, 3.1%-11.2%; noninferiority P <.01)
  • Detection of malignant lesions with AI improvement was 0.057 (95% CI, 0.028-0.087; P <.01)
  • Reading time decrease with AI was 52.7% (95% CI, 41.8%-61.5%; P <.01)
Lead study author Emily F. Conant, MD, professor and chief of breast imaging from the Department of Radiology at the Perelman School of Medicine at the University of Pennsylvania, highlighted the stark improvements of AI based off the study.

“The concurrent use of AI with DBT increases cancer detection and may bring reading times back to about the time it takes to read DM-alone exams,” she said. This significant decrease in reading time eliminates inefficiency worries correlated to DBT. The performance of radiologists, measured by mean AUC, improved as well through the increased detection of malignant lesions by 0.057 with AI.

“The results of this study suggest that both improved efficiency and accuracy could be achieved in clinical practice using an effective AI system," Conant said. Screening with DBT already showcased a more accurate reading than that of digital mammograms but was burdened by its extensive reading time. Concurrent use of AI eliminated this inefficiency while sparking innovation in breast cancer detection.

Reference

Conant EF, Toledano AY, Periaswamy S, et al. Improving accuracy and efficiency with concurrent use of artificial intelligence for digital breast tomosynthesis [published online July 31, 2019]. Radiology: Artificial Intelligence. doi: 10.1148/ryai.2019180096.

 
Copyright AJMC 2006-2019 Clinical Care Targeted Communications Group, LLC. All Rights Reserved.
x
Welcome the the new and improved AJMC.com, the premier managed market network. Tell us about yourself so that we can serve you better.
Sign Up