TY - JOUR
T1 - Artificial intelligence system reduces false-positive findings in the interpretation of breast ultrasound exams
AU - Shen, Yiqiu
AU - Shamout, Farah E.
AU - Oliver, Jamie R.
AU - Witowski, Jan
AU - Kannan, Kawshik
AU - Park, Jungkyu
AU - Wu, Nan
AU - Huddleston, Connor
AU - Wolfson, Stacey
AU - Millet, Alexandra
AU - Ehrenpreis, Robin
AU - Awal, Divya
AU - Tyma, Cathy
AU - Samreen, Naziya
AU - Gao, Yiming
AU - Chhor, Chloe
AU - Gandhi, Stacey
AU - Lee, Cindy
AU - Kumari-Subaiya, Sheila
AU - Leonard, Cindy
AU - Mohammed, Reyhan
AU - Moczulski, Christopher
AU - Altabet, Jaime
AU - Babb, James
AU - Lewin, Alana
AU - Reig, Beatriu
AU - Moy, Linda
AU - Heacock, Laura
AU - Geras, Krzysztof J.
N1 - Funding Information:
The authors would like to thank Mario Videna, Abdul Khaja and Michael Costantino for supporting our computing environment, Benny Huang and Marc Parente for extracting the data, Yizhuo Ma for providing graphical design consultation, and Catriona C. Geras for proofreading the manuscript. We also gratefully acknowledge the support of Nvidia Corporation with the donation of some of the GPUs used in this research. This work was supported in part by grants from the National Institutes of Health (P41EB017183, R21CA225175), the National Science Foundation (1922658), the Gordon and Betty Moore Foundation (9683), the Polish National Agency for Academic Exchange (PPN/ IWA/2019/1/00114/U/00001) and NYU Abu Dhabi.
Publisher Copyright:
© 2021, The Author(s).
PY - 2021/12
Y1 - 2021/12
N2 - Though consistently shown to detect mammographically occult cancers, breast ultrasound has been noted to have high false-positive rates. In this work, we present an AI system that achieves radiologist-level accuracy in identifying breast cancer in ultrasound images. Developed on 288,767 exams, consisting of 5,442,907 B-mode and Color Doppler images, the AI achieves an area under the receiver operating characteristic curve (AUROC) of 0.976 on a test set consisting of 44,755 exams. In a retrospective reader study, the AI achieves a higher AUROC than the average of ten board-certified breast radiologists (AUROC: 0.962 AI, 0.924 ± 0.02 radiologists). With the help of the AI, radiologists decrease their false positive rates by 37.3% and reduce requested biopsies by 27.8%, while maintaining the same level of sensitivity. This highlights the potential of AI in improving the accuracy, consistency, and efficiency of breast ultrasound diagnosis.
AB - Though consistently shown to detect mammographically occult cancers, breast ultrasound has been noted to have high false-positive rates. In this work, we present an AI system that achieves radiologist-level accuracy in identifying breast cancer in ultrasound images. Developed on 288,767 exams, consisting of 5,442,907 B-mode and Color Doppler images, the AI achieves an area under the receiver operating characteristic curve (AUROC) of 0.976 on a test set consisting of 44,755 exams. In a retrospective reader study, the AI achieves a higher AUROC than the average of ten board-certified breast radiologists (AUROC: 0.962 AI, 0.924 ± 0.02 radiologists). With the help of the AI, radiologists decrease their false positive rates by 37.3% and reduce requested biopsies by 27.8%, while maintaining the same level of sensitivity. This highlights the potential of AI in improving the accuracy, consistency, and efficiency of breast ultrasound diagnosis.
UR - http://www.scopus.com/inward/record.url?scp=85115675338&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85115675338&partnerID=8YFLogxK
U2 - 10.1038/s41467-021-26023-2
DO - 10.1038/s41467-021-26023-2
M3 - Article
C2 - 34561440
AN - SCOPUS:85115675338
VL - 12
JO - Nature Communications
JF - Nature Communications
SN - 2041-1723
IS - 1
M1 - 5645
ER -