Related Experiment Video
Updated: Dec 28, 2025
Author Spotlight: Advancing CBCT and Digital Dental Image Integration with AI-Assisted Digitization
Published on: February 23, 2024
Improving Accuracy and Efficiency with Concurrent Use of Artificial Intelligence for Digital Breast Tomosynthesis
Emily F Conant1, Alicia Y Toledano1, Senthil Periaswamy1
1Department of Radiology, Perelman School of Medicine at the University of Pennsylvania, 3400 Spruce St, Philadelphia, PA 19104 (E.F.C.); Biostatistics Consulting, Kensington, Md (A.Y.T.); iCAD, Nashua, NH (S.P., S.V.F., J.G., J.W.H.); and Intrinsic Imaging, Bolton, Mass (J.E.B.).
Artificial intelligence (AI) significantly improved cancer detection accuracy and reduced reading time for digital breast tomosynthesis (DBT) examinations. This AI tool enhanced radiologist performance, leading to better diagnostic outcomes in breast cancer screening.
Area of Science:
- Radiology and Medical Imaging
- Artificial Intelligence in Healthcare
- Breast Cancer Screening Technologies
Background:
- Digital breast tomosynthesis (DBT) is a crucial tool for breast cancer screening.
- Optimizing radiologist reading time and diagnostic accuracy in DBT is an ongoing challenge.
- Artificial intelligence (AI) offers potential solutions for improving efficiency and performance in medical image analysis.
Purpose of the Study:
- To assess the impact of an AI system on the reading time of digital breast tomosynthesis (DBT) images.
- To evaluate whether AI maintains or enhances radiologist accuracy in detecting malignant lesions on DBT.
- To quantify changes in key performance metrics including sensitivity, specificity, and recall rates with AI assistance.
Main Methods:
- A deep learning AI system was developed for identifying suspicious lesions and calcifications in DBT images.
- A reader study involved 24 radiologists evaluating 260 DBT examinations (65 with cancer) with and without AI assistance.
- Performance metrics such as area under the receiver operating characteristic curve (AUC), reading time, sensitivity, specificity, and recall rate were statistically analyzed.
Main Results:
- Mean AUC increased by 0.057 with AI (0.795 to 0.852), indicating improved cancer detection performance.
- Radiologist reading time decreased by 52.7% (64.1 to 30.4 seconds) when using the AI system.
- Sensitivity improved by 8.0%, specificity by 6.9%, and recall rate for non-cancers decreased by 7.2% with AI.
Conclusions:
- The concurrent use of the AI system significantly improved cancer detection efficacy in DBT interpretation.
- AI demonstrated increases in AUC, sensitivity, and specificity, while reducing reading time and recall rates.
- This AI tool shows promise for enhancing diagnostic accuracy and efficiency in breast cancer screening workflows.
