- AI
- Molecular Imaging
- CT
- X-Ray
- Ultrasound
- MRI
- Facility Management
- Mammography
Researchers Note Lack of Benefit with AI in Women with Dense Breasts
In a study of over 1.300 women with dense breasts, the combination of mammography and ultrasound had a recall rate of 11.7 percent, a specificity rate of 89.1 percent and an accuracy rate of 89.2 percent in comparison to a 21.4 percent recall rate, 79.4 percent specificity and 79.5 percent accuracy for the combination of mammography, ultrasound, and artificial intelligence (AI).
Adjunctive use of an artificial intelligence (AI) model, trained and validated on over 170,000 mammography exams, had significantly lower accuracy and specificity, and a higher recall rate in comparison to a combination of mammography and supplemental ultrasound, according to a new study of women with dense breasts.
For the retrospective study, recently published in the American Journal of Roentgenology, researchers compared mammography, stand-alone AI and ultrasound in 1,325 women (mean age of 53) with dense breasts. The study authors also compared the combination of mammography and AI to the combination of mammography and ultrasound and the combination of AI, mammography, and ultrasound.
While stand-alone AI demonstrated a higher cancer detection rate (CDR) (6.8) than mammography (6.0) and ultrasound (6.0), researchers found that it had the highest recall rate and lowest sensitivity and specificity of the three modalities.
In a study of 1,325 women with dense breasts, researchers found the mammography/ultrasound/AI combination had the highest recall rate (21.4 percent) in comparison to mammography/AI (14.9 percent) and mammography/ultrasound (11.7 percent). The mammography/ultrasound/AI combination also had the lowest specificity rate (79.4 percent) in comparison to mammography/AI (85.8 percent) and mammography/ultrasound (89.1 percent). (Images courtesy of the American Journal of Roientgenology.)
Specifically, stand-alone AI had an 11.9 percent recall rate in comparison to 4.4 percent for mammography and 9.2 percent for ultrasound. The study authors noted the sensitivity rate for stand-alone AI was 88.7 percent in comparison to 96.2 percent for mammography and 91.3 percent for ultrasound. Mammography had the highest specificity rate at 96.2 percent followed by 91.3 percent for ultrasound and 88.7 percent for stand-alone AI.
“Screening mammography alone, in comparison with stand-alone AI, screening ultrasound alone, mammography with AI, and mammography with ultrasound, showed significantly higher specificity and accuracy, with significantly lower recall rate and no significant difference in CDR or sensitivity,” wrote study co-author Hee Jung Moon, M.D., who is affiliated with the Department of Radiology at Wonju Severance Christian Hospital and the Yonsei University Wonju College of Medicine in Wonju, Korea, and colleagues.
The researchers also noted that the combination of mammography and supplemental ultrasound had the same sensitivity (100 percent) as the combination of mammography, AI, and ultrasound. The mammography/ultrasound/AI combination had the highest recall rate (21.4 percent) in comparison to mammography/AI (14.9 percent) and mammography/ultrasound (11.7 percent). The mammography/ultrasound/AI combination also had the lowest specificity rate (79.4 percent) in comparison to mammography/AI (85.8 percent) and mammography/ultrasound (89.1 percent).
“The present findings fail to show benefit of AI with respect to screening mammography performed with supplementary breast ultrasound in patients with dense breasts,” noted Moon and colleagues.
Could a Deep Learning Model for Mammography Improve Prediction of DCIS and Invasive Breast Cancer?
April 15th 2024Artificial intelligence (AI) assessment of mammography images may significantly enhance the prediction of invasive breast cancer and ductal carcinoma in situ (DCIS) in women with breast cancer, according to new research presented at the Society for Breast Imaging (SBI) conference.
