Artificial Intelligence vs Radiologists in Breast Cancer Detection With Digital Mammography
In a study reported in the Journal of the National Cancer Institute, Rodriguez-Ruiz et al found that a stand-alone artificial intelligence (AI) system had accuracy comparable to an average breast radiologist in detecting breast cancer using digital mammography.
Study Details
In the retrospective study, nine multireader, multicase study datasets previously used for different research purposes in the United States and six European countries were collected. Each dataset consisted of digital mammography exams acquired with systems from four different vendors, multiple radiologists’ assessments per exam, with breast cancer status verified by histopathologic analysis or follow-up. The total dataset consisted of 2,652 exams, with 653 malignant cases and interpretations by 101 radiologists. An AI system analyzed the exams, and rated level of suspicion of cancer on a scale of 1 to 10 (10 representing highly suspicious of malignancy present).
Performance of AI and Radiologists
Overall, the performance of the AI system in detecting cancer was statistically noninferior to that of the average of the 101 radiologists. The receiver operating characteristic area under the curve (AUC) value was 0.840 for the AI system vs an average of 0.814 for the 101 radiologists. The AUC difference was 0.026 (95% confidence interval = -0.003–0.055). The AI system had an AUC higher than that of 61.4% of the radiologists and higher sensitivity than 57.9% of radiologists. However, its performance was always poorer than that of the most accurate radiologist. The AUC and sensitivity of the AI system were similar to those for the average of radiologists within each of the individual datasets.
The investigators concluded, “The evaluated AI system achieved a cancer detection accuracy comparable to an average breast radiologist in this retrospective setting. Although promising, the performance and impact of such a system in a screening setting needs further investigation.”
Ioannis Sechopoulos, PhD, of the Department of Radiology and Nuclear Medicine, Radboud University Medical Centre, Nijmegen, is the corresponding author for the Journal of the National Cancer Institute article.
Disclosure: The study authors' full disclosures can be found at academic.oup.com.
The content in this post has not been reviewed by the American Society of Clinical Oncology, Inc. (ASCO®) and does not necessarily reflect the ideas and opinions of ASCO®.
