Performance benchmarking

Journal of the National Cancer Institute

5 March 2019


Rodríguez-Ruiz A, Lång K, Gubern-Merida A et al.

Stand-Alone Artificial Intelligence for Breast Cancer Detection in Mammography: Comparison With 101 Radiologists

Rodríguez-Ruiz A, Lång K, Gubern-Merida A, Broeders M, Gennaro G, Clauser P, Helbich T, Chevalier M, Tan T, Mertelmeier T, Wallis M, Andersson I, Zackrisson S, Mann R, Sechopoulos I
Journal of the National Cancer Institute 2019,

Aim and Objective
To independently evaluate the stand-alone breast cancer detection performance of Transpara Artificial Intelligence (AI) system in a multi-vendor multi-center cohort, and to compare it to radiologists.

Materials and Method
Nine independent and external datasets of mammograms from different US and European institutions were collected.
Each mammogram had been retrospectively assessed by multiple radiologists. In this study each mammogram was also assessed by Transpara (version 1.4.0). Mammogram assessments were in terms of probability of malignancy for each exam In total there were 2652 mammograms (653 malignant) and interpretations by 101 radiologists (28 296 independent interpretations). Mammograms were from four different vendors. The breast cancer detection performance (evaluated in terms of area under the receiver operating characteristic curve, AUC) between the radiologists and Transpara was compared using a noninferiority null hypothesis at a margin of 0.05.

The performance of Transpara was comparable to the average of the 101 radiologists. Transpara had an AUC of 0.840 and the average of the radiologists had an AUC of 0.814. The AI system had an AUC higher than 61.4% of the radiologists.

Transpara achieved a cancer detection accuracy comparable to an average breast radiologist in this retrospective setting. AI systems performing at radiologist-like levels in the evaluation of digital mammography (DM) have the potential to significantly improve breast cancer screening accuracy and efficiency.

You might also like