Double-reading mammograms using artificial intelligence technologies: A new model of mass preventive examination organization

Abstract

Aims: The aim of the study is to justify the possibility and effectiveness of using software based on artificial intelligence technology for the first interpretation of digital mammograms while maintaining the practice of the second description of X-ray images by a radiologist.
 Materials and methods: A data set of 100 digital mammography studies (50 Normal, 50 with signs of malignant neoplasms Not normal) was processed by software based on artificial intelligence technology registered in the Russian Federation as a medical device. ROC analysis was performed.
 Results: When set to 80.0% sensitivity: AI specificity was 90.0% (95% CI: 81.7-98.3), accuracy 85.0% (95% CI: 78.0-92.0). When set to 100% specificity: AI showed sensitivity of 56.0% (95% CI: 42.2-69.8), accuracy of 78.0% (95% CI: 69.9-86.1). When set to 100% sensitivity: AI specificity was 54.0% (95% CI: 40.2-67.8), accuracy 77.0% (95% CI: 68.8-85.2). Two approaches have been proposed that provide for an autonomous first interpretation of digital mammography using AI. The first approach is to evaluate the X-ray image using AI with a higher sensitivity than that of the mammograms double reading by radiologists, with a comparable level of specificity. The second approach implies that the software based on artificial intelligence technology will determine the mammograms category ("Normal" or "Not normal"), indicating the degree of its "confidence" in the obtained result, depending on the corridor into which the predicted value falls.
 Conclusions: Both proposed approaches for the use of software based on artificial intelligence technology for the purpose of autonomous first interpretation of digital mammograms are able to provide diagnostic quality that is not inferior to the images double reading by radiologists, and even exceeds it. The economic benefit from the practical implementation of this approach nationwide can range from 0.6 to 5.5 billion rubles annually.