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

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