Diagnostic Efficiency of Various Systems for Automatic Analysis of Radiographs in the Detection of Lung Nodule

The purpose of the study was to compare effectiveness various artificial intelligence systems for detecting foci and rounded lesions in lungs. For testing, we selected four software products based on convolutional neural networks, positioning themselves as a sensitive system evaluating digital chest radiographs. An analytical validation method used clinical evaluation. diagnostics, 3 data samples were formed with identification signs diseases (sample 1–5150 radiographs, detection pathological changes %; sample 2–100 6 3–300 prevalence 50 %). None passed AUC threshold 0.811 all three samples. In samples, have high accuracy sensitivity round formations, which leads rare cases overdiagnosis special underdiagnosis. use X-ray image analysis technologies is promising direction high-quality primarily when considering their young radiologists an additional opinion.