Abstract
Purpose: In the conditions of a constant increase in the number of computer vision algorithms developed based on artificial intelligence (AI) for medical diagnostics, it becomes necessary to determine criteria for deciding whether their practical application for mass preventive studies of the population is appropriate.
 Materials and methods: The study with the participation of several radiologists was conducted on a “Web platform for evaluating radiological studies” on a marked data set containing digital radiographs and fluorograms in an anterior direct projection. On the same data set, using the “Versioning Testing Platform”, responses were obtained from two commercial AI-based computer vision algorithms developed for the analysis of digital radiographs. Evaluation of the results obtained from doctors and algorithms (binary, in terms of “with pathology” and “without pathology”) was carried out using ROC analysis. For the threshold value calculated by the Yuden method, the following metrics were determined: sensitivity, specificity and accuracy.
 Results: diagnostic accuracy metrics were calculated for the average assessment of radiologists and AI-based computer vision algorithms when searching for pathological changes on chest X-rays in anterior direct projection according to ROC analysis. The average values of diagnostic accuracy indicators of radiologists exceeded the indicators of AI services.
 Conclusions: when deciding on the implementation of AI-based computer vision algorithms for preventive research, One should be guided by the metrics of diagnostic accuracy of a particular algorithm and use the average result of doctors in solving this diagnostic problem as the target values of metrics.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.