Artificial intelligence in breast cancer diagnosis: regional experience

Abstract

Aim: to analyze the results of the implementation and use of a product with artificial intelligence (AI) technology in the practice of radiologists during mammographic examination.
Materials and methods: database of patients who underwent mammographic examination within the framework of medical examination of certain groups of the adult population and preventive medical check-ups, whose images were reviewed by specialists of the Reference Center of the Krasnoyarsk Regional Clinical Oncology Dispensary and AI. The results were processed using the StatTech 4.0.6 software. Discordance was considered for clinically significant discrepancies in which the patient's management tactics were changed.
Results: in the Krasnoyarsk Territory, the introduction of AI into the practice of radiologists during mammography examination led to an increase in diagnostically difficult categories of BI-RADS 3,4, which increased the workload on the Reference Center by 40.8%. There was a 1.9% decrease in the discordance rate compared to the period when AI was not used in the region, indicating that doctors are not simply accepting the AI results and sending them to the Reference Center for review, but are analyzing the resulting AI report, and the key decision rests with the radiologist. Conclusion: the use of AI in the practice of a radiologist has both positive and negative sides. The negative ones are mostly related to technical and organizational problems, eliminating which it is possible to improve the quality of mammographic studies and their description.