Automated morphometry of the prostate gland by the results of magnetic resonance imaging

BACKGROUND: Within the framework of the experiment on using the innovative technologies in the field of computer vision for analyzing the medical images and on further usage of these technologies in the healthcare system of the City of Moscow, the research was carried out using the equipment based on the artificial intelligence (AI-service) for the purpose of automatization of the morphometry of the prostate gland using the magnetic resonance imaging (MRI), for the issue is topical due to the high incidence of urological diseases among men. Unlike the 11 previous systems, oriented at the retrospective analysis, this solution helps the radiologists in shortening the time of describing the examination results and in increasing their accuracy. AIM: to evaluate the quality and the validity of automatic morphometry of the prostate gland by the MRI results using the technologies of artificial intelligence in the settings of practical healthcare. METHODS: A prospective diagnostic research in accordance with the methodology of reporting results of scientific research involving the STARD 2015 diagnostic tests was conducted during the period from April until October of 2024. A total of 560 MRI results were used and compared to the data from the morphometric AI-service. RESULTS: An evaluation of the accuracy of using the AI-service for the morphometry of the prostate gland was carried out. A total of 7 clinical monitoring procedures were conducted using 560 MRI datasets with the complete conformity reported in 71.6%. The rate of false-negative cases was 3.9%, technical defects were found in 3.8% of the cases. The integral clinical evaluation has achieved the range of 88.0–97.0%, confirming the high diagnostic quality. The predominant errors were the ones related to the contouring of the gland (52%) and incorrect measurements (13%), often related to the prolapsing of the prostate gland apex. CONCLUSION: The automatization of routine measurements greatly contributes to the standardizing the processes of describing the results obtained by radio-diagnostic methods. This aspect is of special importance from the point of view of providing the continuity of medical aid in case of patients presenting to various medical organizations. The artificial intelligence technologies for the automatization of the prostate gland measurements have demonstrated high clinical value in 92.0%, which indicates their accuracy and quality. These data can be used for developing new MRI-based automated morphometry products.