APPROBATION OF A MODEL FOR ANALYZING X-RAY DATA WITH AN AUTOMATICALLY CONFIGURABLE COLLIMATOR

Golia Roman Dmitrievich

In this article, the approbation of the model for analyzing X-ray data showed a significant reduction in radiation dose, which, along with accurate consideration of the device's features and anatomy, indicates that the system is able to meet user requests in the field of radiation reduction and adaptation to the individual characteristics of patients and medical devices. Additional measures have been taken to ensure maximum consistency of our method: Training the model on raw X-rays rather than on processed images. This ensures that the anatomical model does not depend on post-processing algorithms used in various data collection modes; The use of a diverse dataset that includes images with darkened areas in the background (for example, tables, intravenous infusion lines, electrocardiograms, surgical instruments and hands); Images that went beyond the target procedural phase are included. (for example, from the wrist to the shoulder); An aggressive strategy was used to increase the amount of data.

Keywords:medicine, artificial intelligence, neural networks, radiation dose, data collection, diagnostics, modern algorithms, models.