Implementation of artificial intelligence in speech (voice input) in the conditions of a high-flow endoscopy

Abstract

Artificial intelligence (AI) and digitalization are used in all spheres of life, including medicine: in making diagnoses, in working with patients, in creating and using an Electronic Medical Record (EMC) for fast and accessible document management. Attempts have been made repeatedly to facilitate work with medical protocols. Technologies related to speech recognition are a promising direction in this area. The purpose of the research work was to optimize the time of writing the protocol using voice input. The object of the study was a template protocol of endoscopic examination. This work was carried out on the basis of 2 large endoscopic centers in Moscow, created as part of the pilot project of the DM within the framework of the colorectal cancer screening program in the Russian Federation in 2021. The software of the Voice2Med Speech Technology Center (MDG) was used to fill out the research protocol. Using mathematical calculations, 4 metrics were created to evaluate the work of the software: the time of protocol formation; the proportion of protocol formation time from the study; the level of speech recognition accuracy; satisfaction of doctors when working with technology. Standardized protocol templates for various nosologies related to voice commands have been developed; medical thesaurus. In the course of the work, the accuracy of speech recognition was investigated and improved. Control measurements of the time of writing the protocol were carried out. The results of the evaluation of the use of the technology by endoscopists were obtained - the time of protocol formation decreased by 29%, the evaluation of the work by endoscopists was positive. Thus, the use of speech technologies related to the formation of a doctor’s protocol “from the voice” is a promising direction for the further development of the IT sphere in practical medicine.