Inteligencia artificial para diagnóstico de infecciones en vías aéreas bajas en radiografías de pacientes pediátricos

Abstract

Pneumonia in children represents a crucial point in medical care, its detection has been a diagnostic criterion due to its easy access, cost and reproduction. The use of artificial intelligence (AI) with increasing development in the health area, through algorithms, has shown progress in image recognition tasks. In the practice of pediatric radiology, it is possible to evaluate medical images for the detection, characterization and monitoring of chest diseases. This study will estimate the validity of an artificial intelligence system for the diagnosis of pulmonary involvement in chest radiographs of patients with lower airway infection from a database with patients from 1 month to 17 years (stages of development) with positive results for pneumonia and bronchopneumonia, corroborated by laboratory. We use a convolutional neural network algorithm-type chest evaluation system, which detects whether or not the lung is affected. The group of infants with pneumonia demonstrated greater precision (0.9) in the diagnosis. This work is part of the beginning of a field of development and research, considering it as a tool for care in health systems. (English) [ABSTRACT FROM AUTHOR]