Abordagem computacional de estimativa de temperatura por imagens térmicas para uso em neonatologia

Abstract

Infant radiant heat warmers are devices that assist in the thermoregulation of newborns in intensive care units, providing the necessary warmth and measuring the patient temperature through sensors in direct contact with the patient’s skin. However, precise monitoring of the temperature of these critical patients poses challenges due to the need for sensor attachment to the newborn’s skin, as well as the displacement of these sensors and wires during patient movement. This study presents a computational approach to non-invasively and contactlessly measure the temperature of newborns using infrared thermal imaging. The approach employs computer vision for analysis and processing of thermal images, enabling the extraction of numerical temperature values using knowledge acquired from studying the pseudo-color scale of the infrared camera. Correction algorithms are also applied to address measurement variations inherent to the camera used in the study, including artificial neural networks. The proposed approach allowed for more accurate measurements, reducing error from over 2°C to an average error of 0.1°C with a maximum error of 0.7°C. It is expected that the results of this work will contribute to the improvement of neonatal care by suggesting a technique for non-invasive temperature measurement, eliminating the need for physical contact sensors with the patient skin for monitoring newborns positioned in radiant heat warmers.