Sistema de reconhecimento de imagens para avaliação do movimento toracoabdominal em recém-nascidos

Abstract

This project has been develop in partnership with Trairí’s Health Science Faculty (FACISA – UFRN ) aiming to optimize the evaluation of thoracoabdominal movements in newborns, after receiving physiotherapeutic respiratory maneuver. However, in newborns, there is no effective method to assess respiratory kinematics. The traditional technique for respiratory volume measurement, called plethysmography, is not viable for clinical assessments in neonates. Some papers propose a different solution for this problem; by using lateral pictures of the newborn and processing those images in vector-based software, like CorelDraw or AutoCAD, where the user has to manually add a curve over the thoracoabdominal region to delimitate the wanted areas. On these softwares, each picture has to be individually analyzed, the focus area has to be calculated and converted to cm2, by applying an appropriated scale, which is derived from a standard length value in the image. After compiling all the images, the maximum and minimum point of the respiratory motion is acquired, providing enough data to evaluate the efficiency of the maneuver. Nevertheless, this process is slow and inaccurate, because it relies on user’s proficiency on said softwares. In this context, this project was developed, as a first functional and automated method to determine the thoracoabdominal volume from 2D images of neonates. Developed using MATLAB’s image processing toolboxes, the algorithm is capable of identifying the reference points and demarcating the thorax curvature automatically. This way, the application was able to process and quantify, sequentially, a larger set of images, always using the same standard, determining the maximum and minimum areas of the thorax and the abdomen throughout the entire respiratory motion of newborns, before and after applying the physiotherapeutic respiratory maneuver