Automatic Segmentation of Neonates Thermal Imaging for Evaluation of Trunk Thermal Asymmetry

This paper proposes a new method for automatic segmentation of neonates's trunks in thermal images. The method is based on an algorithm that use a threshold, associated with an active contour technique. The hypothesis is that it is possible to correlate a given temperature map, presented in a t...

Full description

Saved in:

Bibliographic Details
Published in:	2018 IEEE International Conference on Bioinformatics and Biomedicine (BIBM) pp. 989 - 992
Main Authors:	de Farias, Thyago Maia Tavares, de Lima, Manoel Eusebio, da Silva Mattos, Sandra, de Araujo, Juliana Souza S., Mozer, Lucia Roberta D. N., Mourato, Felipe Alves
Format:	Conference Proceeding
Language:	English
Published:	IEEE 01-12-2018
Subjects:	Automatic Segmentation Blood Cameras Handheld computers Image segmentation Neonates Patent Ductus Arteriosus Pediatrics Temperature measurement Thermal Images Trunk
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	This paper proposes a new method for automatic segmentation of neonates's trunks in thermal images. The method is based on an algorithm that use a threshold, associated with an active contour technique. The hypothesis is that it is possible to correlate a given temperature map, presented in a thermal image, with the persistence of the heart ductus arteriosus of a newborn. From this region of interest (ROI), it is possible to capture quantitative and statistical information on the pathology's temperature map, in order to use them as input in a machine learning applications to, in addition to aiding early diagnosis, to enable its automation. However, the presence of clothing, bandages, sheets, and other objects during the catches may alter the temperature to be considered in the diagnosis. Thus, in order to guarantee the consistency of the data extracted from such images, the region of interest must be segmented and analyzed separately. The proposed automatic segmentation obtained low error rates, generating outputs very similar to those obtained in manual segmentation.
DOI:	10.1109/BIBM.2018.8621553