RASIT: Region shrinking based Accurate Segmentation of Inflammatory areas from Thermograms

摘要

Effective segmentation of thermal images reflecting the inflamed region in human body to assist medical diagnosis is a challenging task. In this paper we propose a method for thermal image segmentation, named as "Region shrinking based Accurate Segmentation of Inflammatory areas from Thermograms'', in short RASIT. The method comprising of four steps encompassing thermal image contextual electrostatic force extraction, intensity adjustment as applicable, automated generation of the weighted threshold, and segmentation of thermograms based on the computed threshold. The proposed method is operative devoid of the subjective and possibly questionable task of parameter selection clearly offering an edge over the state-of-the-art methods in terms of usage. The efficacy of our proposed technique is shown by experimenting on abnormal thermograms taken from two datasets: one is newly created knee arthritis thermogram dataset and another is online available Database of Mastology Research (DMR) of breast thermograms. The averages on correct detection rates obtained by the proposed method for both the knee and breast thermograms are 98.2% and 96.98% respectively with favorable inference on basis of Wilcoxon's test. Application of the proposed method minimizes the complexity of parameter selection, time complexity of execution and amount of under segmentation compared to existing state-of-the-art methods of thermogram segmentation. (C) 2018 Nalecz Institute of Biocybernetics and Biomedical Engineering of the Polish Academy of Sciences. Published by Elsevier B.V. All rights reserved.