Enhancing Thyroid Disease Detection through IWSO-based Ontology Matching and En-SwinT+ Classification Model

Abstract

Currently, medical informatics technology, Clinical decision support systems, and the 
medical technology are all making heavy use of medical ontologies. These coordinated and 
well-integrated systems are necessary for the smooth and accurate transmission of data and 
information. This, in turn, can lead to superior patient care and control of illnesses. Therefore, 
incorporating medical ontologies into these pathways is very important to its future success 
and development. Thyroid disease is a complicated health problem that has too much thyroid 
hormone. This is generally diagnosed with techniques such as CT scans and X-rays. The 
thyroid can be seen in the neck, just below the larynx. Thyroid problems can be found and 
solved by using Deep Learning (DL) technology in the proposed treatment. So, in this paper, 
to build semantic links between different datasets, this paper first apply Metaheuristic 
Optimization Strategy at Improved War Strategy Optimization (IWSO) at the beginning of 
our study to match ontologies properly. Subsequent to this, a preprocessing including Log 
and Gaussian filters is done to complete missing values and normalize features data for 
convenient comparison at the times of the modelling training. Then follows feature 
extraction, performed by the Gabor Wavelet Transform (GWT). For early detection of this 
disease, the study uses an Enhanced Swin transformer (En-SwinT+) model. This is a 
specially designed model by deep learning for just such diseases. Swin Transformer Inside 
has been refitted into a thyroid disease detection device. Our proposed model does better than 
the existing models. It attains an accuracy of 99. 45 % in the task of recognizing problems 
related to the thyroid gland. With the up-to-date and revolutionary methods employed in this 
research, thyroid disease identification effectiveness and accuracy can be greatly improved.