Use of artificial neural network algorithm in the immunohistochemical dyeing based diagnosis of thyroid tumor
Ahmet Haltas, Ahmet Alkan, Mustafa Karabulut · 2014
Separation of thyroid nodules which are often morphologically similar is a significant task. Due to this similarity, different pathologist may fail to correctly evaluate and misdiagnose the lesions. Such misdiagnoses lead to substantial problems such as late recovery and unnecessary treatment process and costs. Use of immunohistochemical dyes help pathologists correctly diagnose the cases where morphologically similar lesions are present. In this study, a decision support system is proposed to diagnose nodules into benign and malignant by analyzing data via artificial neural networks (ANNs). In the work, a set of 63 samples are utilized to train and test the neural network based algorithm. As a result, %95 accuracy is reached. Also, the proposed method which is based on ANNs is compared against some well-known classification algorithms.