Early Identification and Classification of Thyroid Nodule in Medical Ultrasound Images

Abstract

In medical imaging, machine learning (ML) and deep learning (DL) plays a significant role to predict symptoms of early disease. DL is one of the growing trends in general data analysis since 2013. It is an improvement of artificial neural networks (ANN), that consists of many hidden layers which permits high level of abstraction from data. In particular, convolutional neural networks (CNNs) have proven to be a potential tool for computer vision tasks. Deep CNNs have a capability to automatically learn raw data especially images. In the field of medical imaging, the accurate assessment or identification of disease depends on both image interpretation and acquisition. Due to the improvement in last decade in image acquisition, devices acquire data at high rate with increased resolution. However, the interpretation of process has recently begun to benefit from computer technology. Mostly these are made by the radiologist, physicians and senior doctors but limited due to its subjectivity and high skilled physicians/ doctors. Computerized tools in the medical imaging field are the key enablers to improve diagnosis by facilitating the findings. Analysis of thyroid ultrasonography (USG) images via visual inspection and manual examination for early identification and classification of thyroid nodule has always been cumbersome. This manual examination of thyroid USG images in order to identify benign and malignant thyroid nodule can be tedious and time-consuming. With the rapid advancement in technology and increase in computational resources various deep learning models have emerged in medical field especially in thyroid nodule classification. Early identification of these nodules can improve the effectiveness of clinical interventions and treatments. Therefore, these days many researchers now advocate the use of computer diagnostic system (CDS) to objectively and quantitatively analyze the ultrasonography images of thyroid nodules. This helps them to solve the differences for radiologists so as to solve the differences in interpretation results. The key motive of this work is to develop an efficient model using DL techniques for thyroid nodule identification and classification. The first phase of this work proposes two models ANN-SVM hybrid model and CNN-SVM hybrid model. The model works in four stages namely data collection, preprocessing, classification using hybridization of ML and DL classifiers and result analysis. The optimization of a model plays a significant role to improve the performance of the model. The second phase of this work proposes an optimized CNN based model using segmentation and boundary detection technique. In this work, Alex-Net, visual geometry group (VGG-16), deep neural network (DNN) and Res-Net-50 models are optimized using grid search optimization (GSO) technique. The third part of this work proposes deep-generative adversial network (Deep-GAN) model to improve an accuracy of the model. VGG-GAN and Alex-GAN models are proposed in this work. The proposed Alex-GAN model outperforms the rest of existing and VGG-GAN model.