3 Proficient implementation toward detection of thyroid nodules for AR/VR environment through deep learning methodologies

Abstract

A disease that commonly exists on a global scale is the thyroid nodule. It is identified by unusual thyroid tissue development. The nodules of the thyroid gland are of two kinds namely, benign and malignant. Accurate diagnosis of thyroid nodules is necessary for appropriate clinical therapy. Ultrasound (US) is one of the most often-employed imaging tools for assessing and evaluating thyroid nodules. It performs well when it comes to distinguishing between the two kinds of thyroid nodules. But diagnosis based on US images is not simple and depends majorly on the radiologists' experience. Radiologists, sometimes, may not notice minor elements of an US image, leading to an incorrect diagnosis. To help physicians and radiologists to diagnose better, several Deep Learning (DL) based models that can accurately classify the nodules as benign and malignant have been implemented. After performing a comparative study of several DL-based models implemented with different classification algorithms on an Open Source dataset, it has been found that InceptionNet V3 gave the best accuracy (~96%); others were F1 Score (0.957), Sensitivity (0.917), etc. A simple and easy-to-use graphical user interface (GUI) has been implemented. A US image of the thyroid gland is to be uploaded by the user following which the classification output is displayed along with a link redirecting the user to a page, which has more information about the reasons for the classification obtained. A downloadable file can also be obtained with the same information as the page, using the link.

Ähnliche Arbeiten