Paper Title
Thyroid Detection Using Ultrasound Imaging With Deep Learning Techniques

Abstract
With the dramatic rise in thyroid cancer cases, the increasing burden on radiologists for accurate sonographic diagnosis underscores the need for a reliable and efficient computer-aided diagnosis system. This work addresses the challenge by automating thyroid nodule detection in ultrasound imaging, leveraging advanced object detection techniques to enhance precision and efficiency. A publicly available thyroid ultrasound dataset was employed to train and validate the proposed system. The approach integrates mechanisms to improve feature extraction, address class imbalance, and enhance the detection of small and overlapping nodules. These enhancements ensure robust performance in identifying abnormalities and supporting radiologists in accurate diagnostic decision-making. Among the models tested, YOLOv5x6 achieved the greatest findings, showing its efficacy in accurately detecting thyroid nodules with a precision of 0.561, recall of 0.835, and mean average precision (mAP) of 0.650. These results demonstrate the system's ability to greatly enhance diagnostic procedures and guarantee accurate and prompt detection of thyroid problems. Keywords - Label Smoothing, YOLOv5 Network, Ultra Sonography, Thyroid Nodules, Computer-Aided Diagnostics, and Attention Module.