Research on The Identification of Benign and Malignant Thyroid Nodule Ultrasound Images Based on Deep Learning Model

Abstract

The incidence of thyroid cancer has been on the rise in recent years and early diagnosis of thyroid nodules can effectively reduce the mortality rate of thyroid cancer. In the field of medical imaging of the thyroid gland ultrasound testing is the preferred option with low harm to the body. Clinically, the manual diagnosis of benign and malignant thyroid nodules is closely related to the clinical experience of the physician. With the continuous development of artificial intelligence technology, deep learning has made tremendous progress in ultrasound map detection. Based on the ultrasound image, we can build a detection model to detect the location of thyroid nodules and classify them as benign or malignant. It can assist doctors in intelligent diagnosis. In this paper, based on the acquisition of thyroid nodules ultrasonogram. Firstly, the image data are preprocessed, and secondly, the YOLOv5 algorithm is improved by adding the SENet attention module to the convolutional layer, while replacing the SiLU activation function with FReLU, and improving the PANet feature fusion network by adding a jump connection, and the new deep learning model constructed can achieve more accurate benign and malignant recognition.

Ähnliche Arbeiten