首页|基于YOLOv5模型的自动乳腺超声乳头目标检测

基于YOLOv5模型的自动乳腺超声乳头目标检测

扫码查看
近年来乳腺疾病尤其是乳腺癌的发病率呈上升趋势,其早期诊断尤为重要.自动乳腺超声(ABUS)作为一种新型三维超声成像技术,有助于提高乳腺癌筛查和诊断的准确性.在ABUS中,乳头区域的定位和分割是一个重要过程.为了解决深度学习标签难以获取的问题,文章首先提出了一种半自动标签标注方法,以简化数据标签的获取过程;其次,提出利用YOLOv5目标检测算法对ABUS中的乳头阴影区域进行检测.试验测试结果表明,该模型可以达到96.5%的精确率以及92.8%的召回率,并能以较高的效率进行推理,基本满足了工程与临床的需求.
ABUS Nipple Object Detection Based on YOLOv5 Model
In recent years,the incidence rate of breast diseases,especially breast cancer,is on the rise,and its early diagnosis is particularly important.As a new three-dimensional ultrasound imaging technology,Automated Breast Ultrasound(ABUS)is helpful to improve the accuracy of breast cancer screening and diagnosis.In ABUS,the localization and segmentation of the nipple area is an important process.In order to solve the problem of difficulty in obtaining deep learning labels,this paper first proposes a semi-automatic label annotation method to simplify the process of obtaining data labels;Secondly,it is proposed to use the YOLOv5 object detection algorithm to detect nipple shadow areas in ABUS.Experimental test results show that the model can achieve an accuracy of 96.5%and a recall rate of 92.8%,and can reason with high efficiency,basically meeting the needs of engineering and clinical practice.

3D ultrasoundnipple detectionYOLO

曾焕城、陈嘉炜

展开 >

汕头大学医学院附属肿瘤医院(广东 汕头 515041)

汕头大学(广东 汕头 515063)

三维超声 乳头检测 YOLO

2024

中国医疗器械信息
中国医疗器械行业协会

中国医疗器械信息

影响因子:0.375
ISSN:1006-6586
年,卷(期):2024.30(19)