计算机工程与应用 ›› 2022, Vol. 58 ›› Issue (13): 204-209.DOI: 10.3778/j.issn.1002-8331.2011-0424

• 图形图像处理 • 上一篇    下一篇

乳腺病灶的多显著性检测方法

方宏文,蔡念,周静雯,白有芳,黎剑,王晗   

  1. 1.广东工业大学 信息工程学院,广州 510006
    2.中山大学 肿瘤中心 诊断和介入超声科,广州 510060
    3.广东工业大学 机电工程学院,广州 510006
  • 出版日期:2022-07-01 发布日期:2022-07-01

Multi-Saliency Detection Method for Breast Lesions

FANG Hongwen, CAI Nian, ZHOU Jingwen, BAI Youfang, LI Jian, WANG Han   

  1. 1.School of Information Engineering, Guangdong University of Technology, Guangzhou 510006, China
    2.Department of Diagnostic and Interventional Ultrasound, Cancer Center, Sun Yat-Sen University, Guangzhou 510060, China
    3.School of Electromechanical Engineering, Guangdong University of Technology, Guangzhou 510006, China
  • Online:2022-07-01 Published:2022-07-01

摘要: 乳腺病灶自动检测对于后续计算机辅助诊断具有重要意义。融合背景先验、自适应中心先验以及频率先验等三种先验知识,提出一种多显著性检测方案实现超声影像中的乳腺病灶检测。该方案包含图像预处理、显著性检测和显著性优化等三个步骤。为了提高检测准确率,提出了一种基于经验累积分布函数的自适应阈值分割方法和一种改进的自适应中心先验检测方法。实验结果证明,提出的多显著性检测方案精确率达92.50%,召回率达87.05%,F-measure值达91.18%,能够更好地检测乳腺超声病灶。

关键词: 乳腺病灶检测, 乳腺超声图, 显著性检测

Abstract: Automatic detection of breast lesions is of great significance for subsequent computer-aided diagnosis. This paper combines three kinds of prior knowledge, i.e. background prior, adaptive center prior and frequency prior to propose a multi-saliency detection scheme, which is employed to detect breast lesions in ultrasound images. The scheme includes three stages:image preprocessing, saliency detection and saliency optimization. In addition, in order to improve the detection accuracy, it proposes an adaptive threshold segmentation method based on empirical cumulative distribution function and an improved adaptive center prior detection method. Experimental results indicate that the proposed multi-saliency detection scheme achieves the precision of 92.50%, recall of 87.05% and F-measure of 91.18%, which demonstrates better performance in detecting breast lesions in ultrasound images.

Key words: breast lesion detection, breast ultrasound images, saliency detection