摘要
急性冠状动脉综合征(ACS)是冠心病(CAD)中最为凶险的,而导致ACS的主要原因是冠状动脉粥样硬化斑块易发斑块破裂,因此早期识别斑块十分重要。为此提出了一个针对冠脉OCT图像的基于局部特征的纤维化斑块自动识别的算法。首先我们对OCT图像进行预处理,提取纤维和非纤维的局部特征作为训练样本,然后利用卷积神经网络(CNN)的学习方法进行训练、测试,通过对待测OCT图像的测试,纤维斑块平均识别率达到91.36%。经有经验的医师检验,实验方法可以对纤维斑块的自动识别有很高的识别率。
Acute coronary syndrome (ACS) in coronary heart disease (CAD) is the most dangerous, and the main cause of ACS is the rupture of plaque in the coronary artery. Therefore, early identification of plaque is very important. In this paper, a method for automatic identification of fibrous plaques based on local features of the coronary OCT images is proposed. First, we preprocessed the OCT image, and extracted the local features of the fibrous plaque and non fibrous plaque as the training sample. Then, the learning method of the convolutional neural network (CNN) was used for training and testing. By treating the test of OCT images, the average identification rate of fibrous plaque was 91.36%. Under the test of experienced doctor, the method of this paper can help to complete the automatic identification of the fibrous plaque.
出处
《激光杂志》
北大核心
2016年第3期57-60,共4页
Laser Journal
基金
国家自然科学基金资助项目(61473112
61203160)
河北省自然科学基金资助项目(F2015201196)
河北省教育厅项目(QN2014166)