基于分类回归树和AdaBoost的眼底图像视网膜血管分割

Retinal Vessel Segmentation in Fundus Images Using CART and AdaBoost

提出一种能有效分割眼底图像中视网膜血管的监督学习方法,为眼底图中的每个像素点构造一个包括局部特征、形态学特征和Gabor特征在内的39维特征向量,用以判定其是否为血管上的像素.在进行分类计算时,以分类回归树作为弱分类器对样本集分类,然后对AdaBoost分类器进行训练得到强分类器,并由此完成各个像素点的分类判定.基于国际公共数据库DRIVE的实验结果表明,该方法的平均精确度达到0.960 7,且敏感度和特异性均优于已有的基于监督学习的方法,适用于眼底图像的计算机辅助定量分析和疾病诊断.

It is proposed an effective method based on supervised learning for retinal vessel segmentation in fundus images.To determine whether a pixel is in the vessel,a 39-dimensional feature vector is extracted for every pixel,consisting of local features,morphological features and Gabor features.Afterwards,the sampled set is first treated by the classification and regression tree （CART） as a weak classifier,and then strengthened by a trained AdaBoost-based classifier as a strong classifier,to classify the pixels.The proposed method is evaluated with the public digital retinal images for vessel extraction （DRIVE） set and experimental results show that the proposed method has a high average accuracy of 0.9607 and performs better than other approaches based on supervised learning in sensitivity and specificity.It is suitable for computer-aided eye disease diagnosis and evaluation using fundus images.