摘要
针对正常与癌变大肠病理切片图像的特征,结合主成分分析(PCA)和K-Means算法提出了一种分割大肠病理切片图像中腺腔和上皮细胞、细胞核、间质的算法,解决了传统K-Means算法确定初始中心的难点,提高了识别分类时的收敛速度.使用基于相关系数矩阵的主成分分析方法确定具有代表性的聚类初始中心,结合K-Means算法将大肠病理切片图像数据分成三类.相关实验证明:提出的改进型K-Means大肠病理切片图像分割算法能够准确地将大肠病理切片图像中的腺腔和上皮细胞、细胞核、间质分类,且使用PCA方法的算法收敛速度比传统使用RANDOM方法的算法更快,取得了良好效果.
Targeting at the characteristics of normal and cancerous large intestine pathological section,and combining PCA （principal component analysis） and K-Means algorithm,an algorithm for segmenting the lumen,epithelial cell and cell nucleus in large intestine pathological section was proposed.This method can solve the difficulties in determining the initial center in traditional K-Means algorithm and improve the rate of convergence in identification and classification.The representative clustering initial center was determined throuth PCA based on correlation coefficient matrix.By combining the K-Means algorithm,the image data was divided into three classes.The experiments show that the improved K-Means algorithm for image segmentation of large intestine pathological section can classify the lumen,epithelial cell and cell nucleus in large intestine pathological section accurately,and rate of convergence of this algorithm is faster than traditional algorithm with RANDOM method.This method has good result.
出处
《浙江工业大学学报》
CAS
2014年第5期581-585,共5页
Journal of Zhejiang University of Technology
