光学与光电技术, 2018, 16 (3): 22, 网络出版: 2018-07-31  

基于视窗聚类的人体外周血白细胞快速无重复自动扫描方法

A Non-Redundant Human Peripheral Blood Leukocyte Fast Automatic Scanning Method Based on Window Clustering
作者单位
四川大学电子信息学院, 四川 成都 610064
摘要
为了实现人体外周血白细胞的快速无重复自动扫描,提出了一种新的聚类方法——视窗聚类法。由于人体外周血白细胞在血液涂片中分布稀疏且不均匀,传统的利用高倍镜直接逐行扫描的人工方法效率比较低下。用高倍物镜有目的地对有白细胞的区域进行拍摄可以提高白细胞的扫描效率。用低倍物镜预扫描血片,由于低倍物镜视场大,在低倍图中,用图像处理的方法可获得白细胞核心坐标。视窗聚类方法首先用最大最小距离聚类方法确定白细胞核心原始聚类中心,再按照高倍镜视窗大小对聚类中心进行调整,在确保各聚类视窗内无重复白细胞存在的同时,尽可能合并原始聚类中心,减少总的聚类中心个数。以新的视窗聚类中心组成高倍物镜扫描路径扫描血片,可保证高倍图像最大程度地包含尽可能多的白细胞,进而减少高倍图像的拍摄次数,实现了白细胞的快速无重复扫描。实验表明,用该方法获取110个无重复的白细胞图像仅需要拍摄高倍图像72~85幅,整个过程大概需要1.5~2.0 min。同时,视窗聚类方法可以在无人监管的情况下完全避免白细胞图像的重复拍摄,实现了人体外周血白细胞的自动扫描。
Abstract
In order to realize fast non-redundant auto-scanning of human peripheral blood leukocytes, a new clustering method called window clustering method is proposed. As a result of human peripheral blood leukocytes’ sparse and uneven distribution in blood smear, the use of traditional method that the medical staffs progressive-scan the blood smear directly use high power lens is low efficiency. Lots of images without leukocyte are captured due to the uneven leukocyte distribution in blood smear. There is no point in capturing the regions without leukocyte, besides it will increase follow-up workload. If we can know the location of the leukocytes in advance, we can scan purposefully. Because low power lens has large field of view, we can obtain coordinates of leukocyte nucleus by pre-scanning the blood smear. Making window clustering on leukocyte nucleus’ coordinate in leukocyte low power microscopic image can greatly reduce the total number of captured images and can avoid redundantly capture. Firstly, the initial clustering center is figured out using max-min clustering method. Then the initial clustering center is adjusted according to the window of high power objective. This method can merge the original clustering center as far as possible and it can also make sure that there is no redundant leukocyte in any new cluster. The new clusters’ center coordinates make up a routine. This method can ensure that high power window can furthest obtain more leukocytes. Therefore, high power images’ capture times will be reduced. The leukocyte non-redundant auto-scanning is realized. The efficiency of the scanning is greatly improved . Experiments show that to obtain 110 non-redundant leukocytes the proposed method needs to capture 72~85 times and the time consumption is 1.5~2.0 min. What’s more, window clustering can completely avoid leukocyte redundant capture without supervise, and it makes leukocyte of human peripheral blood automatic scanning a reality.

王亚品, 曹益平, 陈澄, 付光凯, 万莹莹, 王璐. 基于视窗聚类的人体外周血白细胞快速无重复自动扫描方法[J]. 光学与光电技术, 2018, 16(3): 22. WANG Ya-pin, CAO Yi-ping, CHEN Cheng, FU Guang-kai, WAN Ying-ying, WANG Lu. A Non-Redundant Human Peripheral Blood Leukocyte Fast Automatic Scanning Method Based on Window Clustering[J]. OPTICS & OPTOELECTRONIC TECHNOLOGY, 2018, 16(3): 22.

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!