摘要
结核感染T细胞斑点试验(T-SPOT.TB)是一种可靠的结核感染诊断方法,而确定T细胞菌斑数量是T-SPOT.TB中的一项重要工作。针对现有的菌落自动计数系统对T细胞菌斑图像适用性较差的缺点,提出了一种基于改进分水岭算法,对T细胞菌斑图像进行全自动分割和计数的方法。该方法首先利用霍夫变换将培养皿区域提取出来,再使用中值滤波去除图像噪声,然后结合阈值分割、形态学处理和面积滤波等方法识别出图像的粘连区域,最后使用分水岭算法对图像进行分割,并通过对粘连区域的二次分割实现粘连目标计数修正。本文以50多例培养皿样本作为测试集,与人工计数结果进行对比实验,结果证明该方法准确率达95%。
Tuberculosis infection T-cell spot test(T-SPOT.TB)is a reliable method for the diagnosis of tuberculosis infection,and determining the number of T-cell plaque is an important work in T-SPOT.TB.In view of the poor applicability of the existing colony automatic counting system to T-cell plaque images,a fully automatic segmentation and counting method for T-cell plaque images based on improved watershed algorithm is proposed.Firstly,the Hough transform is used to extract the petri dish area,then median filtering is used to remove image noise,then threshold segmentation,morphological processing and area filtering is combined to identify the adhesion area of the image,finally,the watershed algorithm is used to segment the image and the adhesion target count correction is achieved through a second segmentation processing step of the adhesion area.Taking more than 50 petri dish samples as the test set,experiments comparing with the manual counting results have shown an accuracy up to 95%.
作者
吕铭轩
陈兆学
LV Mingxuan;CHEN Zhaoxue(School of Medical Instrument and Food Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China)
出处
《智能计算机与应用》
2022年第1期41-45,52,共6页
Intelligent Computer and Applications
基金
上海理工大学医疗器械与食品学院微创创新基金项目
关键词
分水岭分割
图像处理
菌落计数
图像分割
Watershed Segmentation
Image Processing
Colony counting
Image segmentation
