数字液滴PCR图像的荧光信息提取

Extraction method for fluorescence information from droplet digital PCR images

针对数字液滴聚合酶链式反应(droplet digital Polymerase Chain Reaction,ddPCR)核酸检测中数字聚合酶链式反应由液滴数量多、尺寸小、排列紧密、荧光强度不均匀而导致的液滴难以计数问题,提出一种图像处理方法,基于灰度遍历法采集图像的信息,通过微分分析法对液滴进行分类和计数,可准确采集ddPCR实验的图像信息。通过卷积算法去除图像中的噪声,使用灰度分布均衡化法增强图像的对比度。以灰度遍历的方式将图像在逐个阈值下二值化,并以几何条件为限制统计液滴数量。通过微分分析法对数据进行分析,得出荧光液滴与全部液滴的计数结果。在以人类gDNA(genomic DNA)为检测样本的ddPCR实验中,该算法的平均检测准确率为99.36%,与商用仪器算法和同类算法相比分别提高了2.24%,2.53%。该方法为ddPCR实验提供了可靠的检测结果,可更好地适用于ddPCR实验。

Droplet digital polymerase chain reaction(ddPCR)is one of the main methods used in nucleic acid detection.Counting the droplets in ddPCR experiments can be difficult if the number of droplets is too high or low,the fluorescence intensity is uneven,or the droplets are too closely arranged.We propose an image analysis method in which the image information is collected via the gray traversal method and droplets are classified and counted by differential analysis.Noise is removed from the image using the convolution algorithm,and the gray contrast of the image is enhanced by gray distribution equalization.The image is then converted to binary form by the gray traversal method with the geometric conditions taken as screening conditions,and the number of droplets is counted by differential analysis.Experiments with human gDNA(genomic DNA)as a detection sample,with approximately 20000 droplets,showed an average detection accuracy of 99.36%as compared to those conducted using commercial instrument algorithms or similar,with an average improvement of 2.24%and 2.53%,respectively,in the detection accuracy.The developed method can therefore be considered to provide reliable results and is suitable for use in ddPCR experiments.