用于寡克隆带检测的脑脊液凝胶电泳半自动图像分析

Semi-automated image analysis of gel electrophoresis of cerebrospinal fluid for oligoclonal band detection.

作者信息

Boudet S, Peyrodie L, Wang Z, Forzy G

出版信息

Annu Int Conf IEEE Eng Med Biol Soc. 2016 Aug;2016:744-747. doi: 10.1109/EMBC.2016.7590809.

DOI:10.1109/EMBC.2016.7590809

Abstract

Detection of oligoclonal electrophoretic bands in cerebrospinal fluid (CSF) is an important diagnostic tool for Multiple Sclerosis (MS). Electrophoretic profiles are difficult to interpret due to low contrast and artefacts. A semi-automated method to ease analysis and to reduce subjectivity is presented. The method sequentially converts color images to grayscale, realigns bands, removes artifacts, then converts 2D images to a signal, before detecting, thresholding and editing peaks to optimize profiles. Such treated profiles (21 positive and 15 negative) are compared to ground truth analysis of an expert biologist. 16 profiles over 21 are well detected positive and 12 profiles over 15 are detected negative, results seem similar to inter-experts variability reported in literature.

摘要

检测脑脊液（CSF）中的寡克隆电泳带是多发性硬化症（MS）的一项重要诊断工具。由于对比度低和伪影，电泳图谱难以解读。本文提出了一种半自动方法，以简化分析并减少主观性。该方法先将彩色图像依次转换为灰度图像，重新对齐条带，去除伪影，然后将二维图像转换为信号，再进行峰值检测、阈值处理和编辑以优化图谱。将经过这种处理的图谱（21份阳性和15份阴性）与专业生物学家的真实分析结果进行比较。21份阳性图谱中有16份被准确检测为阳性，15份阴性图谱中有12份被检测为阴性，结果似乎与文献中报道的专家间差异相似。