基于多光谱成像的宫颈细胞实时筛查系统

RealTime Screening System of Cervical Cells Based on Multispectral Imaging

摘要:

目前常用的宫颈癌细胞筛查方法有TBS分类法和细胞DNA定量分析法,TBS分类法主要是观察细胞形态是否发生异常改变;而DNA定量分析法是统计细胞核内的DNA含量是否发生增值,两种方法需要在两张不同的细胞涂片上分别展开。针对此,设计了一种在复合染色的环境下利用多光谱成像技术的宫颈细胞实时筛查系统。该系统实现了在一张细胞涂片上同时使用TBS分类法和DNA定量分析法进行宫颈癌细胞的筛查。其主要评价参数是细胞核内的DNA含量(即DI指数),同时可以观察细胞形态。该系统可以测量各个细胞核的DI值并且获得整个细胞的彩色图像,从而辅助医生对DI值大于2.5的异常细胞进行定位筛查。实验表明,该系统具有快速、准确和高效的特点,敏感性超过95%,克服了涂片质量差、判断阳性率低的缺点。

Abstract:

The present common methods for screening cervical cancer cells are TBS classification and cell DNA quantitative analysis. The former is carried out by observing whether the cell morphology changes abnormally, and the latter is performed by recording whether or not the DNA content of the nuclei is increased. The methods need to be carried out separately on two different cell smears. In this paper, a realtime screening system for cervical cells using multispectral imaging technology in a complex dyeing environment was proposed. Screening of cervical cancer cells was performed on a cell smear using TBS classification and DNA quantitative analysis. The main evaluation parameters are the DNA content in the nucleus (ie, the DI index), and the morphology of the cells can be observed. Each nucleus has its own DI measurements and color images of the entire cell, allowing physicians to screen for abnormal cells with a DI value of higher than 2.5. Experiments show that the system has the characteristics of speediness, high accuracy and efficiency, and the sensitivity is over 95%, which can overcome the shortcomings of poor smear quality and low positive rate.