基于机器视觉的粪便白细胞自动检测动态库设计与实现

发布时间：2018-05-26 06:03

本文选题：白细胞 + 自动检测　；参考：《山东大学》2015年硕士论文

【摘要】：粪便是人类的主要分泌物,正常的粪便中约3/4是水,剩下1/4为固体,而固体中成分主要为未消化的食物、消化道分泌物和有形成分、食物分解产物和细菌。健康的人粪便一般无白细胞或偶见白细胞,存在的白细胞为中性粒细胞,形态完整者与血液中的粒细胞无差别。若检测到白细胞的数量、形态异常,则机体出现炎症。粪便中的生物白细胞的数量和状态可以作为判断人机体是否存在疾病的先行判决条件,而且可以给医师进一步检测提供方向和线索。医学粪便检测、尤其是白细胞对肠胃等疾病的早期诊断具有良好的效果。现阶段医院粪便白细胞检查均为人工显微镜识别,效率低、定性分析、主观因素大,进而导致不能及时在中早期发现疾病。同时,医院粪便白细胞检测量大、相关资源紧张,导致医患关系紧张,带来不利的社会效益。通过展开粪便白细胞自动检测技术研究,可以解决目前医患关系紧张、检测精度不好等,具有很大的学术和社会价值。同时智能医疗设备也是现代医疗发展趋势所在,相关领域具有良好的前景。虽然粪便白细胞自动检测具有重要价值,但是粪便白细胞检测由于检测目标图样背景的异常复杂及白细胞的形态多变等问题,使得粪便白细胞的定量自动检测存在较大难度。本文针对上述问题,展开了如下几个方面的研究,并取得了较好的研究成果,达到了医学实际使用的目标：a)统计分析了粪便中白细胞特性并进行了相关特性研究。通过查阅相关文献资料、分析样品图样中白细胞的形态结构特性,获取设计算法所需的特征信息。研究过程中,统计分析完5000份样本(每样品有低倍镜图5张,高倍镜图15张)b)统计分析了研究粪便样品图样背景特性及红细胞等其他细胞的特征信息,以便在复杂背景中正确的识别目标物-白细胞,并避免红细胞等对白细胞检测的干扰,为设计白细胞识别算法准备判断信息。研究过程汇总,统计分析完5000份样本(每样品有低倍镜图5张,高倍镜图15张)c)研究设计了基于白细胞特征、粪便样品图样的背景信息及粪便中其他细胞的特征信息等的粪便白细胞定量自动检测算法。在医院实际使用时,实现了针对粪便样本的白细胞阴阳性灵敏度、特异性均高于90%。d)开发了基于本文开发的粪便白细胞自动检测算法的粪便白细胞自动检测动态库开发,并实际满足了相关医疗设备在医院的使用。
[Abstract]:Feces are the main secretions of human beings. About three-quarters of the normal feces are water, and the remaining quarter are solid, while the components in solids are mainly undigested food, digestive tract secretions and tangible components, food decomposition products and bacteria. In healthy human faeces, there are no or occasional white blood cells, and the existing leukocytes are neutrophils, and there is no difference between those with intact morphology and the granulocytes in the blood. If the number of white blood cells detected, abnormal morphology, the body appears inflammation. The quantity and state of biological white blood cells in feces can be used as the first judgment condition to judge the existence of human body disease, and can provide direction and clue for doctors to further test. Medical stool detection, especially white blood cells, has a good effect on the early diagnosis of gastrointestinal diseases. At present, the leucocyte examination of feces in hospital is identified by artificial microscope, which has low efficiency, qualitative analysis and large subjective factors, which leads to the failure to find disease in the middle and early stage in time. At the same time, the measurement of leucocyte in hospital feces is large and the related resources are tight, which leads to the shortage of doctor-patient relationship and brings unfavorable social benefits. It is of great academic and social value to carry out the research on automatic detection of leucocyte in feces, which can solve the tension between doctors and patients and the poor precision of detection. At the same time, intelligent medical equipment is also the development trend of modern medical treatment, related fields have a good prospect. Although the automatic detection of leucocyte in feces is of great value, the quantitative automatic detection of leucocyte in feces is difficult because of the complexity of the background of the target pattern and the variety of the morphology of the white blood cell. In order to solve the above problems, this paper has carried out the following research, and achieved the goal of medical practice. The white blood cell characteristics in feces were statistically analyzed and the related characteristics were studied. By consulting the relevant literature, the morphological and structural characteristics of white blood cells in the sample pattern were analyzed, and the characteristic information needed for the design algorithm was obtained. During the course of the study, 5000 samples (5 low-power and 15 high-power) were statistically analyzed for the background characteristics of fecal samples and the characteristic information of other cells, such as red blood cells. In order to correctly identify the target white blood cell in the complex background, and avoid the interference of red blood cell on the white blood cell detection, and prepare the judgment information for the design of the white blood cell recognition algorithm. The research process was summarized, and 5000 samples (5 low-power and 15 high-power microscopes) were studied and designed based on white blood cell characteristics. The quantitative automatic detection algorithm of leucocyte in fecal samples based on background information of fecal samples and characteristic information of other cells in feces. In the actual use of the hospital, the sensitivity of white blood cells in fecal samples was realized, and the specificity was higher than 90. D) A dynamic library for automatic detection of fecal white blood cells was developed based on the automatic detection algorithm of fecal white blood cells developed in this paper. And the actual use of medical equipment in the hospital is satisfied.
【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：R446.1;TP391.41

【参考文献】