节点文献
面向中医舌诊的舌下静脉特征获取与分析
Feature Acquisition and Analysis of Sublingual Vein for Tongue Diagnosis in Traditional Chinese Medicine
【作者】 闫子飞;
【导师】 王宽全;
【作者基本信息】 哈尔滨工业大学 , 计算机应用技术, 2010, 博士
【摘要】 舌诊是中医学中最常用和最有临床应用价值的诊法之一。借助迅速发展的数字图像处理和模式识别技术,开展自动化舌象获取与分析研究可以避免传统舌诊非量化和主观性的缺点,使中医舌诊为更广阔的领域所认可。至今为止,关于舌面特征提取的研究已经广泛开展,但本领域研究人员对中医舌诊的重要组成部分舌脉这一舌下络脉诊法的主要诊察指标的研究却甚少涉及。本论文以面向中医诊断的舌下静脉的特征量化方法为研究中心,开展了舌下图像样本采集设备的研制、舌下静脉轮廓的分割和舌下静脉特征量化描述等方面的研究。获取能有效描述舌下静脉特征的舌下图像信号是开展舌下静脉特征量化研究的基础。本论文以舌下络脉诊断中的舌脉诊断特征为考虑因素,首先从获取舌脉颜色特征的角度出发,应用彩色舌图像采集设备获取彩色舌下图像,并对所获取的图像进行了预处理。同时,鉴于彩色舌下图像描述舌脉距离信息特征的局限性,结合当前在生物特征识别领域广泛应用的静脉识别原理,系统地探讨了近红外舌下图像获取中的光源、传感器等基础问题,形成了可行的近红外舌下图像信号获取策略,并研制出一套新型近红外舌下图像信号采集系统原型。然后,针对彩色舌下图像中舌脉与其周围舌质的对比度不同的特点,本论文分别提出了处理舌脉与舌质对比度较大的彩色舌下图像的基于像素聚类区域生长法的舌下静脉提取方法,以及针对舌脉与舌质对比度较小的彩色舌下图像的基于贝叶斯决策论的自适应舌脉轮廓提取方法。实验表明,所提出的方法针对舌脉与其周围舌质对比度的不同情况分别显现出较好的舌脉分割性能。应用本论文研制的便携式近红外舌下静脉图像采集仪获取的近红外舌下静脉图像,较彩色舌下图像能够更真实地反映舌脉的生理轮廓。本论文针对所获取的图像的自身特点提出了基于近红外舌下图像的舌下静脉提取方法,为更准确地量化舌脉的距离信息特征奠定了基础。在舌脉颜色特征的提取与量化方面,本论文以传统中医医师对舌脉颜色的判定结果为依据,构建了基于专家知识的舌脉颜色描述模型,进而给出了结合中医理论的舌脉颜色客观化描述;并应用基于KNN的像素级舌脉颜色分类方法在所构建的舌脉颜色模型基础上实现了舌脉颜色分类。实验结果表明,应用所提出的舌脉颜色模型及相应的舌脉颜色分类方法得到的舌脉颜色分类结果与中医诊断的舌脉颜色特征较为接近。在舌脉距离信息特征方面,提出了用于测量舌脉的长度及阔度两个具有中医诊断价值的距离特征的测量方法,并对本论文所应用的两种舌下图像视觉信号,彩色舌下图像和近红外舌下图像,进行了舌脉区域距离信息的测量实验。本论文在舌下图像信号的获取、舌下静脉轮廓提取以及舌下静脉特征量化等方面进行了系统的研究,使今后在舌诊客观化研究中应用多种富有诊断价值的舌下静脉特征辅助疾病的诊断成为可能。
【Abstract】 Tongue diagnosis is one of the most common and valuable diagnostic method in Traditional Chinese Medicine. By means of the rapid development of digital image processing and pattern recognition technology, research on automatic tongue image acquisition and analysis can avoid the disadvantage of non-quantification and subjectivity in traditional tongue diagnosis, and make it widely acknowledged. Till now, research on feature extraction of tongue surface has been gradually perfect. However, research on one of the most important part of tongue diagnosis, the sublingual vein diagnosis, and its primary diagnosing indicator, the sublingual vein, is rarely referred by the researchers of this field.This dissertation takes the feature quantification method of sublingual vein in traditional Chinese medicine as the research centre, carrying out researches including the design of sublingual image acquisition device, segmentation of sublingual vein and feature quantification of sublingual vein, respectively.Acquiring sublingual image signal, which can effectively describe the feature of sublingual vein, is the basis of researching on the feature quantification of sublingual vein. This dissertation takes the diagnostic feature of sublingual vein in sublingual vein diagnosis into consideration. Firstly, from the viewpoint of acquiring the chromatic feature of sublingual vein, applying the color tongue image acquisition device to acquire the color sublingual image, and then preprocessing the acquired images. Meanwhile, whereas the limit of color sublingual image in describing the distance information feature of sublingual vein, combing with the vein recognition principle popularized in current Biometrics field, this dissertation systematically investigate the basic problem in near infrared sublingual image acquisition, such as the light source and sensor, etc., standardized acquisition strategy of near infrared sublingual image is formed. Finally, a novel near infrared sublingual image acquisition prototype is developed.And then, focus on the different contrast between sublingual vein and surrounding tongue proper in color sublingual images, this dissertation proposes the pixel-clustering-based region growing algorithm for segmentation of sublingual veins from color sublingual image with visible contrast between sublingual vein and surrounding tongue proper, and the adaptive sublingual vein segmentation method based on Bayesian decision theory for color sublingual image with low contrast between sublingual vein and surrounding tongue proper, respectively. Experimental results reveal that, the proposed methods possess acceptable performance on sublingual vein segmentation for color sublingual image with different condition of contrast between sublingual vein and surrounding tongue proper.The near infrared sublingual images acquired by using the portable near infrared sublingual vein image acquisition device designed in this dissertation, can more truly reflect the physiological contour of sublingual vein than the color sublingual images. Segmentation method of sublingual vein from the near infrared sublingual image is proposed in this dissertation according to the self characteristic of near infrared sublingual images, which establishes the foundation for quantifying the distance information feature of sublingual veins more precisely.From the aspect of extraction and quantification of the color feature of sublingual veins, this dissertation takes the judgment results to the color of sublingual veins from Chinese physicians as the basis, establishes the color description model of sublingual vein based on expert knowledge, furthermore, proposes the objective description of the color of sublingual vein which take the theory of tongue diagnosis into consideration; moreover, pixel level classification of the color of sublingual vein based on KNN algorithm is applied to realize the classification of sublingual vein color on the basis of the established color description model of sublingual vein. Experimental results reveal that, the classification result of sublingual vein color obtained from the proposed model and the corresponding classification method are close to that given by Chinese physicians.Considering the distance information of sublingual veins, measurement method of the length and width of sublingual veins, which both possess diagnostic value in tongue diagnosis, is proposed. And then, experiments are implemented to measure the distance information of the sublingual vein regions in two kinds of sublingual images, color sublingual image and near infrared sublingual image.This dissertation systematically researches on the acquisition of sublingual images, segmentation of sublingual veins and feature quantification of sublingual veins, which make applying multiple sublingual vein features with diagnostic value to assist diseases diagnosing in the research on modernization of TCM tongue diagnosis come true in future.