Research On Quantitative Analysis Of Lymphocyte Morphology Deformation In Microscopic Images

Microscopic cell imaging plays an important role throughout the clinical test. In the traditional process, doctors diagnose rejection by counting cells and observing the deformation and motion of lymphocytes in patients’ peripheral blood using microscopes, the progress is actually tedious, subjective and empirical. With the increasing development of medical imaging technology and the resolution of image, microscopic cell imaging has been gradually developed from traditional two-dimensional static image to video image(2 dimension image sequences). These novel technologies generate a wealth of data that require automated image processing techniques to turn qualitative data into quantitative values, which brings challenge of multi-dimension data analysis as well as provides exciting and novel insights into cell biology. How to obtain the quantitative and objective features, which can give new insights to the relevant biological processes, from huge amounts of image data by using digital image processing and pattern recognition technology is one of the challenge in the development of medical imaging in clinical disease diagnosis. The dissertation studied measurement of live lymphocyte morphology and intracellular motion(motility) in microscopic images acquired from peripheral blood of mice post skin transplantation. Main research and innovation points are as follows:1. This dissertation proposed an improved segmentation method for cell image based on active contour models. The accurate segmentation and tracking of lymphocyte boundaries is one of the prerequisites for the quantitative analysis of cell morphology and motility. Segmentation and tracking of cell boundaries in phase-contrast microscopic images are challenging due to non-uniform edge intensities and the phase halos. Active contour models, pulling the initial contour to the desired boundaries with various kinds of external force, have shown prominent advantages in object segmentation and tracking. However, current external-force-inspired methods are weak at handling low-contrast edges, and suffer from initialization sensitivity. To alleviate initial location sensitivity, we set the initial contour close to the real boundaries by performing morphological image processing to the image. In order to segment low-contrast boundaries, this dissertation proposed an improved external force field for the evolution of active contours. The proposed external force field combine the region and the edge information of the object at the same time, which got good effects. The experimental results show that the proposed method can accurately segment and track lymphocyte boundaries in microscopic images even in the presence of low-contrast edges.2. A novel algorithm for quantitative analysis of intracellular motility is proposed based on variation optical flow model. In the improved variation optical flow model, the data term is in the form of 2L norm; the smoothness data is controlled by a self-adaptive parameter, using the 2L norm in the intracellular area and 1L norm at the edge of the cell, which can be used to extract intracellular smooth movement fields and deal with the border problems of optical flow. Histograms of oriented optical flow(HOOF) is applied for quantitative analysis of intracellular movement, and the distance of the subsequent HOOF is used as the intracellular movement features. Experimental results show that features extracted based on the improved variation optical flow model and HOOF can effectively distinguish the degree of the intracellular movement.3. A scheme is proposed for quantitatively analyzing of dynamic behavior of lymphocytes in microscopic image sequences, and explored the shape, deformation and intracellular motion features of live lymphocytes. We extracted features of the dynamic behavior of lymphocyte based on the shape, deformation and intracellular motion features from two kinds of data which corresponding to the abnormal and normal lymphocytes. The Wilcoxon rank sum test is applied to choose the best features to form an optimal feature vector. PNN is used to classify lymphocytes dynamic behavior based on the feature vector. Experimental results show that the scheme of combing the shape, deformation and intracellular motion features achieved a better performance than the only shape, deformation or intracellular motion feature method. The experimental results agree well with clinicians’ observation: the activity of lymphocytes is enhanced during graft reaction, i.e., the lymphocytes in the abnormal group have dramatic deformation, while the lymphocytes in the normal group are more stable. The proposed method leads to its potential applications to interventional strategies for the graft rejection.4. The study of shape signature of cell boundary in microscopic images. We analyzed cell deformation based on Contourlet transform. First, the cell boundary was extracted and the shape signature was computed based on it. Then a surface was obtained after the shape signature signal was spread out over a period of time. The wave of the surface denotes the deformation of cell boundary in all directions. After that, the Wavelet and Contourlet transform were applied to the shape signature signal. The Contourlet transform can effectively capture smooth contours that are the dominant feature in natural images, and large coefficients denotes the local feature of the special orientation and position, such as edges and contours. Finally, statistical analysis was applied to the sub-band coefficients, and we apply Wilcoxon rank sum test to identify the extracted features that have the best power in distinguishing the two kinds of data. Experimental results verify the effectiveness of the algorithm.