Contour Inference-based And Sparse Learning-based Approaches To Individual Cell Segmentation In Complex Pathology Images

With improvements in computer image processing techiniques and rapid development of clinical medicine,the pathology image-based automated nucleus/cell detection,counting,and segmentation problems become apparent in digital pathology-aided diagnosis and present significant challenges.On one hand,due to the complex background structures of pathology images,highly dense distribution of cells/nuclei,and cell/nucleus adhesion,accurate individual cellular/nuclear segmentation is an important research topic.On the other hand,the focus of the past literature is dominated by either segmenting a certain type of cells/nuclei or simply splitting clustered objects without contours inference of them.Achieving individuallevel accuracy for detection,segmentation,and inference of densely overlapping cells/ nuclei from complex pathology image and allowing the computers to understand the details of pathology images are important research directions in computer vision and image processing.In this dissertation,we mainly focus on contour inference and sparse learning methods for individual cell/nucleus detection and segmentation in complex pathology images.The major contributions of the paper are:(1)A boundary-to-marker evidence-controlled segmentation and minimum description length-based contour inference method is proposed for overlapping nuclei.Combining the initial segmentation information and concavity measurement,the proposed method first segments clusters of nuclei into individual pieces,avoiding segmentation errors introduced by the scale-constrained Laplacian-of-Gaussian filtering.After that a nuclear boundary-to-marker evidence computing is introduced to delineate individual objects.The obtained evidence set is then modeled by the periodic B-splines with the minimum description length criterion,which achieves accurate complete contour inference of individual nuclei.The algorithm has been tested on the synthetic database as well as real histopathology images.By comparing the proposed method with several state-of-the-art methods,experimental results show the superior recognition performance of our method and indicate the potential applications of analyzing the intrinsic features of nuclei morphology(2)A novel contour-seed pairs learning-based framework for robust and automated cell/nucleus segmentation is proposed.Our method addresses the issues of detection and segmentation by formulating these two tasks in terms of a unified regression problem,where a cascade sparse regression chain model is trained and then applied to return object locations and entire boundaries of clustered objects.In particular,we first learn a set of sparse convolutional features in each layer.Then,in the proposed chain,with the input from the learned features,we iteratively update the locations and clustered object boundaries until convergence.In this way,the boundary evidences of each individual object can be easily delineated and be further fed to a complete contour inference procedure optimized by the minimum description length principle.For any probe image,our method enables to analyze free-lying and overlapping cells with complex shapes.Experimental results show that the proposed method is very generic and performs well on contour inferences of various cell/nucleus types.Compared with the current segmentation techniques,our approach achieves state-of-the-art performances on four challenging datasets.(3)The problems of existing techniques are sovled by reformulating nuclear segmentation in terms of a cascade 2-class classification problem and a multi-layer boosting sparse convolutional model is proposed.The proposed model is constructed by stacking several layers of boosting weak learners as its building blocks,in which discriminative probabilistic binary decision trees are designed as weak learners in each layer to cope with challenging cases.A sparsity-constrained cascade structure enables the ML-BSC model to do better representation learning.Comparing to the existing techniques,our method can accurately separate individual nuclei in complex histopathology images,and it is more robust against chromatin-sparse and heavy background clutter.An evaluation carried out using three disparate datasets demonstrates the superiority of our method over the state-of-the-art supervised approaches in terms of segmentation accuracy.(4)A new method to automatically segment nuclei from complex pathology images with cascade sparse convolution and decision tree ensemble model is proposed.In particular,the sparse separable convolution learning module and the decision tree ensemble learning module are stacked in a cascaded manner to form the model.The former adopts rank-one tensor decomposition learning mechanism that can extract multiscale and multi-directional distributed abstract features;while the latter employs regularized regression mechanism to boost per-pixel regression or classification performance.Compared with the existing deep networks,the proposed model does not require nonlinear activation and backpropagation computation and depends on few parameters.Our model is trained in a layer-wise manner that can achieve an end-to-end pixelwise learning and whose overall processing is simple,fast and does not requires handcrafted feature extraction.We demonstrated the superiority of our segmentation technique by evaluating it on the multi-disease state and multi-organ dataset where consistently higher performance was obtained as compared to convolutional neural networks and fully convolutional networks.