Abstract
Image segmentation is a critical step in numerous medical imaging studies, which can be facilitated by automatic computational techniques. Supervised methods, although highly effective, require large training datasets of manually labeled images that are labor-intensive to produce. Unsupervised methods, on the contrary, can be used in the absence of training data to segment new images. We introduce a new approach to unsupervised image segmentation that is based on the computation of the local center of mass. We propose an efficient method to group the pixels of a one-dimensional signal, which we then use in an iterative algorithm for two- and three-dimensional image segmentation. We validate our method on a 2D X-ray image, a 3D abdominal magnetic resonance (MR) image and a dataset of 3D cardiovascular MR images.
Highlights
Numerous approaches to medical image segmentation have been proposed[1,2]
We introduce a new approach to unsupervised medical image segmentation, which is based on the computation of the one-dimensional (1D) local center of mass (CM)
We implemented our unsupervised segmentation algorithm in MATLAB and evaluated it on 2D and 3D medical images. We compared it with three other unsupervised segmentation algorithms: the watershed segmentation[5], a Gaussian-mixture-model-based hidden-Markov-random-field (GMM-HMRF) model[7] initialized with k-means clustering and the simple linear iterative clustering (SLIC) superpixel algorithm[8]
Summary
Numerous approaches to medical image segmentation have been proposed (see surveys)[1,2]. Provided that a large training dataset of a specific class of images – with ground-truth labels – is available, supervised segmentation can be effective in producing accurate results. Such a dataset can be exploited to train a deep neural network[3] or create probabilistic atlases[4], which can in turn generate accurate segmentations of new images in the same image class. By using the information from the entire signal, we group the pixels based on the CM of the regions where they are located We exploit this method for 2D and 3D image segmentation, by computing the local CMs of image pixels in many different orientations and iteratively updating the segmentation label of each pixel by choosing from the labels at its local CMs. We describe the proposed method in detail (the Methods section), present experimental results (the Results section), discuss them (the Discussion section) and conclude the paper (the Conclusion section)
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
