Conference Paper

Fast Automatic Multi-atlas Segmentation of the Prostate from 3D MR Images

DOI: 10.1007/978-3-642-23944-1_2 Conference: Prostate Cancer Imaging. Image Analysis and Image-Guided Interventions - International Workshop, Held in Conjunction with MICCAI 2011, Toronto, Canada, September 22, 2011. Proceedings
Source: DBLP


A fast fully automatic method of segmenting the prostate from 3D MR scans is presented, incorporating dynamic multi-atlas label fusion. The diffeomorphic demons method is used for non-rigid registration and a comparison of alternate metrics for atlas selection is presented. A comparison of results from an average shape atlas and the multi-atlas approach is provided. Using the same clinical dataset and manual contours from 50 clinical scans as Klein et al. (2008) a median Dice similarity coefficient of 0.86 was achieved with an average surface error of 2.00mm using the multi-atlas segmentation method.

6 Reads
  • Source
    • "In atlas-based methods a pre-computed segmentation or prior information in a template space is propagated towards the image to be segmented via spatial normalization (registration). These methods have been largely used in brain MRI ( [26], [27] ), head and neck CT Scans ( [28], [29], [30] ), cardiac aortic CT [31], pulmonary lobes from CT [32] and prostate MR ( [33], [34] ) . In the atlas based methods image registration is a key element, as label propagation relies on the registration of one or more templates to a target image. "
    [Show abstract] [Hide abstract]
    ABSTRACT: In prostate cancer radiotherapy, the accurate identification of the prostate and organs at risk in planning computer tomography (CT) images is an important part of the therapy planning and optimization. Manually contouring these organs can be a time consuming process and subject to intra- and inter-expert variability. Automatic identification of organ boundaries from these images is challenging due to the poor soft tissue contrast. Atlas-based approaches may provide a priori structural information by propagating manual expert delineations to a new individual space; however the inter-individual variability and registration errors may lead to biased results. Multi-atlas approaches can partly overcome some of these difficulties by selecting the most similar atlases among a large data base but the definition of similarity measure between the available atlases and the query individual has still to be addressed. The purpose of this chapter is to explain atlas-based segmentation approaches and the evaluation of different atlas-based strategies to simultaneously segment prostate, bladder and rectum from CT images. A comparison between single and multiple atlases is performed. Experiments on atlas ranking, selection strategies and fusion decision rules are carried out to illustrate the presented methodology. Propagation of labels using two registration strategies are applied and the results of the comparison with manual delineations are reported.
    Full-text · Chapter · Nov 2013
  • Source
    • "The methods in this category primarily vary depending on the energy minimization framework. For example, in atlas-based methods, a model of the prostate is created from manually segmented training images and intensity difference between the model and a new un-segmented image is minimized (Klein et al., 2008; Dowling et al., 2011). In contrast, in region based level sets prior mean and standard deviation information of the prostate region from manually segmented images are used to maximize the distance between prostate and background regions depending on region based statistical moments and propagate an implicitly defined deformable model whose energy is minimized at the zone of convergence of the two regions (Costa et al., 2007; Rousson et al., 2005; Chen et al., 2009). "
    [Show abstract] [Hide abstract]
    ABSTRACT: Prostate segmentation aids in prostate volume estimation, multi-modal image registration, and to create patient specific anatomical models for surgical planning and image guided biopsies. However, manual segmentation is time consuming and suffers from inter-and intra-observer variabilities. Low contrast images of trans rectal ultrasound and presence of imaging artifacts like speckle, micro-calcifications, and shadow regions hinder computer aided automatic or semi-automatic prostate segmentation. In this paper, we propose a prostate segmentation approach based on building multiple mean parametric models derived from principal component analysis of shape and posterior probabilities in a multi-resolution framework. The model parameters are then modified with the prior knowledge of the optimization space to achieve optimal prostate segmentation. In contrast to traditional statistical models of shape and intensity priors, we use posterior probabilities of the prostate region determined from random forest classification to build our appearance model, initialize and propagate our model. Furthermore, multiple mean models derived from spectral clustering of combined shape and appearance parameters are applied in parallel to improve segmentation accuracies. The proposed method achieves mean Dice similarity coefficient value of 0.91±0.09 for 126 images containing 40 images from the apex, 40 images from the base and 46 images from central regions in a leave-one-patient-out validation framework. The mean segmentation time of the procedure is 0.67±0.02s.
    Full-text · Article · Apr 2013 · Medical image analysis
  • [Show abstract] [Hide abstract]
    ABSTRACT: This paper presents a novel approach to automatically segment the prostate (including seminal vesicles) using a surface that is actively deformed via shape and gray level models. The surface deformation process utilises the results of a multi-atlas registration approach, where training images are matched to the case image via non-rigid registration. Normalised mutual information is then used to measure the similarity between each image in the training set and the case image. The set of training images with a similarity greater than a threshold is then used to build the initialisation and the gray level model of the segmentation process. This case specific gray level model is used to deform the initial surface to more closely match the prostate boundary via normalised cross-correlation based template matching of gray level profiles. Mean and median Dice's Similarity Coefficients of 0.849 and 0.855, as well as a mean surface error of 2.11 mm, were achieved when segmenting 3T Magnetic Resonance clinical scans of fifty patients.
    No preview · Conference Paper · Dec 2011
Show more