Segmentation Of Anatomical Structures Research Articles

Segmentation Methods of Echocardiography Images for Left Ventricle Boundary Detection

Due to acoustic interferences and artifacts which are inherent in echocardiography images, automatic segmentation of anatomical structures in cardiac ultrasound images is a real challenge. This paper surveys state-of-the-art researches on echocardiography data segmentation methods, concentrating on methods techniques developed for clinical data. We present a classification of methodologies for echocardiography image segmentation. By choosing ten recent papers which have proposed innovative ideas that they proved certain clinical advantages or potential especial role to the echocardiography segmentation task. The contribution of the paper would be serving as a tutorial of the field for both clinicians and technologists, providing large number of segmentation techniques in a comprehensive and systematic manner and critically review recent approaches in terms of their performance and degree of clinical evaluation with respect to the final goal of cardiac functional analysis.

An Optimized PatchMatch for multi-scale and multi-feature label fusion

Automatic segmentation methods are important tools for quantitative analysis of Magnetic Resonance Images (MRI). Recently, patch-based label fusion approaches have demonstrated state-of-the-art segmentation accuracy. In this paper, we introduce a new patch-based label fusion framework to perform segmentation of anatomical structures. The proposed approach uses an Optimized PAtchMatch Label fusion (OPAL) strategy that drastically reduces the computation time required for the search of similar patches. The reduced computation time of OPAL opens the way for new strategies and facilitates processing on large databases. In this paper, we investigate new perspectives offered by OPAL, by introducing a new multi-scale and multi-feature framework. During our validation on hippocampus segmentation we use two datasets: young adults in the ICBM cohort and elderly adults in the EADC-ADNI dataset. For both, OPAL is compared to state-of-the-art methods. Results show that OPAL obtained the highest median Dice coefficient (89.9% for ICBM and 90.1% for EADC-ADNI). Moreover, in both cases, OPAL produced a segmentation accuracy similar to inter-expert variability. On the EADC-ADNI dataset, we compare the hippocampal volumes obtained by manual and automatic segmentation. The volumes appear to be highly correlated that enables to perform more accurate separation of pathological populations.

A technique for semiautomatic segmentation of echogenic structures in 3D ultrasound, applied to infant hip dysplasia.

Automatic segmentation of anatomical structures and lesions from medical ultrasound images is a formidable challenge in medical imaging due to image noise, blur and artifacts. In this paper we present a segmentation technique with features highly suited to use in noisy 3D ultrasound volumes and demonstrate its use in modeling bone, specifically the acetabulum in infant hips. Quantification of the acetabular shape is crucial in diagnosing developmental dysplasia of the hip (DDH), a common condition associated with hip dislocation and premature osteoarthritis if not treated. The well-established Graf technique for DDH diagnosis has been criticized for high inter-observer and inter-scan variability. In our earlier work we have introduced a more reliable instability metric based on 3D ultrasound data. Visualizing and interpreting the acetabular shape from noisy 3D ultrasound volumes has been one of the major roadblocks in using 3D ultrasound as diagnostic tool for DDH. For this study we developed a semiautomated segmentation technique to rapidly generate 3D acetabular surface models and classified the acetabulum based on acetabular contact angle (ACA) derived from the models. We tested the feasibility and reliability of the technique compared with manual segmentation. The proposed segmentation algorithm is based on graph search. We formulate segmentation of the acetabulum as an optimal path finding problem on an undirected weighted graph. Slice contours are defined as the optimal path passing through a set of user-defined seed points in the graph, and it can be found using dynamic programming techniques (in this case Dijkstra's algorithm). Slice contours are then interpolated over the 3D volume to generate the surface model. A three-dimensional ACA was calculated using normal vectors of the surface model. The algorithm was tested over an extensive dataset of 51 infant ultrasound hip volumes obtained from 42 subjects with normal to dysplastic hips. The contours generated by the segmentation algorithm closely matched with those obtained from manual segmentation. The average RMS errors between the semiautomated and manual segmentation for the 51 volumes were 0.28 mm/1.1 voxel (with 2 node points) and 0.24 mm/0.9 voxel (with 3 node points). The semiautomatic algorithm gave visually acceptable results on images with moderate levels of noise and was able to trace the boundary of the acetabulum even in the presence of significant shadowing. Semiautomatic contouring was also faster than manual segmentation at 37 versus 56 s per scan. It also improved the repeatability of the ACA calculation with inter-observer and intra-observer variability of 1.4 ± 0.9 degree and 1.4 ± 1.0 degree. The semiautomatic segmentation technique proposed in this work offers a fast and reliable method to delineate the contours of the acetabulum from 3D ultrasound volumes of the hip. Since the technique does not rely upon contour evolution, it is less susceptible than other methods to the frequent missing or incomplete boundaries and noise artifacts common in ultrasound images. ACA derived from the segmented 3D surface was able to accurately classify the acetabulum under the categories normal, borderline and dysplastic. The semiautomatic technique makes it easier to segment the volume and reduces the inter-observer and intra-observer variation in ACA calculation compared with manual segmentation. The method can be applied to any structure with an echogenic boundary on ultrasound (such as a ventricle, blood vessel, organ or tumor), or even to structures with a bright border on computed tomography or magnetic resonance imaging.

Automatic Localization of the Optic Disc Center in Retinal Images based on Angle Detection in Curvature Scale Space

Digital images of the retina is widely used for screening of patients suffering from sight threatening diseases such as Diabetic retinopathy and Glaucoma. The localization of the Optic Disc (OD) center is the first and necessary step identification and segmentation of anatomical structures and in pathological retinal images. From the center of the optic disc spreads the major blood vessels of the retina. Therefore, by considering the high number of vessels and the high number of the angles resulted from the vessels crossing, the authors propose a new method based on the number of angles in the vicinity of optic disc for localization of the center of optic disc. The first step is pre-processing of retinal image for separate the fundus from its background and increase the contrast between contours. In the second step, the authors use the Curvature Scale Space (CSS) for angle detection. In the next step, they move a window about the size of optic disc to count the number of corners. In the final step, they use the center of windows which has the most number of corners for localizing the optic disc center. The proposed method is evaluated on DRIVE, CHASE_DB1 and STARE databases and the success rate is 100, 100 and 96.3%, respectively.

Segmentation of brain magnetic resonance images based on multi-atlas likelihood fusion: testing using data with a broad range of anatomical and photometric profiles.

We propose a hierarchical pipeline for skull-stripping and segmentation of anatomical structures of interest from T1-weighted images of the human brain. The pipeline is constructed based on a two-level Bayesian parameter estimation algorithm called multi-atlas likelihood fusion (MALF). In MALF, estimation of the parameter of interest is performed via maximum a posteriori estimation using the expectation-maximization (EM) algorithm. The likelihoods of multiple atlases are fused in the E-step while the optimal estimator, a single maximizer of the fused likelihoods, is then obtained in the M-step. There are two stages in the proposed pipeline; first the input T1-weighted image is automatically skull-stripped via a fast MALF, then internal brain structures of interest are automatically extracted using a regular MALF. We assess the performance of each of the two modules in the pipeline based on two sets of images with markedly different anatomical and photometric contrasts; 3T MPRAGE scans of pediatric subjects with developmental disorders vs. 1.5T SPGR scans of elderly subjects with dementia. Evaluation is performed quantitatively using the Dice overlap as well as qualitatively via visual inspections. As a result, we demonstrate subject-level differences in the performance of the proposed pipeline, which may be accounted for by age, diagnosis, or the imaging parameters (particularly the field strength). For the subcortical and ventricular structures of the two datasets, the hierarchical pipeline is capable of producing automated segmentations with Dice overlaps ranging from 0.8 to 0.964 when compared with the gold standard. Comparisons with other representative segmentation algorithms are presented, relative to which the proposed hierarchical pipeline demonstrates comparative or superior accuracy.

Progressive Label Fusion Framework for Multi-atlas Segmentation by Dictionary Evolution.

Accurate segmentation of anatomical structures in medical images is very important in neuroscience studies. Recently, multi-atlas patch-based label fusion methods have achieved many successes, which generally represent each target patch from an atlas patch dictionary in the image domain and then predict the latent label by directly applying the estimated representation coefficients in the label domain. However, due to the large gap between these two domains, the estimated representation coefficients in the image domain may not stay optimal for the label fusion. To overcome this dilemma, we propose a novel label fusion framework to make the weighting coefficients eventually to be optimal for the label fusion by progressively constructing a dynamic dictionary in a layer-by-layer manner, where a sequence of intermediate patch dictionaries gradually encode the transition from the patch representation coefficients in image domain to the optimal weights for label fusion. Our proposed framework is general to augment the label fusion performance of the current state-of-the-art methods. In our experiments, we apply our proposed method to hippocampus segmentation on ADNI dataset and achieve more accurate labeling results, compared to the counterpart methods with single-layer dictionary.

Medical image segmentation on GPUs--a comprehensive review.

Segmentation of anatomical structures, from modalities like computed tomography (CT), magnetic resonance imaging (MRI) and ultrasound, is a key enabling technology for medical applications such as diagnostics, planning and guidance. More efficient implementations are necessary, as most segmentation methods are computationally expensive, and the amount of medical imaging data is growing. The increased programmability of graphic processing units (GPUs) in recent years have enabled their use in several areas. GPUs can solve large data parallel problems at a higher speed than the traditional CPU, while being more affordable and energy efficient than distributed systems. Furthermore, using a GPU enables concurrent visualization and interactive segmentation, where the user can help the algorithm to achieve a satisfactory result. This review investigates the use of GPUs to accelerate medical image segmentation methods. A set of criteria for efficient use of GPUs are defined and each segmentation method is rated accordingly. In addition, references to relevant GPU implementations and insight into GPU optimization are provided and discussed. The review concludes that most segmentation methods may benefit from GPU processing due to the methods' data parallel structure and high thread count. However, factors such as synchronization, branch divergence and memory usage can limit the speedup.

Automatic Segmentation of Anatomical Structures using Deformable Models and Bio-Inspired/Soft Computing

This PhD dissertation is focused on the development of algorithms for the automatic segmentation of anatomical structures in biomedical images, usually the hippocampus in histological images from the mouse brain. Such algorithms are based on computer vision techniques and artificial intelligence methods. More precisely, on the one hand, we take advantage of deformable models to segment the anatomical structure under consideration, using prior knowledge from different sources, and to embed the segmentation into an optimization framework. On the other hand, metaheuristics and classifiers can be used to perform the optimization of the target function defined by the shape model (as well as to automatically tune the system parameters), and to refine the results obtained by the segmentation process, respectively. Three new different methods, with their corresponding advantages and disadvantages, are described and tested. A broad theoretical discussion, together with an extensive introduction to the state of the art, has also been included to provide an overview necessary for understanding the developed methods.

A Higher-Order Neural Network Design for Improving Segmentation Performance in Medical Image Series

Segmentation of anatomical structures from medical image series is an ongoing field of research. Although, organs of interest are three-dimensional in nature, slice-by-slice approaches are widely used in clinical applications because of their ease of integration with the current manual segmentation scheme. To be able to use slice-by-slice techniques effectively, adjacent slice information, which represents likelihood of a region to be the structure of interest, plays critical role. Recent studies focus on using distance transform directly as a feature or to increase the feature values at the vicinity of the search area. This study presents a novel approach by constructing a higher order neural network, the input layer of which receives features together with their multiplications with the distance transform. This allows higher-order interactions between features through the non-linearity introduced by the multiplication. The application of the proposed method to 9 CT datasets for segmentation of the liver shows higher performance than well-known higher order classification neural networks.

Integrating edge detection and fuzzy connectedness for automated segmentation of anatomical branching structures

Image segmentation algorithms are critical components of medical image analysis systems. This paper presents a novel and fully automated methodology for segmenting anatomical branching structures in medical images. It is a hybrid approach which integrates the Canny edge detection to obtain a preliminary boundary of the structure and the fuzzy connectedness algorithm to handle efficiently the discontinuities of the returned edge map. To ensure efficient localisation of weak branches, the fuzzy connectedness framework is applied in a sliding window mode and using a voting scheme the optimal connection point is estimated. Finally, the image regions are labelled as tissue or background using a locally adaptive thresholding technique. The proposed methodology is applied and evaluated in segmenting ductal trees visualised in clinical X-ray galactograms and vasculature visualised in angiograms. The experimental results demonstrate the effectiveness of the proposed approach achieving high scores of detection rate and accuracy among state-of-the-art segmentation techniques.

Optimized patchMatch for near real time and accurate label fusion.

Automatic segmentation methods are important tools for quantitative analysis of magnetic resonance images. Recently, patch-based label fusion approaches demonstrated state-of-the-art segmentation accuracy. In this paper, we introduce a new patch-based method using the PatchMatch algorithm to perform segmentation of anatomical structures. Based on an Optimized PAtchMatch Label fusion (OPAL) strategy, the proposed method provides competitive segmentation accuracy in near real time. During our validation on hippocampus segmentation of 80 healthy subjects, OPAL was compared to several state-of-the-art methods. Results show that OPAL obtained the highest median Dice coefficient (89.3%) in less than 1 sec per subject. These results highlight the excellent performance of OPAL in terms of computation time and segmentation accuracy compared to recently published methods.

Automatic segmentation of anatomical structures from CT scans of thorax for RTP.

Modern radiotherapy techniques are vulnerable to delineation inaccuracies owing to the steep dose gradient around the target. In this aspect, accurate contouring comprises an indispensable part of optimal radiation treatment planning (RTP). We suggest a fully automated method to segment the lungs, trachea/main bronchi, and spinal canal accurately from computed tomography (CT) scans of patients with lung cancer to use for RTP. For this purpose, we developed a new algorithm for inclusion of excluded pathological areas into the segmented lungs and a modified version of the fuzzy segmentation by morphological reconstruction for spinal canal segmentation and implemented some image processing algorithms along with them. To assess the accuracy, we performed two comparisons between the automatically obtained results and the results obtained manually by an expert. The average volume overlap ratio values range between 94.30 ± 3.93% and 99.11 ± 0.26% on the two different datasets. We obtained the average symmetric surface distance values between the ranges of 0.28 ± 0.21 mm and 0.89 ± 0.32 mm by using the same datasets. Our method provides favorable results in the segmentation of CT scans of patients with lung cancer and can avoid heavy computational load and might offer expedited segmentation that can be used in RTP.

Semi-supervised clustering for MR brain image segmentation

Magnetic resonance (MR) brain image segmentation of different anatomical structures or tissue types has become a critical requirement in the diagnosis of neurological diseases. Depending on the availability of the training samples, image segmentation can be either supervised or unsupervised. While supervised learning requires a sufficient amount of labelled training data, which is expensive and time-consuming, unsupervised learning techniques suffer from the problem of local traps. Semi-supervised algorithms that includes prior knowledge into the unsupervised learning can enhance the segmentation process without the need of labelled training data. This paper proposes a method to improve the quality of MR brain tissue segmentation and to accelerate the convergence process. The proposed method is a clustering based semi-supervised classifier that does not need a set of labelled training data and uses less human expert analysis than a supervised approach. The proposed classifier labels the voxels clusters of an image slice and then uses statistics and class labels information of the resultant clusters to classify the remaining image slices by applying Gaussian Mixture Model (GMM). The experimental results show that the proposed semi-supervised approach accelerates the convergence and improves the results accuracy when comparing with the classical GMM approach.

Automatic Segmentation Framework of Building Anatomical Mouse Model for Bioluminescence Tomography

Bioluminescence tomography is known as a highly ill-posed inverse problem. To improve the reconstruction performance by introducing anatomical structures as a priori knowledge, an automatic segmentation framework has been proposed in this paper to extract the mouse whole-body organs and tissues, which enables to build up a heterogeneous mouse model for reconstruction of bioluminescence tomography. Finally, an in vivo mouse experiment has been conducted to evaluate this framework by using an X-ray computed tomography system and a multi-view bioluminescence imaging system. The findings suggest that the proposed method can realize fast automatic segmentation of mouse anatomical structures, ultimately enhancing the reconstruction performance of bioluminescence tomography. DOI: http://dx.doi.org/10.11591/telkomnika.v11i9.3241

TH-C-WAB-01: BEST IN PHYSICS (JOINT IMAGING-THERAPY)-Semi-Automated Probabilistic Segmentation of Head and Neck Anatomy Through Structure Specific Feature Selection From Multi-Sequence MRI

Purpose: To develop semi‐automatic methods for robust segmentation of head and neck anatomy through structure‐specific image features selection from multiple MRI sequences. Methods: We developed a semi‐automatic approach for probabilistic segmentation of head and neck anatomy that combines several candidate segmentations using random forests and structure‐specific image features selection from multiple MRI sequences. Starting from user‐drawn line segments on a single axial image, in each tree, structure segmentations are generated from composite feature images computed at the tree split nodes. Composite features are computed through random combinations of the MRI sequences and image features, including texture and Laplacian of Gaussian gradients. We used seven different MRI pulse sequences obtained from a Phillips scanner, including Dixon water only T1, Dixon fat only T1, Dixon in‐phase, Dixon opposed‐phase, T1‐weighted, T2‐weighted and diffusion weighted whole body imaging with background body signal suppression (DWIBS). The highest quality anatomical structure segmentation in a tree is extracted from all available split node segmentations with segmentation quality measured using within‐segment statistics including kurtosis, edge quality, and inter‐segment dissimilarity. The joint segmentation with probabilistic score is obtained by combining all the generated tree segmentations. Results: We used our approach to segment anatomical structures in head and neck images using the seven MRI sequences listed above. The algorithm returned the segmentations, their probabilistic scores, and the highest quality feature images for each structure. We evaluated our approach by computing DICE overlap scores with manually segmented ground truths and obtained an overall segmentation accuracy of 80% for the various radiotherapeutically‐relevant anatomic structures. Conclusion: We developed a semi‐automatic, random forest‐based probabilistic segmentation of head and neck anatomy by combining multiple MRI pulse sequences. Our approach achieves accurate segmentation for various radiotherapeutically‐relevant anatomic structures. Additionally, it extracts the highest quality composite feature images for obtaining optimal segmentation for the same structures. Phillips

Simultaneous Intensity Inhomogensity Correction, Registration and Segmentation of Anatomical Structures From Brain MR Images

Accurate segmentation for magnetic resonance (MR) images is an essential step in quantitative brain image analysis, and hence has attracted extensive research attention. However, due to the existence of noise and intensity inhomogeneity, also named as bias field, many segmentation methods suffer from limited accuracy. This paper presents a novel variational framework for the registration, segmentation and bias estimation simultaneously. We first presented an improved segmentation model by using the intensity statistic distributions with different means and variances in local regions. The model can estimate the bias field meanwhile segmenting images. We also proposed an anisotropic non-rigid registration method by using the structure tensor information and nonlocal information to contain the information of the image details. Finally, we defined a coupled term to combine the segmentation and registration. The registration term can provide shape information as a prior to guide the segmentation and the segmentation term can provide the edge information to guide the registration. The segmentation and registration can obtain benefit from each other. Our statistical results on both synthetic and clinical images show that the proposed method can overcome the difficulties caused by noise and bias fields and obtain more accurate results.

Automatic segmentation of the fetal cerebellum on ultrasound volumes, using a 3D statistical shape model

Previous work has shown that the segmentation of anatomical structures on 3D ultrasound data sets provides an important tool for the assessment of the fetal health. In this work, we present an algorithm based on a 3D statistical shape model to segment the fetal cerebellum on 3D ultrasound volumes. This model is adjusted using an ad hoc objective function which is in turn optimized using the Nelder-Mead simplex algorithm. Our algorithm was tested on ultrasound volumes of the fetal brain taken from 20 pregnant women, between 18 and 24 gestational weeks. An intraclass correlation coefficient of 0.8528 and a mean Dice coefficient of 0.8 between cerebellar volumes measured using manual techniques and the volumes calculated using our algorithm were obtained. As far as we know, this is the first effort to automatically segment fetal intracranial structures on 3D ultrasound data.

Automatic segmentation of male pelvic anatomy on computed tomography images: a comparison with multiple observers in the context of a multicentre clinical trial

BackgroundThis study investigates the variation in segmentation of several pelvic anatomical structures on computed tomography (CT) between multiple observers and a commercial automatic segmentation method, in the context of quality assurance and evaluation during a multicentre clinical trial.MethodsCT scans of two prostate cancer patients (‘benchmarking cases’), one high risk (HR) and one intermediate risk (IR), were sent to multiple radiotherapy centres for segmentation of prostate, rectum and bladder structures according to the TROG 03.04 “RADAR” trial protocol definitions. The same structures were automatically segmented using iPlan software for the same two patients, allowing structures defined by automatic segmentation to be quantitatively compared with those defined by multiple observers. A sample of twenty trial patient datasets were also used to automatically generate anatomical structures for quantitative comparison with structures defined by individual observers for the same datasets.ResultsThere was considerable agreement amongst all observers and automatic segmentation of the benchmarking cases for bladder (mean spatial variations < 0.4 cm across the majority of image slices). Although there was some variation in interpretation of the superior-inferior (cranio-caudal) extent of rectum, human-observer contours were typically within a mean 0.6 cm of automatically-defined contours. Prostate structures were more consistent for the HR case than the IR case with all human observers segmenting a prostate with considerably more volume (mean +113.3%) than that automatically segmented. Similar results were seen across the twenty sample datasets, with disagreement between iPlan and observers dominant at the prostatic apex and superior part of the rectum, which is consistent with observations made during quality assurance reviews during the trial.ConclusionsThis study has demonstrated quantitative analysis for comparison of multi-observer segmentation studies. For automatic segmentation algorithms based on image-registration as in iPlan, it is apparent that agreement between observer and automatic segmentation will be a function of patient-specific image characteristics, particularly for anatomy with poor contrast definition. For this reason, it is suggested that automatic registration based on transformation of a single reference dataset adds a significant systematic bias to the resulting volumes and their use in the context of a multicentre trial should be carefully considered.

Regional appearance modeling based on the clustering of intensity profiles

Model-based image segmentation is a popular approach for the segmentation of anatomical structures from medical images because it includes prior knowledge about the shape and appearance of structures of interest. This paper focuses on the formulation of a novel appearance prior that can cope with large variability between subjects, for instance due to the presence of pathologies. Instead of relying on Principal Component Analysis such as in Statistical Appearance Models, our approach relies on a multimodal intensity profile atlas from which a point may be assigned to several profile modes consisting of a mean profile and its covariance matrix. These profile modes are first estimated without any intra-subject registration through a boosted EM classification based on spectral clustering. Then, they are projected on a reference mesh whose role is to store the appearance information in a common geometric representation. We show that this prior leads to better performance than the classical monomodal Principal Component Analysis approach while relying on fewer profile modes.

AN EFFICIENT KOHONEN-FUZZY NEURAL NETWORK BASED ABNORMAL RETINAL IMAGE CLASSIFICATION SYSTEM

Artificial Neural Network (ANN) is the primary automated AI system preferred for medical applications. Even though ANN possesses multiple advan- tages, the convergence of the ANN is not always guaranteed for the practical appli- cations. This often results in the local minima problem and ultimately yields inac- curate results. This convergence problem is common among ANNs and especially in Kohonen neural networks which employ unsupervised training methodology. In this work, an Efficient Kohonen Fuzzy Neural (EKFN) network is proposed to elim- inate the iteration dependent nature of the conventional system. The suitability of this hybrid automated system is illustrated in the context of pathology identi- fication in retinal images. This disease identification system includes anatomical structure segmentation from retinal images followed by image classification. The performance measures used are accuracy, sensitivity, specificity, positive predictive value and positive likelihood ratio. Experimental results show promising possibili- ties for the hybrid systems in terms of performance measures.