3D deeply supervised network for automated segmentation of volumetric medical images.

Extraction of bone contours from radiographs plays an important role in disease diagnosis, preoperative planning, and treatment analysis. We present a fully automatic method to accurately segment the proximal femur in anteroposterior pelvic radiographs. A number of candidate positions are produced by a global search with a detector. Each is then refined using a statistical shape model together with local detectors for each model point. Both global and local models use Random Forest regression to vote for the optimal positions, leading to robust and accurate results. The performance of the system is evaluated using a set of 839 images of mixed quality. We show that the local search significantly outperforms a range of alternative matching techniques, and that the fully automated system is able to achieve a mean point-to-curve error of less than 0.9 mm for 99% of all 839 images. To the best of our knowledge, this is the most accurate automatic method for segmenting the proximal femur in radiographs yet reported.

Fully Automatic Segmentation of the Proximal Femur Using Random Forest Regression Voting

Modern medical imaging modalities provide large amounts of information in both the spatial and temporal domains and the incorporation of this information in a coherent algorithmic framework is a significant challenge. In this paper, we present a novel and intuitive approach to combine 3-D spatial and temporal (3-D + time) magnetic resonance imaging (MRI) data in an integrated segmentation algorithm to extract the myocardium of the left ventricle. A novel level-set segmentation process is developed that simultaneously delineates and tracks the boundaries of the left ventricle muscle. By encoding prior knowledge about cardiac temporal evolution in a parametric framework, an expectation-maximization algorithm optimally tracks the myocardial deformation over the cardiac cycle. The expectation step deforms the level-set function while the maximization step updates the prior temporal model parameters to perform the segmentation in a nonrigid sense.

/pdf/segmentation-of-the-left-ventricle-of-the-heart-in-3-d-t-mri-t29tcw21bs.pdf

Segmentation of the Left Ventricle of the Heart in 3-D+t MRI Data Using an Optimized Nonrigid Temporal Model

Whole heart segmentation from magnetic resonance imaging or computed tomography is a prerequisite for many clinical applications. Since manual delineation can be tedious and subject to bias, automating such segmentation becomes increasingly popular in the image computing field. However, fully automatic whole heart segmentation is challenging and only limited studies were reported in the literature. This article reviews the existing techniques and analyzes the challenges and methodologies. The techniques are classified in terms of the types of the prior models and the algorithms used to fit the model to unseen images. The prior models include the atlases and the deformable models, and the fitting algorithms are further decomposed into four key techniques including localization of the whole heart, initialization of substructures, refinement of boundary delineation, and regularization of shapes. Finally, the validation issues, challenges, and future directions are discussed.

/pdf/challenges-and-methodologies-of-fully-automatic-whole-heart-7ajzwptl4e.pdf

Challenges and methodologies of fully automatic whole heart segmentation: a review

A comprehensive shape model of the heart

Background: Electrocardiographic methods still provide the bulk of cardiovascular diagnostics Cardiac ischemia is associated with typical alterations in cardiac biosignals that have to be measured, analyzed by mathematical algorithms and allegorized for further clinical diagnostics The fast growing fields of biomedical engineering and applied sciences are intensely focused on generating new approaches to cardiac biosignal analysis for diagnosis and risk stratification in myocardial ischemia

Objectives: To present and review the state of the art in and new approaches to electrocardiologic methods for non-invasive detection and risk stratification in coronary artery disease (CAD) and myocardial ischemia; secondarily, to explore the future perspectives of these methods 

Methods: In follow-up to the Expert Discussion at the 2008 Workshop on "Biosignal Analysis" of the German Society of Biomedical Engineering in Potsdam, Germany, we comprehensively searched the pertinent literature and databases and compiled the results into this review Then, we categorized the state-of-the-art methods and selected new approaches based on their applications in detection and risk stratification of myocardial ischemia Finally, we compared the pros and cons of the methods and explored their future potentials for cardiology

Results: Resting ECG, particularly suited for detecting ST-elevation myocardial infarctions, and exercise ECG, for the diagnosis of stable CAD, are state-of-the-art methods New exercise-free methods for detecting stable CAD include cardiogoniometry (CGM); methods for detecting acute coronary syndrome without ST elevation are Body Surface Potential Mapping, functional imaging and CGM Heart rate variability and blood pressure variability analyses, microvolt T-wave alternans and signal-averaged ECG mainly serve in detecting and stratifying the risk for lethal arrythmias in patients with myocardial ischemia or previous myocardial infarctions Telemedicine and ambient-assisted living support the electrocardiological monitoring of at-risk patients

Conclusions: There are many promising methods for the exercise-free, non-invasive detection of CAD and myocardial ischemia in the stable and acute phases In the coming years, these new methods will help enhance state-of-the-art procedures in routine diagnostics The future can expect that equally novel methods for risk stratification and telemedicine will transition into clinical routine

Electrocardiologic and related methods of non-invasive detection and risk stratification in myocardial ischemia: state of the art and perspectives.

For the analysis of shape variations of the heart and the cardiac motion in a clinical environment it is necessary to segment a large amount of data in order to be able to build statistically significant models. Therefore it has been the aim of this project to find and develop methods that allow the creation of a fully automatic segmentation pipeline for the segmentation of endocardium and myocardium in ECG-triggered MRI images. For this purpose a combination of a number of image processing techniques, from the fields of segmentation, modeling and image registration have been used and extended to create a segmentation pipeline that reduces the need for supplementary manual correction of the segmented labels to a minimum.

Automatic cardiac 4d segmentation using level sets

Cardiogoniometry (CGM), a spatiotemporal electrocardiologic 5-lead method with automated analysis, may be useful in primary healthcare for detecting coronary artery disease (CAD) at rest. Our aim was to systematically develop a stenosis-specific parameter set for global CAD detection. In 793 consecutively admitted patients with presumed non-acute CAD, CGM data were collected prior to elective coronary angiography and analyzed retrospectively. 658 patients fulfilled the inclusion criteria, 405 had CAD verified by coronary angiography; the 253 patients with normal coronary angiograms served as the non-CAD controls. Study patients—matched for age, BMI, and gender—were angiographically assigned to 8 stenosis-specific CAD categories or to the controls. One CGM parameter possessing significance (P < .05) and the best diagnostic accuracy was matched to one CAD category. The area under the ROC curve was .80 (global CAD versus controls). A set containing 8 stenosis-specific CGM parameters described variability of R vectors and R-T angles, spatial position and potential distribution of R/T vectors, and ST/T segment alterations. Our parameter set systematically combines CAD categories into an algorithm that detects CAD globally. Prospective validation in clinical studies is ongoing.

Cardiogoniometric parameters for detection of coronary artery disease at rest as a function of stenosis localization and distribution

Knowledge-based femur detection in conventional radiographs of the pelvis

A solid and accurate proximal femur segmentation technique using the popular active shape model (ASM) is proposed. For generating an optimal shape prior, the minimum description length, based on 200 supervised manual segmented proximal femur shapes, is used. The segmentation is based on a coarse to fine scaling technique including a profile scale space method. The segmentation results are compared using an optimal defined initial pose and a pose based on a registration technique. Using ideal template initialization, 95% of the shapes have been recovered exactly (average point-to-point error {approx}13 pixels, average point-to-boundary error {approx}7 pixels). Using a template-based initialization based on a registration technique, a successful segmentation rate of {approx}89% is achieved, with an average point-to-point error {approx}12 pixels, and an average point-to-boundary error {approx}8 pixels. With an adequate template initialization and an improved ASM, this method seems to provide an accurate tool for segmentation of the proximal femur shapes on conventional hip overview x-ray images.

Roland Pilgram

Papers

Electrocardiologic and related methods of non-invasive detection and risk stratification in myocardial ischemia: state of the art and perspectives.

Automatic cardiac 4d segmentation using level sets

Cardiogoniometric parameters for detection of coronary artery disease at rest as a function of stenosis localization and distribution

Knowledge-based femur detection in conventional radiographs of the pelvis

Proximal femur segmentation in conventional pelvic x ray.