Search

In this study we aimedto democratize access to Convolutional Neural Networks (CNN) for segmenting cartilage volumes, generating state-of-the-art results for specialized, real-world applications in hospitals and research. Segmentation of cross-sectional and/or longitudinalMagnetic Resonance (MR) Images of articular cartilage facilitates both clinical management of joint damage/disease andfundamental research. Manual delineation of such images is a time-consuming task susceptible to high intra- and inter-operator variability and prone to errors. Thus, enabling reliable and efficient analyses of MRIs of cartilage requires automated segmentation of cartilage volumes. Two main limitations arise in the development of hospital- or population-specific Deep Learning (DL) models for image segmentation: specialized knowledge and specialized hardware. We present a relatively easy and accessible implementation of a DL model to automatically segment MR images of human knees with state-of-the-art accuracy. In representative examples we trained CNN models in six to eight hours and obtained results quantitatively comparable to the stateoftheart for every anatomical structure.We established and evaluated our methodsusing two publicly available MRI datasetsoriginating from the Osteoarthritis Initiative, Stryker Imorphics and Zuse Institute Berlin (ZIB), as representative test cases. We use Google Colabfor editing and adapting the Python codesand selecting the runtime environment leveraging high-performance GPUs.We designed our solution for novice users to apply to any dataset with relatively few adaptations requiring only basic programming skills. To facilitate adoption of our methods we provide a complete guideline for using our methods and software, as well as the software tools themselves. This article is protected by copyright. All rights reserved.

Medical imaging is considered one of the most important advances in the history of medicine and has become an essential part of the diagnosis and treatment of patients. Earlier prediction and treatment have been driving the acquisition of higher image resolutions as well as the fusion of different modalities, raising the need for sophisticated hardware and software systems for medical image registration, storage, analysis, and processing. In this scenario and given the new clinical pipelines and the huge clinical burden of hospitals, these systems are often required to provide both highly accurate and real-time processing of large amounts of imaging data. Additionally, lowering the prices of each part of imaging equipment, as well as its development and implementation, and increasing their lifespan is crucial to minimize the cost and lead to more accessible healthcare. This paper focuses on the evolution and the application of different hardware architectures (namely, CPU, GPU, DSP, FPGA, and ASIC) in medical imaging through various specific examples and discussing different options depending on the specific application. The main purpose is to provide a general introduction to hardware acceleration techniques for medical imaging researchers and developers who need to accelerate their implementations.

Background Motion analysis parameters (MAPs) have been extensively validated for assessment of minimally invasive surgical skills. However, there are discrepancies on how specific MAPs, tasks, and skills match with each other, reflecting that motion analysis cannot be generalized independently of the learning outcomes of a task. Additionally, there is a lack of knowledge on the meaning of motion analysis in terms of surgical skills, making difficult the provision of meaningful, didactic feedback. In this study, new higher significance MAPs (HSMAPs) are proposed, validated, and discussed for the assessment of technical skills in box trainers, based on principal component analysis (PCA). Methods Motion analysis data were collected from 25 volunteers performing three box trainer tasks (peg grasping/PG, pattern cutting/PC, knot suturing/KS) using the EVA tracking system. PCA was applied on 10 MAPs for each task and hand. Principal components were trimmed to those accounting for an explained variance >80% to define the HSMAPs. Individual contributions of MAPs to HSMAPs were obtained by loading analysis and varimax rotation. Construct validity of the new HSMAPs was carried out at two levels of experience based on number of surgeries. Results Three new HSMAPs per hand were defined for PG and PC tasks, and two per hand for KS task. PG presented validity for HSMAPs related to insecurity and economy of space. PC showed validity for HSMAPs related to cutting efficacy, peripheral unawareness, and confidence. Finally, KS presented validity for HSMAPs related with economy of space and knotting security. Conclusions PCA-defined HSMAPs can be used for technical skills’ assessment. Construct validation and expert knowledge can be combined to infer how competences are acquired in box trainer tasks. These findings can be exploited to provide residents with meaningful feedback on performance. Future works will compare the new HSMAPs with valid scoring systems such as GOALS.

This research presents a novel low-cost pedagogical environment oriented to ease the use experiential learning methods for training on image-guided robotic surgery technologies. The environment proposes a simplified surgical simulation use case: the movement of insertion and extraction of a needle, similar to an image-guided biopsy situation. The training environment is composed of a 3D-printed phantom, a CT scan of the phantom, a virtual reality environment that includes haptic information, an ad-hoc 1 degree of freedom (DOF) robotic system for insertion/extraction and a Novint Falcon, a low cost 3 DOF manipulator that allows achieving haptic feedback. A first pilot experience has been carried out in the ?Surgical Simulation and Planning? course of the Bachelor of Biomedical Engineering at Universidad Politécnica de Madrid (UPM). Results of the surveys carried out (by teachers and learners) show that the new project-based methodology improves in all cases the values of student?s satisfaction obtained using the classical methodology based on master classes and practices.

We propose a fully automated methodology for hexahedral meshing of patient-specific structures of the human knee obtained from magnetic resonance images, i.e. femoral/tibial cartilages and menisci. We select eight patients from the Osteoarthritis Initiative and validate our methodology using MATLAB on a laptop computer. We obtain the patient-specific meshes in an average of three minutes, while faithfully representing the geometries with well-shaped elements. We hope to provide a fundamentally different means to test hypotheses on the mechanisms of disease progression by integrating our patient-specific FE meshes with data from individual patients. Download both our meshes and software at http://im.engr.uconn.edu/downloads.php.

Accurate estimation of peak wall stress (PWS) is the crux of biomechanically motivated rupture risk assessment for abdominal aortic aneurysms aimed to improve clinical outcomes. Such assessments often use the finite element (FE) method to obtain PWS, albeit at a high computational cost, motivating simplifications in material or element formulations. These simplifications, while useful, come at a cost of reliability and accuracy. We achieve research-standard accuracy and maintain clinically applicable speeds by using novel computational technologies. We present a solution using our custom finite element code based on graphics processing unit (GPU) technology that is able to account for added complexities involved with more physiologically relevant solutions, e.g. strong anisotropy and heterogeneity. We present solutions up to 17× faster relative to an established finite element code using state-of-the-art nonlinear, anisotropic and nearly-incompressible material descriptions. We show a realistic assessment of the explicit GPU FE approach by using complex problem geometry, biofidelic material law, double-precision floating point computation and full element integration. Due to the increased solution speed without loss of accuracy, shown on five clinical cases of abdominal aortic aneurysms, the method shows promise for clinical use in determining rupture risk of abdominal aortic aneurysms. publisher: Elsevier articletitle: Rupture risk in abdominal aortic aneurysms: A realistic assessment of the explicit GPU approach journaltitle: Journal of Biomechanics articlelink: http://dx.doi.org/10.1016/j.jbiomech.2017.02.019 content_type: article copyright: © 2017 Elsevier Ltd. All rights reserved. ispartof: Journal of Biomechanics vol:56 pages:1-9 ispartof: location:United States status: published

Purpose: Accurate delineation of the rectum is of high importance in off-line adaptive radiation therapy since it is a major dose-limiting organ in prostate cancer radiotherapy. The intensity-based deformable image registration (DIR) methods cannot create a correct spatial transformation if there is no correspondence between the template and the target images. The variation of rectal filling, gas, or feces, creates a noncorrespondence in image intensities that becomes a great obstacle for intensity-based DIR. Methods: In this study the authors have designed and implemented a semiautomatic method to create a rectum mask in pelvic computed tomography (CT) images. The method, that includes a DIR based on the demons algorithm, has been tested in 13 prostate cancer cases, each comprising of two CT scans, for a total of 26 CT scans. Results: The use of the manual segmentation in the planning image and the proposed rectum mask method (RMM) method in the daily image leads to an improvement in the DIR performance in pelvic CT images, obtaining a mean value of overlap volume index = 0.89, close to the values obtained using the manual segmentations in both images. Conclusions: The application of the RMM method in the daily image and the manual segmentations in the planning image during prostate cancer treatments increases the performance of the registration in presence of rectal fillings, obtaining very good agreement with a physicianˈs manual contours.

Purpose: Current radiotherapy is progressing to the concept of adaptive radiotherapy, which implies the adaptation of planning along the treatment course. Nonrigid registration is an essential image processing tool for adaptive radiotherapy and image guided radiotherapy, and the three-dimensional (3D) nature of the current radiotherapy techniques requires a 3D quantification of the registration error that existing evaluation methods do not cover appropriately. The authors present a method for 3D evaluation of nonrigid registration algorithms’ performance, based on organ delineations, capable of working with near-spherical volumes even in the presence of concavities. Methods: The evaluation method is composed by a volume shape description stage, developed using a newad hoc volume reconstruction algorithm proposed by the authors, and an error quantification stage. The evaluation method is applied to the organ delineations of prostate and seminal vesicles, obtained by an automatic segmentation method over images of prostate cancer patients treated with intensity modulated radiation therapy. Results: The volume reconstruction algorithm proposed has been shown to accurately model complex 3D surfaces by the definition of clusters of control points. The quantification method, inspired by the Haussdorf–Chebysev distance, provides a measure of the largest registration error per control direction, defining a valid metric for concave-convex volumes. Summarizing, the proposed evaluation methodology presents accurate results with a high spatial resolution in a negligible computation time in comparison with the nonrigid registration time. Conclusions: Experimental results show that the metric selected for quantifying the registration error is of utmost importance in a quantitative evaluation based on measuring distances between volumes. The accuracy of the volume reconstruction algorithm is not so relevant as long as the reconstruction is tight enough on the actual volume of the organ. The new evaluation method provides a smooth and accurate volume reconstruction for both the reference and the registered organ, and a complete 3D description of nonrigid registration algorithms’ performance, resulting in a useful tool for study and comparison of registration algorithms for adaptive radiotherapy.

Due to limitations in conventional medical imaging and the restrictions imposed by both the anatomy and the surgical approach in pancreatic cancer, there is a need for methods to support intraoperative imaging in order to improve their accurate anatomical localization and the characterization of their nature. Laparoscopic ultrasounds (LUS) images and endoscopic videos can be used to extract useful information during the surgical procedures. A fast approach for acquiring an estimation of the tumor positioning and size through laparoscopic ultrasounds images has been developed. Based on the surgical video, endoscope 3D tracking is achieved by means of a Shape-from-Motion technique. Intraoperative imaging algorithms’ validation has been carried out in an ex vivo porcine model and results have shown the viability of exploiting them for structures characterization and their 3D reconstruction.

The existence of residual stresses in human arteries has long been shown experimentally. Researchers have also demonstrated that residual stresses have a significant effect on the distribution of physiological stresses within arterial tissues, and hence on their development, e.g., stress-modulated remodeling. Through progress in medical imaging, image analysis and finite element (FE) meshing tools it is now possible to construct in vivo patient-specific geometries and thus to study specific, clinically relevant problems in arterial mechanics via FE simulations. Classical continuum mechanics and FE methods assume that constitutive models and the corresponding simulations start from unloaded, stress-free reference configurations while the boundary-value problem of interest represents a loaded geometry and includes residual stresses. We present a pragmatic methodology to simultaneously account for both (i) the three-dimensional (3-D) residual stress distributions in the arterial tissue layers, and (ii) the equilibrium of the in vivo patient-specific geometry with the known boundary conditions. We base our methodology on analytically determined residual stress distributions (Holzapfel and Ogden, 2010, J. R. Soc. Interface 7, 787-799) and calibrate it using data on residual deformations (Holzapfel et al., 2007, Ann. Biomed. Eng. 35, 530-545). We demonstrate our methodology on three patient-specific FE simulations calibrated using experimental data. All data employed here are generated from human tissues - both the aorta and thrombus, and their respective layers - including the geometries determined from magnetic resonance images, and material properties and 3-D residual stretches determined from mechanical experiments. We study the effect of 3-D residual stresses on the distribution of physiological stresses in the aortic layers (intima, media, adventitia) and the layers of the intraluminal thrombus (luminal, medial, abluminal) by comparing three types of FE simulations: (i) conventional calculations; (ii) calculations accounting only for prestresses; (iii) calculations including both 3-D residual stresses and prestresses. Our results show that including residual stresses in patient-specific simulations of arterial tissues significantly impacts both the global (organ-level) deformations and the stress distributions within the arterial tissue (and its layers). Our method produces circumferential Cauchy stress distributions that are more uniform through the tissue thickness (i.e., smaller stress gradients in the local radial directions) compared to both the conventional and prestressing calculations. Such methods, combined with appropriate experimental data, aim at increasing the accuracy of classical FE analyses for patient-specific studies in computational biomechanics and may lead to increased clinical application of simulation tools. publisher: Elsevier articletitle: A method for incorporating three-dimensional residual stretches/stresses into patient-specific finite element simulations of arteries journaltitle: Journal of the Mechanical Behavior of Biomedical Materials articlelink: http://dx.doi.org/10.1016/j.jmbbm.2015.03.024 content_type: article copyright: Copyright © 2015 Elsevier Ltd. All rights reserved. ispartof: Journal of the Mechanical Behavior of Biomedical Materials vol:47 pages:147-164 ispartof: location:Netherlands status: published

A correct patient-specific identification of the abdominal aortic aneurysm is useful for both diagnosis and treatment stages, as it locates the disease and represents its geometry. The actual thickness and shape of the arterial wall and the intraluminal thrombus are of great importance when predicting the rupture of the abdominal aortic aneurysms. The authors describe a novel method for delineating both the internal and external contours of the aortic wall, which allows distinguishing between vessel wall and intraluminal thrombus. The method is based on active shape model and texture statistical information. The method was validated with eight MR patient studies. There was high correspondence between automatic and manual measurements for the vessel wall area. Resulting segmented images presented a mean Dice coefficient with respect to manual segmentations of 0.88 and a mean modified Hausdorff distance of 1.14 mm for the internal face and 0.86 and 1.33 mm for the external face of the arterial wall. Preliminary results of the segmentation show high correspondence between automatic and manual measurements for the vessel wall and thrombus areas. However, since the dataset is small the conclusions cannot be generalized. ispartof: Computational and Mathematical Methods in Medicine vol:2015 ispartof: location:United States status: published

A novel method for generating patient-specific high quality conforming hexahedral meshes is presented. The meshes are directly obtained from the segmentation of patient magnetic resonance (MR) images of abdominal aortic aneu-rysms (AAA). The MRI permits distinguishing between struc-tures of interest in soft tissue. Being so, the contours of the lumen, the aortic wall and the intraluminal thrombus (ILT) are available and thus the meshes represent the actual anato-my of the patient?s aneurysm, including the layered morpholo-gies of these structures. Most AAAs are located in the lower part of the aorta and the upper section of the iliac arteries, where the inherent tortuosity of the anatomy and the presence of the ILT makes the generation of high-quality elements at the bifurcation is a challenging task. In this work we propose a novel approach for building quadrilateral meshes for each surface of the sectioned geometry, and generating conforming hexahedral meshes by combining the quadrilateral meshes. Conforming hexahedral meshes are created for the wall and the ILT. The resulting elements are evaluated on four patients? datasets using the Scaled Jacobian metric. Hexahedral meshes of 25,000 elements with 94.8% of elements well-suited for FE analysis are generated.

This work presents a real time non-rigid registration method for a parametric model of the aorta based on computer animation techniques. The endovascular system is modeled as a polygonal mesh and a skeletal structure, both calculated automatically from segmented pre-operative images. The skeleton, registered using linear regression methods, supports an adaptive triangular mesh, both adapting in real time to the longitudinal deformations and changes affecting the aorta during the procedure. These changes are offered to the surgeon as part of a virtual-reality catheterization guidance system. The initial evaluation shows that the algorithm and implementation are able to handle up to 5 situation updates per second while maintaining a mean registration error of 1,012 mm and a maximum of 4.346 mm, thus effectively supporting real-time navigation.

Through progress in medical imaging, image analysis and finite element (FE) meshing tools it is now possible to extract patient-specific geometries from medical images of, e.g., abdominal aortic aneurysms (AAAs), and thus to study clinically relevant problems via FE simulations. Medical imaging is most often performed in vivo, and hence the reconstructed model geometry in the problem of interest will represent the in vivo state, e.g., the AAA at physiological blood pressure. However, classical continuum mechanics and FE methods assume that constitutive models and the corresponding simulations start from an unloaded, stress-free reference condition.

In order to perform finite element (FE) analyses of patient-specific abdominal aortic aneurysms, geometries derived from medical images must be meshed with suitable elements. We propose a semi-automatic method for generating conforming hexahedral meshes directly from contours segmented from medical images. Magnetic resonance images are generated using a protocol developed to give the abdominal aorta high contrast against the surrounding soft tissue. These data allow us to distinguish between the different structures of interest. We build novel quadrilateral meshes for each surface of the sectioned geometry and generate conforming hexahedral meshes by combining the quadrilateral meshes. The three-layered morphology of both the arterial wall and thrombus is incorporated using parameters determined from experiments. We demonstrate the quality of our patient-specific meshes using the element Scaled Jacobian. The method efficiently generates high-quality elements suitable for FE analysis, even in the bifurcation region of the aorta into the iliac arteries. For example, hexahedral meshes of up to 125,000 elements are generated in less than 130 s, with 94.8 % of elements well suited for FE analysis. We provide novel input for simulations by independently meshing both the arterial wall and intraluminal thrombus of the aneurysm, and their respective layered morphologies. QC 20140306

A new method for three-dimensional quantitative evaluation of target volume delineation is presented. It is composed by a new 3D reconstruction method called Origami, based on the combination of two bidimensional analysis of the volume of each organ instead of one 3D analysis, and a 3D description of the error distribution in the space. The Origami method avoid external errors introduced by a 3D rendering and has shown to work correctly in both convex and convex-concave volumes, accurately fitting the contours of the planning study. Its performance has been evaluated comparing it with the Convex Hull algorithm using Delaunay triangulation, resulting in a much more adjusted evaluation. Moreover, Origami computes thousands of control points in less than one second in a common PC.

Two deformable registration methods, the Demons and the Morphon algorithms, have been used for registration of CT datasets to evaluate their usability in radiotherapy planning for prostate cancer. These methods were chosen because they can perform deformable registration in a fully automated way. The experiments show that for intrapatient registration both of the methods give useful results, although some differences exist in the way they deform the template. The Morphon method has, however, some advantageous compared to the Demons method. It is invariant to the image intensity and it does not distort the deformed data. The conclusion is therefore to recommend the Morphon method as a registration tool for this application. A more flexible regularization model is needed, though, in order to be able to catch the full range of deformations required to match the datasets.