RESUMEN
Purpose To develop a deep learning-based method for fully automated quantification of left ventricular (LV) function from short-axis cine MR images and to evaluate its performance in a multivendor and multicenter setting. Materials and Methods This retrospective study included cine MRI data sets obtained from three major MRI vendors in four medical centers from 2008 to 2016. Three convolutional neural networks (CNNs) with the U-NET architecture were trained on data sets of increasing variability: (a) a single-vendor, single-center, homogeneous cohort of 100 patients (CNN1); (b) a single-vendor, multicenter, heterogeneous cohort of 200 patients (CNN2); and (c) a multivendor, multicenter, heterogeneous cohort of 400 patients (CNN3). All CNNs were tested on an independent multivendor, multicenter data set of 196 patients. CNN performance was evaluated with respect to the manual annotations from three experienced observers in terms of (a) LV detection accuracy, (b) LV segmentation accuracy, and (c) LV functional parameter accuracy. Automatic and manual results were compared with the paired Wilcoxon test, Pearson correlation, and Bland-Altman analysis. Results CNN3 achieved the highest performance on the independent testing data set. The average perpendicular distance compared with manual analysis was 1.1 mm ± 0.3 for CNN3, compared with 1.5 mm ± 1.0 for CNN1 (P < .05) and 1.3 mm ± 0.6 for CNN2 (P < .05). The LV function parameters derived from CNN3 showed a high correlation (r2 ≥ 0.98) and agreement with those obtained by experts for data sets from different vendors and centers. Conclusion A deep learning-based method trained on a data set with high variability can achieve fully automated and accurate cine MRI analysis on multivendor, multicenter cine MRI data. © RSNA, 2018 See also the editorial by Colletti in this issue.
Asunto(s)
Aprendizaje Profundo , Ventrículos Cardíacos/diagnóstico por imagen , Interpretación de Imagen Asistida por Computador/métodos , Imagen por Resonancia Cinemagnética/métodos , Función Ventricular/fisiología , Humanos , Estudios RetrospectivosRESUMEN
OBJECTIVES: Tenosynovitis (inflammation of the synovial lining of the sheath surrounding tendons) is frequently observed on MRI of early arthritis patients. Since visual assessment of tenosynovitis is a laborious task, we investigated the feasibility of automatic quantification of tenosynovitis on MRI of the wrist in a large cohort of early arthritis patients. METHODS: For 563 consecutive early arthritis patients (clinically confirmed arthritis ≥ 1 joint, symptoms < 2 years), MR scans of the wrist were processed in three automatic stages. First, super-resolution reconstruction was applied to fuse coronal and axial scans into a single high-resolution three-dimensional image. Next, 10 extensor/flexor tendon regions were segmented using atlas-based segmentation and marker-based watershed. A measurement region of interest (ROI) was defined around the tendons. Finally, tenosynovitis was quantified by identifying image intensity values associated with tenosynovial inflammation using fuzzy clustering and measuring the fraction of voxels with these characteristic intensities within the measurement ROI. A subset of 60 patients was used for training and the remaining 503 patients for validation. Correlation between quantitative measurements and visual scores was assessed through Pearson correlation coefficient. RESULTS: Pearson correlation between quantitative measurements and visual scores across 503 patients was r = 0.90, p < 0.001. False detections due to blood vessels and synovitis present within the measurement ROI contributed to a median offset from zero equivalent to 13.8% of the largest measurement value. CONCLUSION: Quantitative measurement of tenosynovitis on MRI of the wrist is feasible and largely consistent with visual scores. Further improvements in segmentation and exclusion of false detections are warranted. KEY POINTS: ⢠Automatic measurement of tenosynovitis on MRI of the wrist is feasible and largely consistent with visual scores. ⢠Blood vessels and synovitis in the vicinity of evaluated tendons can contribute to false detections in automatic measurements. ⢠Further improvements in segmentation and exclusion of false detections are important directions of future work on the path to a robust quantification framework.
Asunto(s)
Artritis Reumatoide/complicaciones , Imagen por Resonancia Magnética/métodos , Membrana Sinovial/patología , Tenosinovitis/diagnóstico , Articulación de la Muñeca/patología , Artritis Reumatoide/diagnóstico , Estudios de Factibilidad , Femenino , Humanos , Imagenología Tridimensional , Masculino , Persona de Mediana Edad , Reproducibilidad de los Resultados , Tenosinovitis/etiologíaRESUMEN
OBJECTIVE: To establish a good method to determine the retinal shape from MRI using three-dimensional (3D) ellipsoids as well as evaluate its reproducibility. METHODS AND ANALYSIS: The left eyes of 31 volunteers were imaged using high-resolution ocular MRI. The 3D MR-images were segmented and ellipsoids were fitted to the resulting contours. The dependency of the resulting ellipsoid parameters on the evaluated fraction of the retinal contour was assessed by fitting ellipsoids to 41 different fractions. Furthermore, the reproducibility of the complete procedure was evaluated in four subjects. Finally, a comparison with conventional two-dimensional (2D) methods was made. RESULTS: The mean distance between the fitted ellipsoids and the segmented retinal contour was 0.03±0.01 mm (mean±SD) for the central retina and 0.13±0.03 mm for the peripheral retina. For the central retina, the resulting ellipsoid radii were 12.9±0.9, 13.7±1.5 and 12.2±1.2 mm along the horizontal, vertical and central axes. For the peripheral retina, these radii decreased to 11.9±0.6, 11.6±0.4 and 10.4±0.7 mm, which was accompanied by a mean 1.8 mm posterior shift of the ellipsoid centre. The reproducibility of the ellipsoid fitting was 0.3±1.2 mm for the central retina and 0.0±0.1 mm for the peripheral retina. When 2D methods were used to fit the peripheral retina, the fitted radii differed a mean 0.1±0.1 mm from the 3D method. CONCLUSION: An accurate and reproducible determination of the 3D retinal shape based on MRI is provided together with 2D alternatives, enabling wider use of this method in the field of ophthalmology.
RESUMEN
PURPOSE: The quantification of vessel wall morphology and plaque burden requires vessel segmentation, which is generally performed by manual delineations. The purpose of our work is to develop and evaluate a new 3D model-based approach for carotid artery wall segmentation from dual-sequence MRI. METHODS: The proposed method segments the lumen and outer wall surfaces including the bifurcation region by fitting a subdivision surface constructed hierarchical-tree model to the image data. In particular, a hybrid segmentation which combines deformable model fitting with boundary classification was applied to extract the lumen surface. The 3D model ensures the correct shape and topology of the carotid artery, while the boundary classification uses combined image information of 3D TOF-MRA and 3D BB-MRI to promote accurate delineation of the lumen boundaries. The proposed algorithm was validated on 25 subjects (48 arteries) including both healthy volunteers and atherosclerotic patients with 30% to 70% carotid stenosis. RESULTS: For both lumen and outer wall border detection, our result shows good agreement between manually and automatically determined contours, with contour-to-contour distance less than 1 pixel as well as Dice overlap greater than 0.87 at all different carotid artery sections. CONCLUSIONS: The presented 3D segmentation technique has demonstrated the capability of providing vessel wall delineation for 3D carotid MRI data with high accuracy and limited user interaction. This brings benefits to large-scale patient studies for assessing the effect of pharmacological treatment of atherosclerosis by reducing image analysis time and bias between human observers.
Asunto(s)
Arterias Carótidas/diagnóstico por imagen , Procesamiento de Imagen Asistido por Computador/métodos , Aprendizaje Automático , Imagen por Resonancia Magnética , Adulto , Anciano , Automatización , Humanos , Masculino , Reconocimiento de Normas Patrones AutomatizadasRESUMEN
PURPOSE: Recent studies on ocular shape have raised increased interest in the peripheral characteristics of the eye, as it potentially triggers changes in the central vision. Current techniques are, however, not capable of accurately measuring the three-dimensional shape of the retina. We describe a new magnetic resonance imaging (MRI)-based method to obtain the retinal shape with high precision and use it to assess if differences in retinal shape could explain previously described trends in peripheral refraction. METHODS: Twenty-one healthy subjects were examined using high-field ocular MRI. The resulting data were automatically segmented and processed to calculate the retinal topographic map. We validated the method against partial coherence interferometry and assessed the reproducibility for four subjects. RESULTS: The retinal topographic maps describe the retinal shape with subpixel reproducibility (SD between sessions = 0.11 mm). Comparison with partial coherence interferometry showed a mean difference of 0.08 mm, 95% confidence interval -0.39 to 0.55 mm, with a standard deviation of 0.23 mm. The data give a possible geometric explanation for the previously described trend in myopic eyes toward relatively hyperopic refraction in the periphery, with full three-dimensional information. The retinal maps furthermore show small, submillimeter, irregularities that could have an important influence on the subjects' peripheral vision. CONCLUSIONS: The possibility to quantitatively characterize the full three-dimensional retinal shape by MRI offers new ophthalmologic possibilities, such as quantitative geometric description of staphyloma. It could in addition be used as a validation technique, independent of standard optical methods, to measure the peripheral retinal shape.
Asunto(s)
Catarata/diagnóstico , Procesamiento de Imagen Asistido por Computador , Imagen por Resonancia Magnética/métodos , Retina/patología , Adulto , Femenino , Humanos , Masculino , Persona de Mediana Edad , Valores de Referencia , Reproducibilidad de los Resultados , Adulto JovenRESUMEN
The intensity or gray-level derivatives have been widely used in image segmentation and enhancement. Conventional derivative filters often suffer from an undesired merging of adjacent objects because of their intrinsic usage of an inappropriately broad Gaussian kernel; as a result, neighboring structures cannot be properly resolved. To avoid this problem, we propose to replace the low-level Gaussian kernel with a bi-Gaussian function, which allows independent selection of scales in the foreground and background. By selecting a narrow neighborhood for the background with regard to the foreground, the proposed method will reduce interference from adjacent objects simultaneously preserving the ability of intraregion smoothing. Our idea is inspired by a comparative analysis of existing line filters, in which several traditional methods, including the vesselness, gradient flux, and medialness models, are integrated into a uniform framework. The comparison subsequently aids in understanding the principles of different filtering kernels, which is also a contribution of this paper. Based on some axiomatic scale-space assumptions, the full representation of our bi-Gaussian kernel is deduced. The popular γ-normalization scheme for multiscale integration is extended to the bi-Gaussian operators. Finally, combined with a parameter-free shape estimation scheme, a derivative filter is developed for the typical applications of curvilinear structure detection and vasculature image enhancement. It is verified in experiments using synthetic and real data that the proposed method outperforms several conventional filters in separating closely located objects and being robust to noise.
Asunto(s)
Algoritmos , Vasos Sanguíneos/anatomía & histología , Procesamiento de Imagen Asistido por Computador/métodos , Humanos , Modelos Biológicos , Modelos Teóricos , Curva ROC , Reproducibilidad de los ResultadosRESUMEN
Nonrigid image registration is an important, but time-consuming task in medical image analysis. In typical neuroimaging studies, multiple image registrations are performed, i.e., for atlas-based segmentation or template construction. Faster image registration routines would therefore be beneficial. In this paper we explore acceleration of the image registration package elastix by a combination of several techniques: (i) parallelization on the CPU, to speed up the cost function derivative calculation; (ii) parallelization on the GPU building on and extending the OpenCL framework from ITKv4, to speed up the Gaussian pyramid computation and the image resampling step; (iii) exploitation of certain properties of the B-spline transformation model; (iv) further software optimizations. The accelerated registration tool is employed in a study on diagnostic classification of Alzheimer's disease and cognitively normal controls based on T1-weighted MRI. We selected 299 participants from the publicly available Alzheimer's Disease Neuroimaging Initiative database. Classification is performed with a support vector machine based on gray matter volumes as a marker for atrophy. We evaluated two types of strategies (voxel-wise and region-wise) that heavily rely on nonrigid image registration. Parallelization and optimization resulted in an acceleration factor of 4-5x on an 8-core machine. Using OpenCL a speedup factor of 2 was realized for computation of the Gaussian pyramids, and 15-60 for the resampling step, for larger images. The voxel-wise and the region-wise classification methods had an area under the receiver operator characteristic curve of 88 and 90%, respectively, both for standard and accelerated registration. We conclude that the image registration package elastix was substantially accelerated, with nearly identical results to the non-optimized version. The new functionality will become available in the next release of elastix as open source under the BSD license.