Several ophthalmological and systemic diseases are manifested through pathological changes in the properties and the distribution of the retinal blood vessels. The characterization of such alterations requires the segmentation of the vasculature, which is a tedious and time-consuming task that is infeasible to be performed manually. Numerous attempts have been made to propose automated methods for segmenting the retinal vasculature from fundus photographs, although their application in real clinical scenarios is usually limited by their ability to deal with images taken at different resolutions. This is likely due to the large number of parameters that have to be properly calibrated according to each image scale. In this paper we propose to apply a novel strategy for automated feature parameter estimation, combined with a vessel segmentation method based on fully connected conditional random fields. The estimation model is learned by linear regression from structural properties of the images and known optimal configurations, that were previously obtained for low resolution data sets. Our experiments in high resolution images show that this approach is able to estimate appropriate configurations that are suitable for performing the segmentation task without requiring to re-engineer parameters. Furthermore, our combined approach reported state of the art performance on the benchmark data set HRF, as measured in terms of the F1-score and the Matthews correlation coefficient.
KEYWORDS: Image segmentation, 3D modeling, Prostate, Medical imaging, Bladder, 3D image processing, Human-machine interfaces, Magnetic resonance imaging, Chemical elements, Heart
Segmenting 2D and 3D images is a crucial and challenging problem in medical image analysis. Although several image segmentation algorithms have been proposed for different applications, no universal method currently exists. Moreover, their use is usually limited when detection of complex and multiple adjacent objects of interest is needed. In addition, the continually increasing volumes of medical imaging scans require more efficient segmentation software design and highly usable applications. In this context, we present an extension of our previous segmentation framework which allows the combination of existing explicit deformable models in an efficient and transparent way, handling simultaneously different segmentation strategies and interacting with a graphic user interface (GUI). We present the object-oriented design and the general architecture which consist of two layers: the GUI at the top layer, and the processing core filters at the bottom layer. We apply the framework for segmenting different real-case medical image scenarios on public available datasets including bladder and prostate segmentation from 2D MRI, and heart segmentation in 3D CT. Our experiments on these concrete problems show that this framework facilitates complex and multi-object segmentation goals while providing a fast prototyping open-source segmentation tool.
A novel algorithm to evaluate the quality of surface segmentations extracted from 3D images is presented. The procedure calculates the volume enclosed between the segmented object represented by a triangular mesh and a reference one. The indicator is computed by means of a robust and efficient ray-tracing algorithm. This algorithm is fully parallelizable, and it can run even on GPUs architectures. The method is validated against synthetic cases and segmentations of real medical images.
Ultrasound is widely used as an inexpensive, real-time method for imaging vascular tissue. However, sonographs often lack automatic or semi-automatic software for measuring vascular diameter precisely, especially in low- and mid-income countries or institutions. Tools can be developed to perform this task, but they must be validated before being accepted for clinic use. For that purpose, in this work we present low-cost phantoms that resemble vascular tissue when subjected to ultrasound. Several materials are analysed and a step-by-step recipe for building a simple phantom is presented. Qualitatively, models were imaged by an ultrasound expert physician, and several characteristic are assessed. Quantitatively, a comparison between ultrasound and caliper measurements of the phantoms is presented. Finally, a discussion about the results and the recommended materials for low-cost vascular phantoms is carried out.
Background: Intravascular ultrasound (IVUS) provides axial gray-scale images, allowing the assessment of vessel morphology and tissues. Automated segmentation of lumen-intima and media-adventitia interfaces is valuable to identify artery occlusion.
Purpose: Bifurcations, shadows and echogenic plaques usually affect proper segmentation of the vessel wall. Thus, identification of these morphological structures is an advisable step when developing segmentation techniques, which have been dealing with this issue by using different features and methods in the past. The aim of this work is to develop a simultaneous classification method for IVUS image sectors into bifurcations, shadows, echogenic plaques and normal, as an intermediate step for the arterial wall segmentation.
Methods: A 22-dimensional feature vector, mainly composed by current existing methods, is computed for each column in the polar image. To deal with this multiclass classification problem, Random Forest (RF) is used as classifier. Due to the high skewness of the problem, RFs are successively trained by resampling the training data, specifically the majority class.
Results: Fscore reaches 0.62, when the RF is trained with 15% of the normal samples of the training set. Thresholds found in the RF are close to the previously reported values for the features in the literature.
Conclusion: Random Forest demonstrates good performance to classify morphological structures in IVUS. Random undersampling for training was useful to deal with the imbalanced data, and to manage the trade-off between precision and recall of minority classes. However, better features must be developed to improve the classification of the structures, specially in the case of the echogenic plaque.
Most current systems for automated glaucoma detection in fundus images rely on segmentation-based features, which are known to be influenced by the underlying segmentation methods. Convolutional Neural Networks (CNNs) are powerful tools for solving image classification tasks as they are able to learn highly discriminative features from raw pixel intensities. However, their applicability to medical image analysis is limited by the non-availability of large sets of annotated data required for training. In this article we present results of analysis of the viability of using CNNs that are pre-trained from non-medical data for automated glaucoma detection. Two different CNNs, namely OverFeat and VGG-S, were applied to fundus images to generate feature vectors. Preprocessing techniques such as vessel inpainting, contrast-limited adaptive histogram equalization (CLAHE) or cropping around the optic nerve head (ONH) area were explored within this framework to evaluate the improvement in feature discrimination, combined with both ℓ1 and ℓ2 regularized logistic regression models. Results on the Drishti-GS1 dataset, evaluated in terms of area under the average ROC curve, suggests the viability of this approach and offer significant evidence of the importance of well-chosen image pre-processing for transfer learning when the amount of data is not sufficient for fine-tuning the network.
Background and motivation: Real-time ultrasound simulation refers to the process of computationally creating fully synthetic ultrasound images instantly. Due to the high value of specialized low cost training for healthcare professionals, there is a growing interest in the use of this technology and the development of high fidelity systems that simulate the acquisitions of echographic images. The objective is to create an efficient and reproducible simulator that can run either on notebooks or desktops using low cost devices. Materials and methods: We present an interactive ultrasound simulator based on CT data. This simulator is based on ray-casting and provides real-time interaction capabilities. The simulation of scattering that is coherent with the transducer position in real time is also introduced. Such noise is produced using a simplified model of multiplicative noise and convolution with point spread functions (PSF) tailored for this purpose. Results: The computational efficiency of scattering maps generation was revised with an improved performance. This allowed a more efficient simulation of coherent scattering in the synthetic echographic images while providing highly realistic result. We describe some quality and performance metrics to validate these results, where a performance of up to 55fps was achieved. Conclusion: The proposed technique for real-time scattering modeling provides realistic yet computationally efficient scatter distributions. The error between the original image and the simulated scattering image was compared for the proposed method and the state-of-the-art, showing negligible differences in its distribution.
Cardiovascular events are one of the main causes of death in the world. Considering that most of these events are produced by asymptomatic lesions, early and noninvasive analysis and detection of hints for such afflictions are quite desirable. In this work, we study a method known as Lagrangian Speckle Model Estimator (LSME) for assessing the full strain tensor that describes tissue motion between two states during vascular pulse. We then evaluate the results of the LSME method using computer generated images, varying certain mechanical and computational parameters. The strain maps obtained were then assessed, showing percent variation within range of the exact solution. Qualitatively, strain patterns were similar to the analytical solution. Finally, a discussion about further improvements and error causes is carried out.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.