Gamma mixture classifier for plaque detection in intravascular ultrasonic images

Outdoor THz fading modeling by means of gaussian and gamma mixture distributions

Terahertz (THz) band offers a vast amount of bandwidth and is envisioned to become a key enabler for a number of next generation wireless applications. In this direction, appropriate channel models, encapsulating the large and small-scale fading phenomena, need to be developed for both indoor and outdoor communications environments. The THz large-scale fading characteristics have been extensively investigated for both indoor and outdoor scenarios. The study of indoor THz small-scale fading has recently gained the momentum, while the small-scale fading of outdoor THz wireless channels has not yet been investigated. Motivated by this, this contribution introduces Gaussian mixture (GM) distribution as a suitable small-scale fading model for outdoor THz wireless links. In more detail, multiple outdoor THz wireless measurements recorded at different transceiver separation distance are fed to an expectation-maximization fitting algorithm, which returns the parameters of the GM probability density function. The fitting accuracy of the analytical GMs is evaluated in terms of the Kolmogorov-Smirnov, Kullback-Leibler (KL) and root-mean-square-error (RMSE) tests. The results reveal that as the number of mixtures increases the resulting analytical GMs perform a better fit to the empirical distributions. In addition, the KL and RMSE metrics indicate that the increase of mixtures beyond a particular number result to no significant improvement of the fitting accuracy. Finally, following the same approach as in the case of GM, we examine the suitability of mixture Gamma to capture the small-scale fading characteristics of the outdoor THz channels.

Performance Analysis of Machine Learning and Deep Learning Architectures on Early Stroke Detection Using Carotid Artery Ultrasound Images

Atherosclerotic plaque deposit in the carotid artery is used as an early estimate to identify the presence of cardiovascular diseases. Ultrasound images of the carotid artery are used to provide the extent of stenosis by examining the intima-media thickness and plaque diameter. A total of 361 images were classified using machine learning and deep learning approaches to recognize whether the person is symptomatic or asymptomatic. CART decision tree, random forest, and logistic regression machine learning algorithms, convolutional neural network (CNN), Mobilenet, and Capsulenet deep learning algorithms were applied in 202 normal images and 159 images with carotid plaque. Random forest provided a competitive accuracy of 91.41% and Capsulenet transfer learning approach gave 96.7% accuracy in classifying the carotid artery ultrasound image database.

CurveletTransform-Based Texture Analysis of Carotid B-mode Ultrasound Images in Asymptomatic Men With Moderate and Severe Stenoses: A Preliminary Clinical Study

The curvelet transform, which represents images in terms of their geometric and textural characteristics, was investigated toward revealing differences between moderate (50%-69%, n = 11) and severe (70%-100%, n = 14) stenosis asymptomatic plaque from B-mode ultrasound. Texture features were estimated in original and curvelet transformed images of atheromatous plaque (PL), the adjacent arterial wall (intima-media [IM]) and the plaque shoulder (SH) (i.e., the boundary between plaque and wall), separately at end systole and end diastole. Seventeen features derived from the original images were significantly different between the two groups (4 for IM, 3 for PL and 10 for SH; 9 for end diastole and 8 for end systole); 19 of 234 features (2 for IM and 17 for SH; 8 for end systole and 11 for end diastole) derived from curvelet transformed images were significantly higher in the patients with severe stenosis, indicating higher magnitude, variation and randomness of image gray levels. In these patients, lower body height and higher serum creatinine concentration were observed. Our findings suggest that (a) moderate and severe plaque have similar curvelet-based texture properties, and (b) IM and SH provide useful information about arterial wall pathophysiology, complementary to PL itself. The curvelet transform is promising for identifying novel indices of cardiovascular risk and warrants further investigation in larger cohorts.

Automatic estimation of aortic and mitral valve displacements in dynamic CTA with 4D graph-cuts

The location of the mitral and aortic valves in dynamic cardiac imaging is useful for extracting functional derived parameters such as ejection fraction, valve excursions, and global longitudinal strain, and when performing anatomical structures tracking using slice following or valve intervention's planning. Completely automatic segmentation methods are still challenging tasks because of their fast movements and the different positions that prevent good visibility of the leaflets along the full cardiac cycle. In this article, we propose a processing pipeline to track the displacement of the aortic and mitral valve annuli from high-resolution cardiac four-dimensional computed tomographic angiography (4D-CTA). The proposed method is based on the dynamic separation of left ventricle, left atrium and aorta using statistical shape modeling and an energy minimization algorithm based on graph-cuts and has been evaluated on a set of 15 electrocardiography-gated 4D-CTAs. We report a mean agreement distance between manual annotations and our proposed method of 2.52±1.06 mm for the mitral annulus and 2.00±0.69 mm for the aortic valve annulus based on valve locations detected from manual anatomical landmarks. In addition, we show the effect of detecting the valvular planes on derived functional parameters (ejection fraction, global longitudinal strain, and excursions of the mitral and aortic valves).

Left ventricle segmentation using a Bayesian approach with distance dependent shape priors

We propose a method for segmentation of the left ventricle in magnetic resonance cardiac images. The framework consists of an initial Bayesian segmentation of the central slice of the volume. This segmentation is used to locate a shape prior for the LV myocardial tissue. This shape prior is determined using the fact that the myocardium is approximately annular as seen in the short-axis. Then a second Bayesian segmentation is performed to obtain the final result. This procedure is repeated for the rest of the slices. An extrapolation of the area of the LV is used to determine a stopping criterion. The method was evaluated on the databases of the Cardiac Atlas project. Our results demonstrate a suitable accuracy for myocardial segmentation (≈0.8 Dice's coefficient). For the endocardium and the epicardium the Dice's coefficients are 0.94 and 0.9 respectively. The accuracy was also evaluated in terms of the Hausdorff distance and the average distance. For the myocardium we obtain 8 mm and 2 mm respectively. Our results demonstrate the capability and merits of the proposed method to estimate the structure of the LV. The method requires minimal user input and generates results with quality comparable to more complex approaches. This paper suggests a new efficient approach for automatic LV quantification based on a Bayesian technique with shape priors with errors comparable to state-of-the-art techniques.

Carotid artery ultrasound image analysis: A review of the literature

Stroke is one of the prominent causes of death in the recent days. The existence of susceptible plaque in the carotid artery can be used in ascertaining the possibilities of cardiovascular diseases and long-term disabilities. The imaging modality used for early screening of the disease is B-mode ultrasound image of the person in the artery area. The objective of this article is to give a widespread review of the imaging modes and methods used for studying the carotid artery for identifying stroke, atherosclerosis and related cardiovascular diseases. We encompass the review in methods used for artery wall tracking, intima-media, and lumen segmentation which will help in finding the extent of the disease. Due to the characteristics of the imaging modality used, the images have speckle noise which worsens the image quality. Adaptive homomorphic filtering with wavelet and contourlet transforms, Levy Shrink, gamma distribution were used for image denoising. Learning-based neural network approaches for denoising give better edge preservation. Domain knowledge-based segmentation approaches have proved to provide more accurate intima-media thickness measurements. There is a requirement of useful fully automatic segmentation approaches, 3D, 4D systems, and plaque motion analysis. Taking into consideration the image priors like geometry, imaging physics, intensity and temporal data, image analysis has to be performed. Encouragingly more research has focused on content-specific segmentation and classification techniques. With the evaluation of machine learning algorithms, classifying the image as with or without a fat deposit has gained better accuracy and sensitivity. Machine learning-based approaches like self-organizing map, k-nearest neighborhood and support vector machine achieve promising accuracy and sensitivity in classification. The literature reveals that there is more scope in identifying a patient-specific model in a fully automatic manner.

Delineation of Human Carotid Plaque Features In Vivo by Exploiting Displacement Variance

While in vivo acoustic radiation force impulse (ARFI)-induced peak displacement (PD) has been demonstrated to have high sensitivity and specificity for differentiating soft from stiff plaque components in patients with carotid plaque, the parameter exhibits poorer performance for distinguishing between plaque features with similar stiffness. To improve discrimination of carotid plaque features relative to PD, we hypothesize that signal correlation and signal-to-noise ratio (SNR) can be combined, outright or via displacement variance. Plaque feature detection by displacement variance, evaluated as the decadic logarithm of the variance of acceleration and termed "log(VoA)," was compared to that achieved by exploiting SNR, cross correlation coefficient, and ARFI-induced PD outcome metrics. Parametric images were rendered for 25 patients undergoing carotid endarterectomy, with spatially matched histology confirming plaque composition and structure. On average, across all plaques, log(VoA) was the only outcome metric with values that statistically differed between regions of lipid-rich necrotic core (LRNC), intraplaque hemorrhage (IPH), collagen (COL), and calcium (CAL). Further, log(VoA) achieved the highest contrast-to-noise ratio (CNR) for discriminating between LRNC and IPH, COL and CAL, and grouped soft (LRNC and IPH) and stiff (COL and CAL) plaque components. More specifically, relative to the previously demonstrated ARFI PD parameter, log(VoA) achieved 73% higher CNR between LRNC and IPH and 59% higher CNR between COL and CAL. These results suggest that log(VoA) enhances the differentiation of LRNC, IPH, COL, and CAL in human carotid plaques, in vivo, which is clinically relevant to improving stroke risk prediction and medical management.

Performance of acoustic radiation force impulse ultrasound imaging for carotid plaque characterization with histologic validation

OBJECTIVE

Stroke is commonly caused by thromboembolic events originating from ruptured carotid plaque with vulnerable composition. This study assessed the performance of acoustic radiation force impulse (ARFI) imaging, a noninvasive ultrasound elasticity imaging method, for delineating the composition of human carotid plaque in vivo with histologic validation.

METHODS

Carotid ARFI images were captured before surgery in 25 patients undergoing clinically indicated carotid endarterectomy. The surgical specimens were histologically processed with sectioning matched to the ultrasound imaging plane. Three radiologists, blinded to histology, evaluated parametric images of ARFI-induced peak displacement to identify plaque features such as necrotic core (NC), intraplaque hemorrhage (IPH), collagen (COL), calcium (CAL), and fibrous cap (FC) thickness. Reader performance was measured against the histologic standard using receiver operating characteristic curve analysis, linear regression, Spearman correlation (ρ), and Bland-Altman analysis.

RESULTS

ARFI peak displacement was two-to-four-times larger in regions of NC and IPH relative to regions of COL or CAL. Readers detected soft plaque features (NC/IPH) with a median area under the curve of 0.887 (range, 0.867-0.924) and stiff plaque features (COL/CAL) with median area under the curve of 0.859 (range, 0.771-0.929). FC thickness measurements of two of the three readers correlated with histology (reader 1: R² = 0.64, ρ = 0.81; reader 2: R² = 0.89, ρ = 0.75).

CONCLUSIONS

This study suggests that ARFI is capable of distinguishing soft from stiff atherosclerotic plaque components and delineating FC thickness.

Statistical characterization of noise for spatial standardization of CT scans: Enabling comparison with multiple kernels and doses

Computerized tomography (CT) is a widely adopted modality for analyzing directly or indirectly functional, biological and morphological processes by means of the image characteristics. However, the potential utilization of the information obtained from CT images is often limited when considering the analysis of quantitative information involving different devices, acquisition protocols or reconstruction algorithms. Although CT scanners are calibrated as a part of the imaging workflow, the calibration is circumscribed to global reference values and does not circumvent problems that are inherent to the imaging modality. One of them is the lack of noise stationarity, which makes quantitative biomarkers extracted from the images less robust and stable. Some methodologies have been proposed for the assessment of non-stationary noise in reconstructed CT scans. However, those methods focused on the non-stationarity only due to the reconstruction geometry and are mainly based on the propagation of the variance of noise throughout the whole reconstruction process. Additionally, the philosophy followed in the state-of-the-art methods is based on the reduction of noise, but not in the standardization of it. This means that, even if the noise is reduced, the statistics of the signal remain non-stationary, which is insufficient to enable comparisons between different acquisitions with different statistical characteristics. In this work, we propose a statistical characterization of noise in reconstructed CT scans that leads to a versatile statistical model that effectively characterizes different doses, reconstruction kernels, and devices. The statistical model is generalized to deal with the partial volume effect via a localized mixture model that also describes the non-stationarity of noise. Finally, we propose a stabilization scheme to achieve stationary variance. The validation of the proposed methodology was performed with a physical phantom and clinical CT scans acquired with different configurations (kernels, doses, algorithms including iterative reconstruction). The results confirmed its suitability to enable comparisons with different doses, and acquisition protocols.

Ventricular Geometry From Non-contrast Non-ECG-gated CT Scans: An Imaging Marker of Cardiopulmonary Disease in Smokers

RATIONALE AND OBJECTIVES

Imaging-based assessment of cardiovascular structure and function provides clinically relevant information in smokers. Non-cardiac-gated thoracic computed tomographic (CT) scanning is increasingly leveraged for clinical care and lung cancer screening. We sought to determine if more comprehensive measures of ventricular geometry could be obtained from CT using an atlas-based surface model of the heart.

MATERIALS AND METHODS

Subcohorts of 24 subjects with cardiac magnetic resonance imaging (MRI) and 262 subjects with echocardiography were identified from COPDGene, a longitudinal observational study of smokers. A surface model of the heart was manually initialized, and then automatically optimized to fit the epicardium for each CT. Estimates of right and left ventricular (RV and LV) volume and free-wall curvature were then calculated and compared to structural and functional metrics obtained from MRI and echocardiograms.

RESULTS

CT measures of RV dimension and curvature correlated with similar measures obtained using MRI. RV and LV volume obtained from CT inversely correlated with echocardiogram-based estimates of RV systolic pressure using tricuspid regurgitation jet velocity and LV ejection fraction respectively. Patients with evidence of RV or LV dysfunction on echocardiogram had larger RV and LV dimensions on CT. Logistic regression models based on demographics and ventricular measures from CT had an area under the curve of >0.7 for the prediction of elevated right ventricular systolic pressure and ventricular failure.

CONCLUSIONS

These data suggest that non-cardiac-gated, non-contrast-enhanced thoracic CT scanning may provide insight into cardiac structure and function in smokers.

Identification of Ultrasonic Echolucent Carotid Plaques Using Discrete Fréchet Distance Between Bimodal Gamma Distributions

OBJECTIVE

Echolucent carotid plaques are associated with acute cardiovascular and cerebrovascular events (ACCEs) in atherosclerotic patients. The aim of this study was to develop a computer-aided method for identifying echolucent plaques.

METHODS

A total of 315 ultrasound images of carotid plaques (105 echo-rich, 105 intermediate, and 105 echolucent) collected from 153 patients were included in this study. A bimodal gamma distribution was proposed to model the pixel statistics in the gray scale images of plaques. The discrete Fréchet distance features (DFDFs) of each plaque were extracted based on the statistical model. The most discriminative features (MDFs) were obtained from DFDFs by the linear discriminant analysis, and a k-nearest-neighbor classifier was implemented for classification of different types of plaques.

RESULTS

The classification accuracy of the three types of plaques using MDFs can reach 77.46%. When a receiver operating characteristics curve was produced to identify echolucent plaques, the area under the curve was 0.831.

CONCLUSION

Our results indicate potential feasibility of the method for identifying echolucent plaques based on DFDFs.

SIGNIFICANCE

Our method may potentially improve the ability of noninvasive ultrasonic examination in risk prediction of ACCEs for patients with plaques.

Local Transverse-Slice-Based Level-Set Method for Segmentation of 3-D High-Frequency Ultrasonic Backscatter From Dissected Human Lymph Nodes

OBJECTIVE

To detect metastases in freshly excised human lymph nodes (LNs) using three-dimensional (3-D), high-frequency, quantitative ultrasound (QUS) methods, the LN parenchyma (LNP) must be segmented to preclude QUS analysis of data in regions outside the LNP and to compensate ultrasound attenuation effects due to overlying layers of LNP and residual perinodal fat (PNF).

METHODS

After restoring the saturated radio-frequency signals from PNF using an approach based on smoothing cubic splines, the three regions, i.e., LNP, PNF, and normal saline (NS), in the LN envelope data are segmented using a new, automatic, 3-D, three-phase, statistical transverseslice-based level-set (STS-LS) method that amends Lankton's method. Due to ultrasound attenuation and focusing effects, the speckle statistics of the envelope data vary with imaged depth. Thus, to mitigate depth-related inhomogeneity effects, the STS-LS method employs gamma probabilitydensity functions to locally model the speckle statistics within consecutive transverse slices.

RESULTS

Accurate results were obtained on simulated data. On a representative dataset of 54 LNs acquired from colorectal-cancer patients, the Dice similarity coefficient for LNP, PNF, and NS were 0.938 ± 0.025, 0.832 ± 0.086, and 0.968 ± 0.008, respectively, when compared to expert manual segmentation.

CONCLUSION

The STS-LS outperforms the established methods based on global and local statistics in our datasets and is capable of accurately handling the depth-dependent effects due to attenuation and focusing.

SIGNIFICANCE

This advance permits the automatic QUS-based cancer detection in the LNs. Furthermore, the STS-LS method could potentially be used in a wide range of ultrasound-imaging applications suffering from depth-dependent effects.

Influence of ultrasound speckle tracking strategies for motion and strain estimation

Speckle Tracking is one of the most prominent techniques used to estimate the regional movement of the heart based on ultrasound acquisitions. Many different approaches have been proposed, proving their suitability to obtain quantitative and qualitative information regarding myocardial deformation, motion and function assessment. New proposals to improve the basic algorithm usually focus on one of these three steps: (1) the similarity measure between images and the speckle model; (2) the transformation model, i.e. the type of motion considered between images; (3) the optimization strategies, such as the use of different optimization techniques in the transformation step or the inclusion of structural information. While many contributions have shown their good performance independently, it is not always clear how they perform when integrated in a whole pipeline. Every step will have a degree of influence over the following and hence over the final result. Thus, a Speckle Tracking pipeline must be analyzed as a whole when developing novel methods, since improvements in a particular step might be undermined by the choices taken in further steps. This work presents two main contributions: (1) We provide a complete analysis of the influence of the different steps in a Speckle Tracking pipeline over the motion and strain estimation accuracy. (2) The study proposes a methodology for the analysis of Speckle Tracking systems specifically designed to provide an easy and systematic way to include other strategies. We close the analysis with some conclusions and recommendations that can be used as an orientation of the degree of influence of the models for speckle, the transformation models, interpolation schemes and optimization strategies over the estimation of motion features. They can be further use to evaluate and design new strategy into a Speckle Tracking system.

A Review on Carotid Ultrasound Atherosclerotic Tissue Characterization and Stroke Risk Stratification in Machine Learning Framework

Cardiovascular diseases (including stroke and heart attack) are identified as the leading cause of death in today's world. However, very little is understood about the arterial mechanics of plaque buildup, arterial fibrous cap rupture, and the role of abnormalities of the vasa vasorum. Recently, ultrasonic echogenicity characteristics and morphological characterization of carotid plaque types have been shown to have clinical utility in classification of stroke risks. Furthermore, this characterization supports aggressive and intensive medical therapy as well as procedures, including endarterectomy and stenting. This is the first state-of-the-art review to provide a comprehensive understanding of the field of ultrasonic vascular morphology tissue characterization. This paper presents fundamental and advanced ultrasonic tissue characterization and feature extraction methods for analyzing plaque. Additionally, the paper shows how the risk stratification is achieved using machine learning paradigms. More advanced methods need to be developed which can segment the carotid artery walls into multiple regions such as the bulb region and areas both proximal and distal to the bulb. Furthermore, multimodality imaging is needed for validation of such advanced methods for stroke and cardiovascular risk stratification.

Unifying Concepts of Statistical and Spectral Quantitative Ultrasound Techniques

Quantitative ultrasound (QUS) techniques using radiofrequency (RF) backscattered signals have been used for tissue characterization of numerous organ systems. One approach is to use the magnitude and frequency dependence of backscatter echoes to quantify tissue structures. Another approach is to use first-order statistical properties of the echo envelope as a signature of the tissue microstructure. We propose a unification of these QUS concepts. For this purpose, a mixture of homodyned K-distributions is introduced to model the echo envelope, together with an estimation method and a physical interpretation of its parameters based on the echo signal spectrum. In particular, the total, coherent and diffuse signal powers related to the proposed mixture model are expressed explicitly in terms of the structure factor previously studied to describe the backscatter coefficient (BSC). Then, this approach is illustrated in the context of red blood cell (RBC) aggregation. It is experimentally shown that the total, coherent and diffuse signal powers are determined by a structural parameter of the spectral Structure Factor Size and Attenuation Estimator. A two-way repeated measures ANOVA test showed that attenuation (p-value of 0.077) and attenuation compensation (p-value of 0.527) had no significant effect on the diffuse to total power ratio. These results constitute a further step in understanding the physical meaning of first-order statistics of ultrasound images and their relations to QUS techniques. The proposed unifying concepts should be applicable to other biological tissues than blood considering that the structure factor can theoretically model any spatial distribution of scatterers.

Improved Segmentation of Multiple Cavities of the Heart in Wide-View 3-D Transesophageal Echocardiograms

Minimally invasive interventions in the heart such as in electrophysiology are becoming more and more important in clinical practice. Currently, preoperative computed tomography angiography (CTA) is used to provide anatomic information during electrophysiology interventions, but this does not provide real-time feedback and burdens the patient with additional radiation and side effects of the contrast agent. Three-dimensional transesophageal echocardiography (TEE) is an excellent modality for visualization of anatomic structures and instruments in real time, but some cavities, especially the left atrium, suffer from the limited coverage of the 3-D TEE volumes. This leads to difficulty in segmenting the left atrium. We propose replacing or complementing pre-operative CTA imaging with wide-view TEE. We tested this proposal on 20 patients for which TEE image volumes covering the left atrium and CTA images were acquired. The TEE images were manually registered, and wide-view volumes were generated. Five heart cavities in single-view and wide-view TEE were segmented and compared with atlas based-segmentations derived from the CTA images. We found that the segmentation accuracy (Dice coefficients) improved relative to segmentation of single-view images by 5, 15 and 9 percentage points for the left atrium, right atrium and aorta, respectively. Average anatomic coverage was improved by 2, 29, 62 and 49 percentage points for the right ventricle, left atrium, right atrium and aorta, respectively. This finding confirms that wide-view 3-D TEE can be useful in supporting electrophysiology interventions.

Segmentation of multiple heart cavities in 3-D transesophageal ultrasound images

Three-dimensional transesophageal echocardiography (TEE) is an excellent modality for real-time visualization of the heart and monitoring of interventions. To improve the usability of 3-D TEE for intervention monitoring and catheter guidance, automated segmentation is desired. However, 3-D TEE segmentation is still a challenging task due to the complex anatomy with multiple cavities, the limited TEE field of view, and typical ultrasound artifacts. We propose to segment all cavities within the TEE view with a multi-cavity active shape model (ASM) in conjunction with a tissue/blood classification based on a gamma mixture model (GMM). 3-D TEE image data of twenty patients were acquired with a Philips X7-2t matrix TEE probe. Tissue probability maps were estimated by a two-class (blood/tissue) GMM. A statistical shape model containing the left ventricle, right ventricle, left atrium, right atrium, and aorta was derived from computed tomography angiography (CTA) segmentations by principal component analysis. ASMs of the whole heart and individual cavities were generated and consecutively fitted to tissue probability maps. First, an average whole-heart model was aligned with the 3-D TEE based on three manually indicated anatomical landmarks. Second, pose and shape of the whole-heart ASM were fitted by a weighted update scheme excluding parts outside of the image sector. Third, pose and shape of ASM for individual heart cavities were initialized by the previous whole heart ASM and updated in a regularized manner to fit the tissue probability maps. The ASM segmentations were validated against manual outlines by two observers and CTA derived segmentations. Dice coefficients and point-to-surface distances were used to determine segmentation accuracy. ASM segmentations were successful in 19 of 20 cases. The median Dice coefficient for all successful segmentations versus the average observer ranged from 90% to 71% compared with an inter-observer range of 95% to 84%. The agreement against the CTA segmentations was slightly lower with a median Dice coefficient between 85% and 57%. In this work, we successfully showed the accuracy and robustness of the proposed multi-cavity segmentation scheme. This is a promising development for intraoperative procedure guidance, e.g., in cardiac electrophysiology.

Anisotropic diffusion filter with memory based on speckle statistics for ultrasound images

Ultrasound (US) imaging exhibits considerable difficulties for medical visual inspection and for development of automatic analysis methods due to speckle, which negatively affects the perception of tissue boundaries and the performance of automatic segmentation methods. With the aim of alleviating the effect of speckle, many filtering techniques are usually considered as a preprocessing step prior to automatic analysis methods or visual inspection. Most of the state-of-the-art filters try to reduce the speckle effect without considering its relevance for the characterization of tissue nature. However, the speckle phenomenon is the inherent response of echo signals in tissues and can provide important features for clinical purposes. This loss of information is even magnified due to the iterative process of some speckle filters, e.g., diffusion filters, which tend to produce over-filtering because of the progressive loss of relevant information for diagnostic purposes during the diffusion process. In this paper, we propose an anisotropic diffusion filter with a probabilistic-driven memory mechanism to overcome the over-filtering problem by following a tissue selective philosophy. In particular, we formulate the memory mechanism as a delay differential equation for the diffusion tensor whose behavior depends on the statistics of the tissues, by accelerating the diffusion process in meaningless regions and including the memory effect in regions where relevant details should be preserved. Results both in synthetic and real US images support the inclusion of the probabilistic memory mechanism for maintaining clinical relevant structures, which are removed by the state-of-the-art filters.

Fully automatic detection of salient features in 3-d transesophageal images

Most automated segmentation approaches to the mitral valve and left ventricle in 3-D echocardiography require a manual initialization. In this article, we propose a fully automatic scheme to initialize a multicavity segmentation approach in 3-D transesophageal echocardiography by detecting the left ventricle long axis, the mitral valve and the aortic valve location. Our approach uses a probabilistic and structural tissue classification to find structures such as the mitral and aortic valves; the Hough transform for circles to find the center of the left ventricle; and multidimensional dynamic programming to find the best position for the left ventricle long axis. For accuracy and agreement assessment, the proposed method was evaluated in 19 patients with respect to manual landmarks and as initialization of a multicavity segmentation approach for the left ventricle, the right ventricle, the left atrium, the right atrium and the aorta. The segmentation results revealed no statistically significant differences between manual and automated initialization in a paired t-test (p > 0.05). Additionally, small biases between manual and automated initialization were detected in the Bland-Altman analysis (bias, variance) for the left ventricle (-0.04, 0.10); right ventricle (-0.07, 0.18); left atrium (-0.01, 0.03); right atrium (-0.04, 0.13); and aorta (-0.05, 0.14). These results indicate that the proposed approach provides robust and accurate detection to initialize a multicavity segmentation approach without any user interaction.