Handbook of Speckle Filtering and Tracking in Cardiovascular Ultrasound Imaging and Video
2: Department of Computer Science, University of Cyprus, Limassol, Cyprus
3: Department of Cardiovascular Sciences, University of Leuven, Belgium
Ultrasound imaging technology has experienced a dramatic change in the last 30 years. Because of its non-invasive nature and continuing improvements in image quality, ultrasound imaging is progressively achieving an important role in the assessment and characterization of cardiovascular imaging. Speckle is inherent in ultrasound imaging giving rise to a granular appearance instead of homogeneous, flat shades of gray, as is visible and as such, speckle can severely compromise interpretation of ultrasound images, particularly in discrimination of small structures. On the other hand, speckle can be used in the detection of time varying phenomena, or tracking tissue motion. The objective of this book is to provide a reference edited volume covering the whole spectrum of speckle phenomena, theoretical background and modelling, algorithms and selected applications in cardiovascular ultrasound imaging and video processing and analysis. The book is organized under the following four parts, Part I: Introduction to Speckle Noise; Part II: Speckle Filtering; Part III: Speckle Tracking; Part IV: Selected Applications in Cardiovascular Imaging.
Inspec keywords: medical image processing; biomedical ultrasonics; object tracking; video signal processing; cardiovascular system; image filtering
Other keywords: speckle noise; speckle filtering; ultrasound video; cardiovascular ultrasound imaging; speckle tracking
Subjects: General and management topics; Digital signal processing; Biology and medical computing; Sonic and ultrasonic radiation (medical uses); General electrical engineering topics; Sonic and ultrasonic applications; Biomedical engineering; Sonic and ultrasonic radiation (biomedical imaging/measurement); Optical, image and video signal processing; Textbooks
- Book DOI: 10.1049/PBHE013E
- Chapter DOI: 10.1049/PBHE013E
- ISBN: 9781785612909
- e-ISBN: 9781785612916
- Page count: 706
- Format: PDF
-
Front Matter
- + Show details - Hide details
-
p.
(1)
-
Part I: Introduction to speckle noise in ultrasound imaging and video
1 A brief review of ultrasound imaging and video
- + Show details - Hide details
-
p.
3
–24
(22)
In this introductory chapter, a pedagogical overview on the technical foundations of ultrasound imaging will be provided to guide readers who are new to the biomedical ultrasound field. The intent here is to equip readers with fundamental principles that would serve well as background knowledge to understand the broad range of technical concepts covered in this book. Not only will the general physics of ultrasound and the key imaging considerations be outlined, engineering aspects such as system hardware will also be covered. In addition, commentary on the emerging trend toward high-frame-rate imaging will be included to highlight latest innovation thrusts in ultrasound imaging.
2 Speckle physics
- + Show details - Hide details
-
p.
25
–54
(30)
In this chapter, we will discuss the basic physical origins of speckle and present different, but related, mathematical descriptions of speckle in simple model systems. We will start from a general description of scattered waves as phasors, deriving first-order statistics applicable to both laser speckle and ultrasound speckle. Subsequently, we discuss the correspondences and differences between laser speckle and ultrasound speckle, before presenting higher order statistics and imaging implications in an ultrasound context. We also introduce the impact of the imaging system characteristics on speckle appearance and statistics.
3 Statistical models for speckle noise and Bayesian deconvolution of ultrasound images
- + Show details - Hide details
-
p.
55
–75
(21)
This chapter first summarizes existing results related to the statistical properties of US images based on both radio-frequency (RF) and envelope signals. In a second part of the chapter, we present a Bayesian deconvolution method that can be viewed as a general despeckling technique.
4 Summary
- + Show details - Hide details
-
p.
77
(1)
This chapter summaries Part 1 of this book.
-
Part II: Speckle filtering
5 Introduction to speckle filtering
- + Show details - Hide details
-
p.
81
–93
(13)
This chapter aims to introduce the reader into the field of speckle filtering, by emphasizing its particular characteristics that are relevant for every practitioner in the field. We begin with an introduction to filtering in medical imaging where we stress the importance of information preservation over complete filtering. This broad view motivates the following discussion on the main issues concerning speckle filtering and provides a reasonable framework to approach it. We elaborate on topics that we believe are rather specific to speckle and should be taken into account whenever the different filters described in following chapters are utilized.
6 An overview of despeckle-filtering techniques
- + Show details - Hide details
-
p.
95
–109
(15)
This chapter provides an introduction and a brief overview of selected despeckle-filtering techniques for ultrasound imaging and video expanded from [1]. A despeckle-filtering evaluation protocol is proposed, a brief literature review, as well as an image despeckle-filtering (IDF) toolbox [3] and a video despeckle filtering (VDF) [4] software toolbox are presented. Moreover, selected applications for ultrasound image and VDF techniques are illustrated. Speckle is a multiplicative noise [1-4], which degrades ultrasound images and videos and negatively influences the image and video interpretation, diagnosis and visual appearance [5]. Noise speckle reduction is therefore essential for improving the visual observation quality or as a preprocessing step for further automated analysis, such as image/ video segmentation, texture analysis and encoding in ultrasound image and video. On the other hand, speckle can also be used as an information carrier on the underlying tissue properties. As illustrated in the previous chapters in Part I (see also Part IV), this implies it can be used, for example, for tissue classification. Alternatively, assuming that speckle moves in the image in the same way as the underlying tissue, it allows for tissue motion estimation using one of the many speckle tracking approaches presented in literature. A large number of despeckle-filtering techniques have been proposed in the past years for ultrasound images and very few for videos, which is usually applied for improving their visualization and interpretation or as a preprocessing step for further image/video analysis. This analysis includes segmentation, feature extraction, image and video compression, data transfer and registration. The present review study discusses, compares and evaluates ultrasound image and VDF techniques for the common carotid artery (CCA) introduced so far in the literature. Applications of the techniques are presented on simulated and real ultrasound images and videos (see also Chapters 7-10) of the CCA.
7 Linear despeckle filtering
- + Show details - Hide details
-
p.
111
–152
(42)
This chapter provides the basic theoretical background of linear despeckle filtering techniques together with their algorithmic implementation, MATLAB® code for selected filters and practical examples on phantom and real ultrasound images. There are three groups of filters presented in this chapter, first-order statistics filtering, local statistics filtering and homogeneous mask area filtering. Despeckle filtering was evaluated for all filters presented in this chapter on phantom ultrasound carotid artery images and real ultrasound images and videos of the common carotid artery (CCA). Furthermore, we present an evaluation and comparison of five linear despeckle filtering algorithms presented in this chapter. The evaluation is carried out on a phantom image, an artificial image and on real carotid and cardiac ultrasound images. Furthermore, findings on video despeckling are presented.
8 Nonlinear despeckle filtering
- + Show details - Hide details
-
p.
153
–171
(19)
In this chapter, we will review some of the methods proposed in literature to remove the speckle pattern from ultrasound data based on nonlinear processing. Note that some filters that can also be considered as nonlinear are left aside since they will be deeply treated in other chapters. That is the case of the wavelet-based methods and Bayesian methods. Other methods also treated in other chapters, like diffusion-based schemes are only briefly reviewed in order to place them inside the global partial differential equation (PDE) classification. On the other hand, note that some of the filters here reviewed are not initially proposed for ultrasound imaging but derived for Synthetic Aperture Radar (SAR) images, where noise can be modeled similarly. In those images, the multiplicative model for speckle holds and therefore, many of the methods defined in literature for SAR can be easily extrapolated to ultrasound denoising. This is the case of some of the most popular speckle filters. In addition, we would like to remark that the effectiveness of many of these schemes lays on a proper modeling of the speckle statistics. For some purposes, a simple multiplicative model will suffice, while for some specific applications, more accurate models must be used. Finally, as we stated in the previous chapter, the filtering method must be selected following the specific needs of the problem. There is no all-purpose filter that, with the same configuration parameters, could perform excellent in all situations.
9 Wavelet despeckle filtering
- + Show details - Hide details
-
p.
173
–199
(27)
In this chapter, the effect of transform features (shift sensitivity and directional selectivity) has been examined by implementing the homomorphic and non-homomorphic speckle suppressors in three alternative wavelet domains-DWT, RDWT and CWT. The experimental results demonstrate that performance of DWT is worst on all type of images whereas the performance of RDWT and DT-CWT is comparable. On certain types of images (images dominated by uniform areas), RDWT performs better while on others (textured images), DT-CWT-based methods are the best. This shows that the shift-dependence of a transform causes a significant degradation in performance of a despeckling technique and good directional selectivity is essential for representing the textured images optimally. From these investigations, it is concluded that both DT-CWT and RDWT are equally good for designing wavelet-based de-noising applications. However, the low computational complexity of DT-CWT and the textured nature of medical US images, favours the use of DT-CWT in comparison to RDWT for despeckling applications. More work is required to develop shorter length filters for the DTCWT (e.g. analogous to Haar wavelet) in order to further improve the despeckling performance of the US images.
10 A comparative evaluation on linear and nonlinear despeckle filtering techniques
- + Show details - Hide details
-
p.
201
–232
(32)
In this chapter, the methods of texture analysis, image quality evaluation, distance measures, univariate statistical analysis and the k-nearest-neighbor (kNN) classifier, which are used to evaluate despeckle filtering on ultrasound imaging and video, are presented. For speckle reduction, 16 different despeckle filtering methods, already described in [1,2], were applied to each image or video prior to intima-media complex (IMC) or atherosclerotic plaque segmentation. Despeckle filtering was applied after image or video normalization , either to the entire image or to an ROI, selected by the user. The selected area of interest (ROI) can be of any shape but the image despeckle filtering (IDF) software doesn't support multiple ROIs selection. In the latter case, where the user of the system is interested only in the selected ROI, the area outside the ROI can be blurred using the DsFlsmv filter operating with a sliding moving window of [15x15] pixels and a number iterations 5. It should be noted that the blurring is applied outside of the ROI if the user of the system is not interested to subjectively evaluate this area. The input parameters of the 16 different despeckle filters for the IDF and video despeckle filtering (VDF) software toolboxes can be selected by the user as it was documented in [1-4]. The 16 despeckle filters evaluated in this chapter were applied on a large number of asymptomatic (AS) and of symptomatic (SY) ultrasound images (220 vs 220) of the common carotid artery (CCA). Four despeckle filters (DsFlsmv, DsFhmedian, DsFkuwahara, DsFsrad) were further applied to ten videos of the carotid artery bifurcation. A large number of texture features (61 different texture features) were extracted from the original and despeckle images and videos, and the most discriminant ones are presented. The performance of these filters is investigated for discriminating between AS and SY images using the statistical kNN classifier. Moreover, 16 different image quality evaluation metrics were computed, as well as visual evaluation scores carried out by two experts.
11 Summary and future directions
- + Show details - Hide details
-
p.
233
–237
(5)
This chapter summaries Part II of this book.
-
Part III: Speckle tracking
12 Introduction to speckle tracking in ultrasound video
- + Show details - Hide details
-
p.
241
–257
(17)
Ultrasound imaging is widely used in the medical field since the modality is relatively cheap and can be applied nearly in all clinical environments due to its portability. Static images have been used to assess anatomical and geometrical features, but one of the unique features of ultrasound is its capability of examining dynamic events. In addition to anatomical and echogenicity features, ultrasound can provide information regarding movement of tissues. Quantification of tissue motion will be of interest in fundamental and clinical questions; from the motion, the deformability of the tissue can be quantified. When this deformation is induced by a force applied onto the tissue, the deformation is associated with its mechanical structure and composition. But it can also reveal functional behaviour when the deformation is representing the function of the targeted tissue. There is a vast amount of ultrasound techniques for the detection of tissue motion (functional imaging). For many years, M-mode imaging played an important role in evaluation of rapid motions because of its high sampling rate. Other techniques based on the Doppler effect or applying block-matching algorithms for tracking tissue motion are available. Nowadays, as a result of rapid developments in ultrafast ultrasound imaging, techniques are available that permit fast and complex motions to be measured more accurately. This chapter will introduce the most commonly used techniques in clinical practice and will provide an overview of past and current developments in functional ultrasound imaging.
13 Principles of speckle tracking
- + Show details - Hide details
-
p.
259
–287
(29)
A vast amount of speckle tracking techniques has been developed to extract tissue motion from a time series of medical US images. The purpose of this chapter is to explore the underlying principles of these techniques while highlighting the methodological variety. This diversity also implies that arguably many different classification schemes could be devised to group them. Furthermore, it is important to realize that labeling techniques belonging to a single category can be further complicated given that some methods are conceptually hybrid approaches that combine the strengths of their individual constituents while mitigating their disadvantages as much as possible. As such, rather than being an exhaustive overview, this chapter should instead be used as a practical guide in the ever-evolving US tracking literature.
14 Techniques for speckle tracking: block matching
- + Show details - Hide details
-
p.
289
–320
(32)
The clinical need for in vivo motion and deformation quantification and the many applications have been clearly explained in the previous chapters. In general, we distinguish between (1) quasi-static elastography: speckle tracking during compression or palpation of tissue to estimate local strains and elasticity and (2) dynamic elastography, where speckle tracking on actively deforming tissue such as the heart, arteries and skeletal muscles is performed. The third class, where motion of the tissue is induced with a vibrational device or a push by the transducer (ARFI, shear wave imaging) is not dealt with in this chapter. We will elaborate on a sub-set of speckle tracking and/or strain imaging techniques, which are based on the so-called block-matching techniques. Hence, Doppler-based techniques will not be discussed.
15 Techniques for tracking: image registration
- + Show details - Hide details
-
p.
321
–351
(31)
This chapter is focused on the image registration techniques that estimate the motion and strain by tracking the speckle pattern directly from the intensity of the B-mode US images assuming a temporal relationship between images. The fixed and moving images are noted as It and It-1 with t meaning a specific time.
16 Cardiac strain estimation
- + Show details - Hide details
-
p.
353
–376
(24)
An important application domain of ultrasound-based motion estimation is the heart which leads to specific challenges and boundary conditions. In this chapter, a more detailed discussion is given on applying speckle tracking algorithms for cardiac applications.
17 Combined techniques of filtering and speckle tracking
- + Show details - Hide details
-
p.
377
–392
(16)
Speckle tracking allows estimating tissue motion with ultrasound. Since tracking quality directly depends on image quality, several preprocessing techniques have been used to improve image quality and ease the tracking problem. The objective of this chapter is therefore to review such techniques and to provide some qualitative guidelines as to in which conditions they should and should not be employed.
18 Summary and future directions
- + Show details - Hide details
-
p.
393
–401
(9)
Although ultrasound motion estimation has many applications in distinct scientific domains, within the medical field, motion estimation naturally goes hand in hand with cardiac imaging. As such, it is not surprising to see that most practical examples discussed in this part of the book relate to the heart despite the fact the motion estimation methodologies presented are generic.
-
Part IV: Selected applications
19 Segmentation of the carotid artery IMT in ultrasound
- + Show details - Hide details
-
p.
405
–424
(20)
In this chapter, we will present the most widely used computer techniques for the carotid IMT segmentation from cardiovascular ultrasound images. We will focus on ultrasound image and carotid artery characteristics, and the difference between semiautomatic and completely automatic segmentation methods, also presenting numerous used and established techniques.
20 Ultrasound carotid plaque video segmentation
- + Show details - Hide details
-
p.
425
–443
(19)
Border identification of the atherosclerotic carotid plaque, the common carotid artery (CCA), degree of stenosis, as well as the characteristics of the arterial wall (plaque size, composition and elasticity), may add additional clinical information for the assessment of future cardiovascular events. We propose and evaluate in this chapter an integrated system for the segmentation of atherosclerotic carotid plaque in ultrasound videos of the CCA. The system is based on video frame normalization, speckle reduction filtering, M-mode-state-based identification, parametric active contours and snake's segmentation. The cardiac cycle in each video is first identified and the video M-mode is generated, thus identifying systolic and diastolic states. The video is segmented for a time period of at least one full cardiac cycle by initializing the algorithm in the first video frame. Human manual assistance may be provided if needed. The atherosclerotic plaque borders are tracked and segmented in the subsequent frames. We also propose an initialization method for positioning the snake as close as possible to the plaque borders, based on morphology operators, where initial contours are estimated every 20 video frames. The performance of the algorithm is evaluated on 43 real CCA digitized videos from B-mode longitudinal ultrasound segments and is compared to the manual segmentations of an expert, available every 20 frames in a time span of 3-5 s, covering in general two cardiac cycles. The segmentation results were very promising, according to the expert objective evaluation, with a true-negative fraction (TNF) specificity of 83.7% + 7.6%, a true-positive fraction (TPF) sensitivity of 85.42% + 8.1%, between the ground truth and the proposed segmentation method, a kappa index (KI) of 84.6% and an overlap index (OI) of 74.7%. We also computed the cardiac state identification for the CCA. It is shown that the integrated system presented in this chapter can be used for the video segmentation of the CCA plaque in ultrasound videos.
21 Ultrasound asymptomatic carotid plaque image analysis for the prediction of the risk of stroke
- + Show details - Hide details
-
p.
445
–459
(15)
This chapter presents methods that can be used for evaluation of high cerebrovascular risk using the texture and morphological appearance of plaques in combination with a number of clinical factors for the estimation of the stroke risk. Images and results presented are based on the analysis of images that were collected through a multicenter cohort study of patients with asymptomatic ICA called Asymptomatic Carotid Stenosis and Risk of Stroke (ACSRS).
22 3D segmentation and texture analysis of the carotid arteries
- + Show details - Hide details
-
p.
461
–495
(35)
In this chapter, we review 3D ultrasound-based methods for segmentation of carotid plaques and their use in quantifying plaque composition using image texture metrics. Specifically, we review algorithms used to segment the media-adventitia and lumen-intima of CCA, internal carotid artery (ICA), and external carotid artery (ECA). We also review methods, which have used these segmented boundaries to provide information on plaque composition using image texture metrics.
23 Carotid artery mechanics assessed by ultrasound
- + Show details - Hide details
-
p.
497
–517
(21)
The carotid artery is an important blood vessel, supplying the brain with oxygen and nutrients. Two carotid arteries are found, one on either side of the neck. It bifurcates into the external carotid artery, that is mainly responsible for supplying the facial muscles with blood, and the internal carotid artery that ensures blood flow to the brain. It connects (together with the vertebral arteries) to the circle of Willis, from which the entire brain is supplied with blood. The circle of Willis introduces redundancy of supplying vessels, thus ensuring that occlusion of one carotid artery does not result in cerebral ischemia.
24 Carotid artery wall motion and strain analysis using tracking
- + Show details - Hide details
-
p.
519
–539
(21)
In this chapter, a number of speckle-tracking-based methodologies are outlined, suitable for the estimation of motion and strain of the carotid artery from ultrasound images. Various versions of intensity- and phase-based techniques have been suggested and validated mostly in phantoms, in in silico and in vitro data. Waveforms showing displacements, velocities and accelerations can be obtained from these methods, and a number of indices can then be calculated. Spatial mapping (imaging) of tissue strains can be achieved with elastography, a major application of motion analysis. Through their application in real data, these methods are promising for revealing valuable quantitative in vivo information about arterial mechanics. Compared to other ultrasound-based indices, a major advantage of motion-derived indices is that they provide functional, rather than mere anatomical, information, which is more sensitive to early wall changes. Their full potential in predicting, diagnosing and monitoring carotid-related disorders, such as cerebrovascular events, as well as in characterising the burden of other diseases, remains to be confirmed in large clinical trials, towards an integrated personalised approach for disease management and increased patient safety.
25 IVUS tracking: advantages and disadvantages of intravascular ultrasound in the detection of artery geometrical features and plaque type morphology
- + Show details - Hide details
-
p.
541
–569
(29)
Intravascular ultrasound (IVUS) is an invasive coronary imaging modality which provides insights in the diagnosis and therapy of coronary artery disease. The main advantage of IVUS compared to other coronary imaging techniques such as traditional invasive angiography, is that it allows the tomographic assessment of: the lumen area, the external elastic membrane area, the plaque size, the distribution and the composition of the plaque. It can safely detect and characterize the following types of plaques: necrotic core, fibrous plaque and calcified plaque. One other aspect worth mentioning is that it also provides accurate measurement of the luminal dimensions of the artery of interest. With IVUS, lipid-laden lesions appear hypoechoic, fibromuscular lesions generate low-intensity echoes and fibrous or calcified tissues are echogenic. However, IVUS suffers from a few drawbacks directly connected to its innate image quality. For example, calcium obscures the underlying wall, a phenomenon called acoustic shadowing. One other major image quality issue that emerges from the use of IVUS is the so-called speckle noise, which is the main reason for the grainy texture of IVUS images. This results to poor quality images, which makes their interpretation and processing nontrivial for computed-based diagnostic systems.
26 Introduction to speckle tracking in cardiac ultrasound imaging
- + Show details - Hide details
-
p.
571
–598
(28)
In this chapter, we will first recall some basic principles of speckle tracking. The fundamentals of speckle tracking in a wider context are essentially described in Chapter 13. We will then treat speckle-tracking echocardiography and echocardiographic particle image velocimetry (echo-PIV) and indicate a number of clinical applications in the context of evaluation of cardiac function. We will then briefly introduce color Doppler approaches complementary to speckle tracking. We will finally present how speckle-tracking techniques could benefit from highframe-rate echocardiography (also called “ultrafast echocardiography”). We will conclude with the expected contribution of high-frame-rate ultrasound for speckle tracking in three-dimensional (3-D) echocardiography.
27 Assessment of systolic and diastolic heart failure
- + Show details - Hide details
-
p.
599
–619
(21)
Heart failure is becoming more common in western societies as their populations age. In the Rotterdam community study of nearly 8,000 people, the lifetime risk of developing heart failure from the age of 55 years was 33 per cent for men and 29 per cent for women, and its prevalence in those aged more than 85 years was 17 per cent [1]. Since heart failure is a severe disease with a poor prognosis, imaging is important for early detection, accurate diagnosis, estimating prognosis, and planning and monitoring responses to treatment. Many diagnostic targets can be studied to establish the particular phenotype in a person who is suspected of having heart failure (Table 27.1), and most invasive and non-invasive diagnostic techniques have been used to study them. In this chapter, we concentrate on evidence relating to the use of speckle tracking to study left ventricular, left atrial, and right ventricular function in heart failure. We review its technical strengths and limitations from a clinical perspective and describe how it is applied to determine the pathophysiological mechanisms of heart failure, to identify the underlying aetiology and to estimate prognosis. We also suggest some technical developments that if feasible might assist clinicians in improving treatment and outcomes for patients with heart failure.
28 Myocardial elastography and electromechanical wave imaging
- + Show details - Hide details
-
p.
621
–653
(33)
Cardiovascular diseases remain America's primary killer by a large margin, claiming the lives of more Americans than the next two main causes of death combined (cancer and pulmonary complications). In particular, coronary artery disease (CAD) is by far the most lethal, causing 17% of all (cardiac related or not) deaths every year. One of the main reasons for this high death toll is the severe lack of effective and accessible imaging tools upon anomaly detected on the electrocardiogram, especially at the early stages when CAD can be stabilized with appropriate pharmacological regimen. On the other hand, arrhythmias refer to the disruption of the natural heart rhythm. Cardiac arrhythmias lead to a significant number of cardiovascular morbidity and mortality. This irregular heart rhythm causes the heart to suddenly stop pumping blood. Atrial pathologies are the most common arrhythmias with atrial fibrillation and atrial flutter being the most prevalent.
29 Summary and future directions
- + Show details - Hide details
-
p.
655
–660
(6)
-
Back Matter
- + Show details - Hide details
-
p.
(1)