Efficient ultrasonic signal processing techniques for aided medical diagnostics

Ultrasound imaging is widely used in medical diagnostics as it is the fastest, least invasive, and least expensive imaging modality. However, ultrasound images are intrinsically difficult to be interpreted. In this scenario, Computer Aided Detection (CAD) systems can be used to support physicians during diagnosis providing them a second opinion. This thesis discusses efficient ultrasound processing techniques for computer aided medical diagnostics, focusing on two major topics: (i) Ultrasound Tissue Characterization (UTC), aimed at characterizing and differentiating between healthy and diseased tissue; (ii) Ultrasound Image Segmentation (UIS), aimed at detecting the boundaries of anatomical structures to automatically measure organ dimensions and compute clinically relevant functional indices. Research on UTC produced a CAD tool for Prostate Cancer detection to improve the biopsy protocol. In particular, this thesis contributes with: (i) the development of a robust classification system; (ii) the exploitation of parallel computing on GPU for real-time performance; (iii) the introduction of both an innovative Semi-Supervised Learning algorithm and a novel supervised/semi-supervised learning scheme for CAD system training that improve system performance reducing data collection effort and avoiding collected data wasting. The tool provides physicians a risk map highlighting suspect tissue areas, allowing them to perform a lesion-directed biopsy. Clinical validation demonstrated the system validity as a diagnostic support tool and its effectiveness at reducing the number of biopsy cores requested for an accurate diagnosis. For UIS the research developed a heart disease diagnostic tool based on Real-Time 3D Echocardiography. Thesis contributions to this application are: (i) the development of an automated GPU based level-set segmentation framework for 3D images; (ii) the application of this framework to the myocardium segmentation. Experimental results showed the high efficiency and flexibility of the proposed framework. Its effectiveness as a tool for quantitative analysis of 3D cardiac morphology and function was demonstrated through clinical validation.

Mesh-based vs. Image-based Statistical Appearance Model of the Human Femur: a Preliminary Comparison Study for the Creation of Finite Element Meshes

Statistical models have been recently introduced in computational orthopaedics to investigate the bone mechanical properties across several populations. A fundamental aspect for the construction of statistical models concerns the establishment of accurate anatomical correspondences among the objects of the training dataset. Various methods have been proposed to solve this problem such as mesh morphing or image registration algorithms. The objective of this study is to compare a mesh-based and an image-based statistical appearance model approaches for the creation of nite element(FE) meshes. A computer tomography (CT) dataset of 157 human left femurs was used for the comparison. For each approach, 30 finite element meshes were generated with the models. The quality of the obtained FE meshes was evaluated in terms of volume, size and shape of the elements. Results showed that the quality of the meshes obtained with the image-based approach was higher than the quality of the mesh-based approach. Future studies are required to evaluate the impact of this finding on the final mechanical simulations.

Digital signal processing and digital system design using discrete cosine transform [student course]

The discrete cosine transform (DCT) is an important functional block for image processing applications. The implementation of a DCT has been viewed as a specialized research task. We apply a micro-architecture based methodology to the hardware implementation of an efficient DCT algorithm in a digital design course. Several circuit optimization and design space exploration techniques at the register-transfer and logic levels are introduced in class for generating the final design. The students not only learn how the algorithm can be implemented, but also receive insights about how other signal processing algorithms can be translated into a hardware implementation. Since signal processing has very broad applications, the study and implementation of an extensively used signal processing algorithm in a digital design course significantly enhances the learning experience in both digital signal processing and digital design areas for the students.

Distance measures for image segmentation evaluation

The task considered in this paper is performance evaluation of region segmentation algorithms in the ground-truth-based paradigm. Given a machine segmentation and a ground-truth segmentation, performance measures are needed. We propose to consider the image segmentation problem as one of data clustering and, as a consequence, to use measures for comparing clusterings developed in statistics and machine learning. By doing so, we obtain a variety of performance measures which have not been used before in image processing. In particular, some of these measures have the highly desired property of being a metric. Experimental results are reported on both synthetic and real data to validate the measures and compare them with others.

Digital processing and data compilation approach for using remotely sensed imagery to identify geological lineaments in hard-rock terrains : an application for groundwater explorations in Nicaragua

Sustainable yields from water wells in hard-rock aquifers are achieved when the well bore intersects fracture networks. Fracture networks are often not readily discernable at the surface. Lineament analysis using remotely sensed satellite imagery has been employed to identify surface expressions of fracturing, and a variety of image-analysis techniques have been successfully applied in “ideal” settings. An ideal setting for lineament detection is where the influences of human development, vegetation, and climatic situations are minimal and hydrogeological conditions and geologic structure are known. There is not yet a well-accepted protocol for mapping lineaments nor have different approaches been compared in non-ideal settings. A new approach for image-processing/synthesis was developed to identify successful satellite imagery types for lineament analysis in non-ideal terrain. Four satellite sensors (ASTER, Landsat7 ETM+, QuickBird, RADARSAT-1) and a digital elevation model were evaluated for lineament analysis in Boaco, Nicaragua, where the landscape is subject to varied vegetative cover, a plethora of anthropogenic features, and frequent cloud cover that limit the availability of optical satellite data. A variety of digital image processing techniques were employed and lineament interpretations were performed to obtain 12 complementary image products that were evaluated subjectively to identify lineaments. The 12 lineament interpretations were synthesized to create a raster image of lineament zone coincidence that shows the level of agreement among the 12 interpretations. A composite lineament interpretation was made using the coincidence raster to restrict lineament observations to areas where multiple interpretations (at least 4) agree. Nine of the 11 previously mapped faults were identified from the coincidence raster. An additional 26 lineaments were identified from the coincidence raster, and the locations of 10 were confirmed by field observation. Four manual pumping tests suggest that well productivity is higher for wells proximal to lineament features. Interpretations from RADARSAT-1 products were superior to interpretations from other sensor products, suggesting that quality lineament interpretation in this region requires anthropogenic features to be minimized and topographic expressions to be maximized. The approach developed in this study has the potential to improve siting wells in non-ideal regions.

The influence of colour on oculomotor behaviour during image perception

The aim of this study was to investigate how oculomotor behaviour depends on the availability of colour information in pictorial stimuli. Forty study participants viewed complex images in colour or grey-scale, while their eye movements were recorded. We found two major effects of colour. First, although colour increases the complexity of an image, fixations on colour images were shorter than on their grey-scale versions. This suggests that colour enhances discriminability and thus affects low-level perceptual processing. Second, colour decreases the similarity of spatial fixation patterns between participants. The role of colour on visual attention seems to be more important than previously assumed, in theoretical as well as methodological terms.

Application of MFBD Algorithms to Image Reconstruction Under Anisoplanatic Conditions

All optical systems that operate in or through the atmosphere suffer from turbulence induced image blur. Both military and civilian surveillance, gun-sighting, and target identification systems are interested in terrestrial imaging over very long horizontal paths, but atmospheric turbulence can blur the resulting images beyond usefulness. My dissertation explores the performance of a multi-frame-blind-deconvolution technique applied under anisoplanatic conditions for both Gaussian and Poisson noise model assumptions. The technique is evaluated for use in reconstructing images of scenes corrupted by turbulence in long horizontal-path imaging scenarios and compared to other speckle imaging techniques. Performance is evaluated via the reconstruction of a common object from three sets of simulated turbulence degraded imagery representing low, moderate and severe turbulence conditions. Each set consisted of 1000 simulated, turbulence degraded images. The MSE performance of the estimator is evaluated as a function of the number of images, and the number of Zernike polynomial terms used to characterize the point spread function. I will compare the mean-square-error (MSE) performance of speckle imaging methods and a maximum-likelihood, multi-frame blind deconvolution (MFBD) method applied to long-path horizontal imaging scenarios. Both methods are used to reconstruct a scene from simulated imagery featuring anisoplanatic turbulence induced aberrations. This comparison is performed over three sets of 1000 simulated images each for low, moderate and severe turbulence-induced image degradation. The comparison shows that speckle-imaging techniques reduce the MSE 46 percent, 42 percent and 47 percent on average for low, moderate, and severe cases, respectively using 15 input frames under daytime conditions and moderate frame rates. Similarly, the MFBD method provides, 40 percent, 29 percent, and 36 percent improvements in MSE on average under the same conditions. The comparison is repeated under low light conditions (less than 100 photons per pixel) where improvements of 39 percent, 29 percent and 27 percent are available using speckle imaging methods and 25 input frames and 38 percent, 34 percent and 33 percent respectively for the MFBD method and 150 input frames. The MFBD estimator is applied to three sets of field data and the results presented. Finally, a combined Bispectrum-MFBD Hybrid estimator is proposed and investigated. This technique consistently provides a lower MSE and smaller variance in the estimate under all three simulated turbulence conditions.

Looking left with left neglect: the role of spatial attention when active vision selects local image features for fixation

When we actively explore the visual environment, our gaze preferentially selects regions characterized by high contrast and high density of edges, suggesting that the guidance of eye movements during visual exploration is driven to a significant degree by perceptual characteristics of a scene. Converging findings suggest that the selection of the visual target for the upcoming saccade critically depends on a covert shift of spatial attention. However, it is unclear whether attention selects the location of the next fixation uniquely on the basis of global scene structure or additionally on local perceptual information. To investigate the role of spatial attention in scene processing, we examined eye fixation patterns of patients with spatial neglect during unconstrained exploration of natural images and compared these to healthy and brain-injured control participants. We computed luminance, colour, contrast, and edge information contained in image patches surrounding each fixation and evaluated whether they differed from randomly selected image patches. At the global level, neglect patients showed the characteristic ipsilesional shift of the distribution of their fixations. At the local level, patients with neglect and control participants fixated image regions in ipsilesional space that were closely similar with respect to their local feature content. In contrast, when directing their gaze to contralesional (impaired) space neglect patients fixated regions of significantly higher local luminance and lower edge content than controls. These results suggest that intact spatial attention is necessary for the active sampling of local feature content during scene perception.

High Resolution Image Correspondences for Video Post-Production

We present an algorithm for estimating dense image correspondences. Our versatile approach lends itself to various tasks typical for video post-processing, including image morphing, optical flow estimation, stereo rectification, disparity/depth reconstruction, and baseline adjustment. We incorporate recent advances in feature matching, energy minimization, stereo vision, and data clustering into our approach. At the core of our correspondence estimation we use Efficient Belief Propagation for energy minimization. While state-of-the-art algorithms only work on thumbnail-sized images, our novel feature downsampling scheme in combination with a simple, yet efficient data term compression, can cope with high-resolution data. The incorporation of SIFT (Scale-Invariant Feature Transform) features into data term computation further resolves matching ambiguities, making long-range correspondence estimation possible. We detect occluded areas by evaluating the correspondence symmetry, we further apply Geodesic matting to automatically determine plausible values in these regions.

Faces and emotions: brain electric field sources during covert emotional processing

Covert brain activity related to task-free, spontaneous (i.e. unrequested), emotional evaluation of human face images was analysed in 27-channel averaged event-related potential (ERP) map series recorded from 18 healthy subjects while observing random sequences of face images without further instructions. After recording, subjects self-rated each face image on a scale from “liked” to “disliked”. These ratings were used to dichotomize the face images into the affective evaluation categories of “liked” and “disliked” for each subject and the subjects into the affective attitudes of “philanthropists” and “misanthropists” (depending on their mean rating across images). Event-related map series were averaged for “liked” and “disliked” face images and for “philanthropists” and “misanthropists”. The spatial configuration (landscape) of the electric field maps was assessed numerically by the electric gravity center, a conservative estimate of the mean location of all intracerebral, active, electric sources. Differences in electric gravity center location indicate activity of different neuronal populations. The electric gravity center locations of all event-related maps were averaged over the entire stimulus-on time (450 ms). The mean electric gravity center for disliked faces was located (significant across subjects) more to the right and somewhat more posterior than for liked faces. Similar differences were found between the mean electric gravity centers of misanthropists (more right and posterior) and philanthropists. Our neurophysiological findings are in line with neuropsychological findings, revealing visual emotional processing to depend on affective evaluation category and affective attitude, and extending the conclusions to a paradigm without directed task.

Dual-Domain Image Denoising

Image denoising methods have been implemented in both spatial and transform domains. Each domain has its advantages and shortcomings, which can be complemented by each other. State-of-the-art methods like block-matching 3D filtering (BM3D) therefore combine both domains. However, implementation of such methods is not trivial. We offer a hybrid method that is surprisingly easy to implement and yet rivals BM3D in quality.

A new approach to detect structural differences in lung parenchyma using digital image analysis

Morphometric investigations using a point and intersection counting strategy in the lung often are not able to reveal the full set of morphologic changes. This happens particularly when structural modifications are not expressed in terms of volume density changes and when rough and fine surface density alterations cancel each other at different magnifications. Making use of digital image processing, we present a methodological approach that allows to easily and quickly quantify changes of the geometrical properties of the parenchymal lung structure and reflects closely the visual appreciation of the changes. Randomly sampled digital images from light microscopic sections of lung parenchyma are filtered, binarized, and skeletonized. The lung septa are thus represented as a single-pixel wide line network with nodal points and end points and the corresponding internodal and end segments. By automatically counting the number of points and measuring the lengths of the skeletal segments, the lung architecture can be characterized and very subtle structural changes can be detected. This new methodological approach to lung structure analysis is highly sensitive to morphological changes in the parenchyma: it detected highly significant quantitative alterations in the structure of lungs of rats treated with a glucocorticoid hormone, where the classical morphometry had partly failed.

Digital Data Collection and Processing in Field and Lab: The Approach of Kinneret Regional Project

The paper showcases the field- and lab-documentation system developed for Kinneret Regional Project, an international archaeological expedition to the Northwestern shore of the Sea of Galilee (Israel) under the auspices of the University of Bern, the University of Helsinki, Leiden University and Wofford College. The core of the data management system is a fully relational, server-based database framework, which also includes time-based and static GIS services, stratigraphic analysis tools and fully indexed document/digital image archives. Data collection in the field is based on mobile, hand-held devices equipped with a custom-tailored stand-alone application. Comprehensive three-dimensional documentation of all finds and findings is achieved by means of total stations and/or high-precision GPS devices. All archaeological information retrieved in the field – including tachymetric data – is synched with the core system on the fly and thus immediately available for further processing in the field lab (within the local network) or for post-excavation analysis at remote institutions (via the WWW). Besides a short demonstration of the main functionalities, the paper also presents some of the key technologies used and illustrates usability aspects of the system’s individual components.

Image registration for triggered and non-triggered DTI of the human kidney: Reduced variability of diffusion parameter estimation

BACKGROUND: To investigate if non-rigid image-registration reduces motion artifacts in triggered and non-triggered diffusion tensor imaging (DTI) of native kidneys. A secondary aim was to determine, if improvements through registration allow for omitting respiratory-triggering. METHODS: Twenty volunteers underwent coronal DTI of the kidneys with nine b-values (10-700 s/mm2 ) at 3 Tesla. Image-registration was performed using a multimodal nonrigid registration algorithm. Data processing yielded the apparent diffusion coefficient (ADC), the contribution of perfusion (FP ), and the fractional anisotropy (FA). For comparison of the data stability, the root mean square error (RMSE) of the fitting and the standard deviations within the regions of interest (SDROI ) were evaluated. RESULTS: RMSEs decreased significantly after registration for triggered and also for non-triggered scans (P < 0.05). SDROI for ADC, FA, and FP were significantly lower after registration in both medulla and cortex of triggered scans (P < 0.01). Similarly the SDROI of FA and FP decreased significantly in non-triggered scans after registration (P < 0.05). RMSEs were significantly lower in triggered than in non-triggered scans, both with and without registration (P < 0.05). CONCLUSION: Respiratory motion correction by registration of individual echo-planar images leads to clearly reduced signal variations in renal DTI for both triggered and particularly non-triggered scans. Secondarily, the results suggest that respiratory-triggering still seems advantageous.J. Magn. Reson. Imaging 2014. (c) 2014 Wiley Periodicals, Inc.