In the medical field, the detection and diagnosis of diseases continue to improve. Developments in diagnostic techniques have helped to improve treatment in the early stages and avoid many risks to patients. One relatively new diagnostic technique is optical coherence tomography (OCT), which is used in many medical applications to perform internal microstructure imaging of the human body at high resolution (typically 10 micro metre), at high speed and in real time. OCT is non-invasive and can be used as a contact or non-contact technique to obtain an image. In medicine, there are many applications that involve OCT, such as in ophthalmology, gastroenterology, cardiology and oncology. This work demonstrates the design, development and implementation of a high resolution swept laser OCT system for the imaging and diagnosis of tissues in laboratory and clinical experiments. It reports an investigation to measure the thickness of the peritoneal membrane and the use of optical imaging contrast agents such as gold nanorods. There is also an account of the design of an endoscope-catheter fast scanning OCT system for biomedical studies of the gastrointestinal tract and gynaecological areas. These results were achieved by using a swept tuneable laser source with a very high tuning speed of 16 kHz over a wide range of wavelengths: 1260 nm to 1390 nm. The laser sweeps across 110 nm at a 16 kHz repetition rate. The real axial line speed is limited by the source that is used in the OCT system. The axial resolution of the system is 7 µm and its transverse resolution is 15 µm. The bandwidth of the source is up to DeltaGamma = 110 nm, centred at Gamma0 = 1325 nm, and the coherent length is 7 µm. On the sample arm of the interferometer, the swept laser OCT technique is combined with an optical probe and endoscope in order to develop a novel diagnostic imaging device to visualize tissue in vivo for animal and human experimental trials.

From an engineering background, it is often believed that the human anatomy has already been fully described. Radiology has greatly contributed to understand the inside of the human body without surgical intervention. Despite great advances in clinical CT scanning, image quality is still related to a limited amount X-ray exposure for the patient safety. This limitation prevents fine anatomical structures to be visible and, more importantly, to be detected. Where such modality is of great advantage for screening patients, extracting parameters like surface area and volume implies the bone structure to be large enough in relation to the scan resolution. The mastoid, located in the temporal bone, houses an air cell system whose cells have a variation in size that can go far below current conventional clinical CT scanner resolution. Therefore, the mastoid air cell system is only partially represented on a CT scan. Any statistical analysis will be biased towards air cells of smaller size. To allow a complete representation of the mastoid air cell system, a micro-CT scanner is more adequate. Micro-CT scanning uses approximately the same amount of X-rays but for a much longer exposure time compared to what is normally allowed for patients. Human temporal bone specimens are therefore necessary when using such scanning method. Where the conventional clinical CT scanner lacks level of minutes details, micro-CT scanning provides an overwhelming amount of fine details. Prior to any image analysis of medical data, visualization of the data is often needed to learn how to extract the structures of interest for further processing. Visualization of micro-CT scans is of no exception. Due to the high resolution nature of the data, visualization of such data not only requires modern and powerful computers, but also necessitates a tremendous amount of time to adjust the hiding of irrelevant structures, to find the correct orientation, while emphasising the structure of interest. Once the quality of the data has been assessed, and a strategy for the image processing has been decided, the image processing can start, to in turn extract metrics such as the surface area or volume and draw statistics from it. The temporal bone being one of the most complex in the human body, visualization of micro-CT scanning of this bone awakens the curiosity of the experimenter, especially with the correct visualization settings. This thesis first presents a statistical analysis determining the surface area to volume ratio of the mastoid air cell system of human temporal bone, from micro-CT scanning using methods previously applied for conventional clinical CT scannings. The study compared current resul s with previous studies, with successive downsampling the data down to a resolution found in conventional clinical CT scanning. The results from the statistical analysis showed that all the small mastoid air cells, that cannot be detected in conventional clinical CT scans, do heavily contribute to the estimation of the surface area, and in consequence to the estimation of the surface area to volume ratio by a factor of about 2.6. Such a result further strengthens the idea of the mastoid to play an active role in pressure regulation and gas exchange. Discovery of micro-channels through specific use of a non-traditional transfer function was then reported, where a qualitative and a quantitative preanalysis was performed are described. To gain more knowledge about these micro-channels, a local structure tensor analysis was applied where structures are described in terms of planar, tubular, or isotropic structures. The results from this structural tensor analysis, also reported in this thesis, suggest these micro-channels to potentially be part of a more complex framework, which hypothetically would provide a separate blood supply for the mucosa lining the mastoid air cell system.

Non disponible

Two brain segmentation approaches based on Hidden Markov Models are proposed. The first approach aims to segment normal brain 3D multi-channel MR images into three tissues WM, GM, and CSF. Linear Discriminant Analysis, LDA, is applied to separate voxels belonging to different tissues as well as to reduce their features vector size. The second approach aims to detect MS lesions in Brain 3D multi-channel MR images and to label WM, GM, and CSF tissues. Preprocessing is applied in both approaches to reduce the noise level and to address sudden intensity and global intensity correction. The proposed techniques are tested using 3D images from Montereal BrainWeb data set. In the first approach, the results were numerically assessed and compared to results reported using techniques based on single channel data and applied to the same data sets. The results obtained using the multi channel HMM-based algorithm were better than the results reported for single channel data in terms of an objective measure of overlap, Dice coefficient, compared to other methods. In the second approach, the segmentation accuracy is measured using Dice coefficient and total lesions load percentage

The propagation mechanism of signals for multiple input multiple output (MIMO) channels can be explained via a random matrix. Random matrix theory is a very powerful tool to understand behaviour of such channels and analyse their performance measure of MIMO systems. In this work we study: The asymptotic eigenvalue distribution and the mutual information of a multiuser (MU) multiple-input multiple output (MIMO) channel with a certain fraction of users experiencing line-of-sight. It shows that the AED of the channel matrix decomposes into two separate bulks for practically relevant parameter choices and differs very much from the common assumption of independent identically distributed (iid) entries which induces the quarter circle law. This happens even without antenna correlation at either side of the channel. In order to tackle this problem the paper makes use of recent developments in free probability theory which allow to deal with complex-valued eigenvalue distributions of non-Hermitian matrices. Moreover to understand behaviour of MIMO channels we derived asymptotic complex-valued eigenvalue distributions of practically relevant channels models by means of their respective square equivalent and singular equivalent of channel matrices. Finally we derived an explicit mutual information formula which allows us calculate the mutual information (in general) analytically in high signal-to noiseratio (SNR) regime for numerous practical important scenarios. Furthermore the numerical result shows that, high-SNR approximation draws reliable portrait even for quite moderate SNR level.

Reverberation is defined as equally-spaced, bright linear echoes resulting from reflection from specular-type interfaces. They are provoked by the acoustic Impedance change between the tissue and transducer front surface. B. Angelsen developed a mathematical approach to correct this ultrasound artifact by coupling the ultrasound transducer with an ideal electrical load in order to obtain zero reflection coefficients on the transducer from face [1]. However, when analyzing Impedance spectroscopy this approach cannot be achieved by using passive electronics in most simulated cases. Active impedance may be seen as a trend to improve image quality. This research therefore implements and simulates methods for active impedance synthesis [5] and applies Particle Swarm Optimization to the proposed electronics. Finally, the simulations are compared through a LAB experiment.

The commonly used simulation method Field II, which is based on the spatial impulse response approach, has excellent accuracy in linear domain. However the computational time can be up to many days for one simulation. One of the solutions to this problem is a convolution-based methodology called COLE. It is much faster than Field II and has very good approximation. It generates the data by reducing multi-dimensional convolution model to multiple single-dimensional convolutions. This thesis is about implementing COLE on the FieldSim 3 platform and using it for blood flow imaging. This platform is written in MATLAB with object-oriented programming and it is now under development at department of circulation and medical imaging. Both Field II and real scanner have been used to compare with COLE. The simulated phantom for both simulators was a straight tube with scatterers moving inside, whereas a string phantom was used to get the data from the scanner. The computational time of COLE with 2D Doppler mode scan in FieldSim 3 achieved 85 times faster than Field II. The plotted PW Doppler spectra and the 2D power spectra showed that the velocity resolutions of both simulators were at the same level. COLE had higher noise floor than Field II and scanner in Doppler mode scan. COLE had relatively high sampling frequency requirement compared with Field II. If the sampling frequency was not high enough, COLE would produce side lobes in the PW Doppler spectra.

X-ray computed tomography (CT) is a medical imaging framework. It takes measured projections of X-rays through two-dimensional cross-sections of an object from multiple angles and incorporates algorithms in building a sequence of two-dimensional reconstructions of the interior structure. This thesis comprises a review of the different types of algebraic algorithms used in X-ray CT. Using simulated test data, I evaluate the viability of algorithmic alternatives that could potentially reduce overexposure to radiation, as this is seen as a major health concern and the limiting factor in the advancement of CT [36, 34]. Most of the current evaluations in the literature [31, 39, 11] deal with low-resolution reconstructions and the results are impressive, however, modern CT applications demand very high-resolution imaging. Consequently, I selected ve of the fundamental algebraic reconstruction algorithms (ART, SART, Cimmino's Method, CAV, DROP) for extensive testing and the results are reported in this thesis. The quantitative numerical results obtained in this study, con rm the qualitative suggestion that algebraic techniques are not yet adequate for practical use. However, as algebraic techniques can actually produce an image from corrupt and/or missing data, I conclude that further re nement of algebraic techniques may ultimately lead to a breakthrough in CT. / UOIT

Medical image analysis is a ubiquitous and essential part of modem health care. A crucial first step to this is segmentation, which is often complicated by many factors including subject diversity, pathology, noise corruption, and poor image resolution. Traditionally, manual tracing by experts was done. While considered accurate, this process is time consuming and tedious, especially when performed slice-by-slice on three-dimensional (3D) images over large datasets or on two-dimensional (2D) but topologically complicated images such as a retinography. On the other hand, fully-automated methods are typically faster, but work best with data-dependent, carefully tuned parameters and still require user validation and refinement. This thesis contributes to the field of medical image segmentation by proposing a highly-automated, interactive approach that effectively merges user knowledge and efficient computing. To this end, our work focuses on graph-based methods and offer globally optimal solutions. First, we present a novel method for 3D segmentation based on a 3D Livewire approach. This approach is an extension of the 2D Livewire framework, and this method is capable of handling objects with large protrusions, concavities, branching, and complex arbitrary topologies. Second, we propose a method for efficiently segmenting 2D vascular networks, called ‘Live-Vessel’. Live-Vessel simultaneously extracts vessel centrelines and boundary points, and globally optimizes over both spatial variables and vessel radius. Both of our proposed methods are validated on synthetic data, real medical data, and are shown to be highly reproducible, accurate, and efficient. Also, they were shown to be resilient to high amounts of noise and insensitive to internal parameterization.

Medical imaging plays a vital role in the medical field because it is widely used in diseases diagnosis and treatment of patients. There are different modalities of medical imaging such as ultrasounds, x-rays, Computed Tomography (CT), Magnetic Resonance (MR), and Positron Emission Tomography (PET). Most of these modalities usually suffer from noise and other sampling artifacts. The diagnosis process in these modalities depends mainly on the interpretation of the radiologists. Consequently, the diagnosis is subjective as it is based on the radiologist experience. Medical image segmentation is an important process in the field of image processing. It has a significant role in many applications such as diagnosis, therapy planning, and advanced surgeries. There are many segmentation techniques to be applied on medical images. However, most of these techniques are still depending on the experts, especially for initializing the segmentation process. The artifacts of images can affect the segmentation output. In this thesis, we propose a new approach for automatic prostate segmentation of Trans-Rectal UltraSound (TRUS) images by dealing with the speckle not as noise but as informative signals. The new approach is an automation of the conventional region growing technique. The proposed approach overcomes the requirement of manually selecting a seed point for initializing the segmentation process. In addition, the proposed approach depends on unique features such as the intensity and the spatial Euclidean distance to overcome the effect of the speckle noise of the images. The experimental results of the proposed approach show that it is fast and accurate. Moreover, it performs well on the ultrasound images, which has the common problem of the speckle noise.