<p>KTH, Division of Neuronic Engineering, have a finite element model of the head. However, this model does not contain detailed modeling of the brain. This thesis project consists of finding a method to extract brain tissues from T1-weighted MR images of the head. The method should be automatic to be suitable for patient individual modeling.</p><p>A summary of the most common segmentation methods is presented and one of the methods is implemented. The implemented method is based on the assumption that the probability density function (pdf) of an MR image can be described by parametric models. The intensity distribution of each tissue class is modeled as a Gaussian distribution. Thus, the total pdf is a sum of Gaussians. However, the voxel values are also influenced by intensity inhomogeneities, which affect the pdf. The implemented method is based on the expectation-maximization algorithm and it corrects for intensity inhomogeneities. The result from the algorithm is a classification of the voxels. The brain is extracted from the classified voxels using morphological operations.</p>

This research presents a procedure for interactive segmentation and automatic tracking of moving objects in a video sequence. The user outlines the region of interest (ROI) in the initial frame; the procedure builds a refined mask of the dominant object within the ROI. The refined mask is used to model a spline template of the object to be tracked. The tracking algorithm then employs a motion model to track the template through a sequence of frames and gathers the 3D affine motion parameters of the object from each frame. The extracted template is compared with a previously stored library of 3D shapes to determine the closest 3D object. If the extracted template is completely new, it is used to model a new 3D object which is added to the library. To recreate the motion, the motion parameters are applied to the 3D object in a virtual environment. The procedure described here can be applied to industrial problems such as traffic management and material flow congestion analysis.

KTH, Division of Neuronic Engineering, have a finite element model of the head. However, this model does not contain detailed modeling of the brain. This thesis project consists of finding a method to extract brain tissues from T1-weighted MR images of the head. The method should be automatic to be suitable for patient individual modeling. A summary of the most common segmentation methods is presented and one of the methods is implemented. The implemented method is based on the assumption that the probability density function (pdf) of an MR image can be described by parametric models. The intensity distribution of each tissue class is modeled as a Gaussian distribution. Thus, the total pdf is a sum of Gaussians. However, the voxel values are also influenced by intensity inhomogeneities, which affect the pdf. The implemented method is based on the expectation-maximization algorithm and it corrects for intensity inhomogeneities. The result from the algorithm is a classification of the voxels. The brain is extracted from the classified voxels using morphological operations.

This research presents a procedure for interactive segmentation and automatic tracking of moving objects in a video sequence. The user outlines the region of interest (ROI) in the initial frame; the procedure builds a refined mask of the dominant object within the ROI. The refined mask is used to model a spline template of the object to be tracked. The tracking algorithm then employs a motion model to track the template through a sequence of frames and gathers the 3D affine motion parameters of the object from each frame. The extracted template is compared with a previously stored library of 3D shapes to determine the closest 3D object. If the extracted template is completely new, it is used to model a new 3D object which is added to the library. To recreate the motion, the motion parameters are applied to the 3D object in a virtual environment. The procedure described here can be applied to industrial problems such as traffic management and material flow congestion analysis.

Anatomic changes occur throughout head and neck radiotherapy, and a new treatment plan is often required to mitigate the resulting changes in delivered dose to key structures. This process is known as adaptive radiotherapy (ART), and can be labour-intensive. The aim of this thesis is to optimise ART, addressing some of the technical and clinical challenges facing its routine clinical implementation. Optimising the frequency and timing of adaptive replanning is important, and it has been shown here that intensity modulated radiotherapy (IMRT) and volumetric modulated arc therapy (VMAT) are equally robust to weight loss during head and neck radiotherapy. Plan adaptation strategies that have previously been developed for IMRT are therefore applicable to VMAT.Contour propagation is an important component of ART, and it is essential to ensure that propagated contours are accurate. A method for assessing the suitability of a metric for measuring automatic segmentation accuracy has been developed and applied to the head and neck. For the parotids and larynx, metrics based on surface agreement were better than the commonly used Dice similarity coefficient. By establishing a consensus on which metrics should be used to assess segmentation accuracy, comparison of different algorithms is more objective and should lead to more accurate automatic segmentation. A novel method of assessing contour propagation accuracy on a patient-specific basis has also been developed. This was demonstrated on a cohort of head and neck patients and shows potential as a tool for identifying propagated contours that are subject to a high degree of uncertainty. This is a novel tool that will increase the efficiency of automatic segmentation and, therefore, ART.Optimum ART requires consideration of different radiotherapy-related toxicities, and image-based data mining is a powerful technique for spatially localising dose-response relationships. Correction for multiple comparisons through permutation testing is essential, but has so far only been applied to categorical data. A novel method has been developed for performing permutation testing and image-based data mining with a continuously variable clinical endpoint. Application to trismus for head and neck radiotherapy identified a region with a dose-response relationship in the ipsilateral masseter. Sparing this structure during radiotherapy should reduce the severity of radiation-induced trismus. ART mitigates the dosimetric effects of anatomic changes, and this thesis has addressed technical and clinical challenges that have so far limited its clinical implementation. Detailed knowledge of dose-response relationships will enable selection of patients for ART based on potential clinical benefit, and accurate contour propagation will make ART more efficient, facilitating its routine implementation.

abstract: Recent new experiments showed that wide-field imaging at millimeter scale is capable of recording hundreds of neurons in behaving mice brain. Monitoring hundreds of individual neurons at a high frame rate provides a promising tool for discovering spatiotemporal features of large neural networks. However, processing the massive data sets is impossible without automated procedures. Thus, this thesis aims at developing a new tool to automatically segment and track individual neuron cells. The new method used in this study employs two major ideas including feature extraction based on power spectral density of single neuron temporal activity and clustering tree to separate overlapping cells. To address issues associated with high-resolution imaging of a large recording area, focused areas and out-of-focus areas were analyzed separately. A static segmentation with a fixed PSD thresholding method is applied to within focus visual field. A dynamic segmentation by comparing maximum PSD with surrounding pixels is applied to out-of-focus area. Both approaches helped remove irrelevant pixels in the background. After detection of potential single cells, some of which appeared in groups due to overlapping cells in the image, a hierarchical clustering algorithm is applied to separate them. The hierarchical clustering uses correlation coefficient as a distance measurement to group similar pixels into single cells. As such, overlapping cells can be separated. We tested the entire algorithm using two real recordings with the respective truth carefully determined by manual inspections. The results show high accuracy on tested datasets while false positive error is controlled within an acceptable range. Furthermore, results indicate robustness of the algorithm when applied to different image sequences. / Dissertation/Thesis / Masters Thesis Electrical Engineering 2016

A perda volumétrica cerebral ou atrofia é um importante índice de destruição tecidual e pode ser usada para apoio ao diagnóstico e para quantificar a progressão de diversas doenças com componente degenerativo, como a esclerose múltipla (EM), por exemplo. Nesta doença ocorre perda tecidual regional, com reflexo no volume cerebral total. Assim, a presença e a progressão da atrofia podem ser usadas como um indexador da progressão da doença. A quantificação do volume cerebral é um procedimento relativamente simples, porém, quando feito manualmente é extremamente trabalhoso, consome grande tempo de trabalho e está sujeito a uma variação muito grande inter e intra-observador. Portanto, para a solução destes problemas há necessidade de um processo automatizado de segmentação do volume encefálico. Porém, o algoritmo computacional a ser utilizado deve ser preciso o suficiente para detectar pequenas diferenças e robusto para permitir medidas reprodutíveis a serem utilizadas em acompanhamentos evolutivos. Neste trabalho foi desenvolvido um algoritmo computacional baseado em Imagens de Ressonância Magnética para medir atrofia cerebral em controles saudáveis e em pacientes com EM, sendo que para a classificação dos tecidos foi utilizada a teoria da entropia generalizada de Tsallis. Foram utilizadas para análise exames de ressonância magnética de 43 pacientes e 10 controles saudáveis pareados quanto ao sexo e idade para validação do algoritmo. Os valores encontrados para o índice entrópico q foram: para o líquido cerebrorraquidiano 0,2; para a substância branca 0,1 e para a substância cinzenta 1,5. Nos resultados da extração do tecido não cerebral, foi possível constatar, visualmente, uma boa segmentação, fato este que foi confirmado através dos valores de volume intracraniano total. Estes valores mostraram-se com variações insignificantes (p>=0,05) ao longo do tempo. Para a classificação dos tecidos encontramos erros de falsos negativos e de falsos positivos, respectivamente, para o líquido cerebrorraquidiano de 15% e 11%, para a substância branca 8% e 14%, e substância cinzenta de 8% e 12%. Com a utilização deste algoritmo foi possível detectar um perda anual para os pacientes de 0,98% o que está de acordo com a literatura. Desta forma, podemos concluir que a entropia de Tsallis acrescenta vantagens ao processo de segmentação de classes de tecido, o que não havia sido demonstrado anteriormente. / The loss of brain volume or atrophy is an important index of tissue destruction and it can be used to diagnosis and to quantify the progression of neurodegenerative diseases, such as multiple sclerosis. In this disease, the regional tissue loss occurs which reflects in the whole brain volume. Similarly, the presence and the progression of the atrophy can be used as an index of the disease progression. The objective of this work was to determine a statistical segmentation parameter for each single class of brain tissue using generalized Tsallis entropy. However, the computer algorithm used should be accurate and robust enough to detect small differences and allow reproducible measurements in following evaluations. In this work we tested a new method for tissue segmentation based on pixel intensity threshold. We compared the performance of this method using different q parameter range. We could find a different optimal q parameter for white matter, gray matter, and cerebrospinal fluid. The results support the conclusion that the differences in structural correlations and scale invariant similarities present in each single tissue class can be accessed by the generalized Tsallis entropy, obtaining the intensity limits for these tissue class separations. Were used for analysis of magnetic resonance imaging examinations of 43 patients and 10 healthy controls matched on the sex and age for validation of the algorithm. The values found for the entropic index q were: for the cerebrospinal fluid 0.2; for the white matter 0.1 and for gray matter 1.5. The results of the extraction of the tissue not brain can be seen, visually, a good target, which was confirmed by the values of total intracranial volume. These figures showed itself with variations insignificant (p >= 0.05) over time. For classification of the tissues find errors of false negatives and false positives, respectively, for cerebrospinal fluid of 15% and 11% for white matter 8% and 14%, and gray matter of 8% and 12%. With the use of this algorithm could detect an annual loss for the patients of 0.98% which is in line with the literature. Thus, we can conclude that the entropy of Tsallis adds advantages to the process of target classes of tissue, which had not been demonstrated previously.

This thesis tries to answer if a semi-automatic tool can speed up the process of segmenting tumors to find the area of a slice in the tumor or the volume of the entire tumor. A few different 2D semi-automatic tools were considered. The final choice was to implement live-wire. The implemented live-wire was evaluated and improved upon with hands-on testing from developers. Two methods were found for extending live-wire to 3D bodies. The first method was to interpolate the seed points and create new contours using the new seed points. The second method was to let the user segment contours in two orthogonal projections. The intersections between those contours and planes in the third orthogonal projection were then used to create automatic contours in this third projection. Both tools were implemented and evaluated. The evaluation compared the two tools to manual segmentation on two cases posing different difficulties. Time-on-task and accuracy were measured during the evaluation. The evaluation revealed that the semi-automatic tools could indeed save the user time while maintaining acceptable (80%) accuracy. The significance of all results were analyzed using two-tailed t-tests.

This thesis introduces a tool for analysis of audiovisual records. The tool uses the audio and closed captions supplied by the user to prepare text annotation. The annotation contains a transcript of the show which is based on the closed captions. In addition, speaker diarization is performed to mark who spoke when. The diarization is performed by a third party library. The library is evaluated on data from DIALOG corpus. The inner workings of the library are described. To assign the right portions of the text to the right section of the record Kaldi, a speech recognition toolkit, is used. Furthermore the thesis contains an overview describing how closed captions are created; overview of speech corpora creation; and a brief review of literature on record analysis. 1

This thesis investigates if support vector machine classification is a suitable approach when performing automatic segmentation of knee cartilage using quantitative magnetic resonance imaging data. The data sets used are part of a clinical project that investigates if patients that have suffered recent knee damage will develop cartilage damage. Therefore the thesis also investigates if the segmentation results can be used to predict the clinical outcome of the patients. Two methods that perform the segmentation using support vector machine classification are implemented and evaluated. The evaluation indicates that it is a good approach for the task, but the implemented methods needs to be further improved and tested on more data sets before clinical use. It was not possible to relate the cartilage properties to clinical outcome using the segmentation results. However, the investigation demonstrated good promise of how the segmentation results, if they are improved, can be used in combination with quantitative magnetic resonance imaging data to analyze how the cartilage properties change over time or vary between knees.