The segmentation of medical images is a difficult task due to the inhomogeneous intensity variations that occurs during digital image acquisition, the complicated shape of the object, and the medical expert’s lack of semantic knowledge. Automated segmentation algorithms work well for some medical images, but no algorithm has been general enough to work for all medical images. In practice, most of the time the segmentation results are corrected by the experts before the actual use. In this work, we are motivated to determine how to make use of manually segmented data in automatic segmentation. The key idea is to transfer the ground truth segmentation from the database of train images to a given test image. The ground truth segmentation of MR images is done by experts. The process includes a hierarchical image decomposition approach that performs the shape matching of test images at several levels, starting with the image as a whole (i.e. level 0) and then going through a pyramid decomposition (i.e. level 1, level 2, etc.) with the database of the train images and the given test image. The goal of pyramid decomposition is to find the section of the training image that best matches a section of the test image of a different level. After that, a re-composition approach is taken to place the best matched sections of the training image to the original test image space. Finally, the ground truth segmentation is transferred from the best training images to their corresponding location in the test image. We have tested our method on a hip joint MR image database and the experiment shows successful results on level 0, level 1 and level 2 re-compositions. Results improve with deeper level decompositions, which supports our hypotheses.

The quantitative assessment of lymph node size plays an important role in treatment of diseases like cancer. In current clinical practice, lymph nodes are analyzed manually based on very rough measures of long and/or short axis length, which is error prone. In this paper we present a graph-based lymph node segmentation method to enable the computer-aided three-dimensional (3D) assessment of lymph node size. Our method has been validated on 111 cases of enlarged lymph nodes imaged with X-ray computed tomography (CT). For unsigned surface positioning error, Hausdorff distance and Dice coefficient, the mean was around 0.5 mm, under 3.26 mm and above 0.77 respectively. On average, 5.3 seconds were required by our algorithm for the segmentation of a lymph node.

Medical imaging modalities provide effective information for anatomic or metabolic activity of tissues and organs in the body. Therefore, medical imaging technology is a critical component in diagnosis and treatment of various illnesses. Medical image segmentation plays an important role in converting medical images into anatomically, functionally or surgically identifiable structures, and is used in various applications. In this study, some of the major medical image segmentation methods are examined and applied to 2D CT images of upper torso for segmentation of heart, lungs, bones, and muscle and fat tissues. The implemented medical image segmentation methods are thresholding, region growing, watershed transformation, deformable models and a hybrid method / watershed transformation and region merging. Moreover, a comparative analysis is performed among these methods to obtain the most efficient segmentation method for each tissue and organ in torso. Some improvements are proposed for increasing accuracy of some image segmentation methods.

Medical image segmentation is a challenging problem, and is studied widely. In this thesis, the main goal is to develop automatic segmentation techniques of human mimic muscles and to compare them with ground truth data in order to determine the method that provides best segmentation results. The segmentation methods are based on Bayesian with Markov Random Field (MRF) and Level Set (Active Contour) models. Proposed segmentation methods are multi step processes including preprocess, main muscle segmentation step and post process, and are applied on three types of data: Magnetic Resonance Imaging (MRI) data, Computerized Tomography (CT) data and unified data, in which case, information coming from both modalities are utilized. The methods are applied both in three dimensions (3D) and two dimensions (2D) data cases. A simulation data and two patient data are utilized for tests. The patient data results are compared statistically with ground truth data which was labeled by an expert radiologist.

Στα πλαίσια της παρούσης εργασίας πραγματοποιήθηκε μελέτη της μεθόδου Συνόλων Επιπέδου για την κατάτμηση καροτίδων από τρισδίαστατες εικόνες. Ειδικότερα πραγματοποιήθηκε μελέτη των παθολογιών που συνδέονται με αυτές προκειμένου να καταστούν εμφανή τα κίνητρα της παρούσης εργασίας, όσον αφορά στη συμβολή της στην κλινική σημασία και ιατρική πρακτική. Κατ’αυτόν τον τρόπο, αφού παρουσιάστηκε η ανατομία των καροτίδων και οι δυσκολίες που ενέχει το εγχείρημα της κατάτμησής τους καθώς και μια ανασκόπηση των μεθόδων Συνόλων Επιπέδου (Level-Sets) για κατάτμηση ιατρικής εικόνας και δη καροτίδων, παρουσιάστηκε το γενικό μοντέλο και ο μαθηματικός φορμαλισμός της μεθόδου που χρησιμοποιήθηκε. Εν συνεχεία παρουσιάστηκαν τα τρισδιάστατα δεδομένα και η διαχείρησή τους, οι προγραμματιστικές διαπαφές και υποδομές με τις οποίες υλοποιήθηκαν δύο παραλλαγές της μεθόδου. Επίσης παρουσιάζονται τα αποτελέσματα της μεθόδου οπτικοποιημένα και τέλος συγκρίνονται με αντίστοιχα αποτελέσματα ενός ειδικού ακτινολόγου στη βάση κάποιων κατάλληλων μετρικών. Τέλος παρουσιάζονται τα συμπεράσματα που προέκυψαν καθώς και κάποιες ιδέες για μελλοντική δουλειάπου μπορεί να γίνει στη βάση αυτής που έγινε στα πλαίσια της εν λόγω μεταπτυχιακής διατριβής. / The present thesis outlines the methods we have developed for segmenting both normal and pathological carotid images, acquired with the Computed Tomography (CT) protocol. The layout of the thesis is the following: Chapter 2 analyses the methodological background of the current study. At first, section 2.1 provides an overview to the anatomy of carotids. Section 2.2 reviews the literature of segmentation methods based on level sets for medical images and at last reviews the level set methods developed for segmenting carotids. In addition, section 2.3 presents the conceptual model deployed in the current study, following with the analysis of the particular class we used. Next, section 2.4 treats of the level set method, presenting its basic derivation and furthermore discriminating between the two algorithms used according to their speed function. Chapter 3 refers to the materials and methods. It begins in section 3.1 with a description of the data provided for the experimental demonstration, and the programming interface by deployment of which the experimental procedure took place. Later on, in section 3.2 the implementation of the deployed methods in the programming interface used is presented with an analysis of their components. At last, all intermediate outputs and the final results of each method are illustrated. Chapter 4 presents the evaluation of the results of each method by comparison with a corresponding manual segmentation result on the basis of appropriate metrics. At last, refers to the conclusions occurred and to future work that can be carried out based on the current Msc thesis. In Appendix A some subsidiary methods, for the sake of a coherent flow are stated and analyzed independently.

During the last few years, segmentation architectures based on deep learning achieved promising results. On the other hand, attention networks have been invented years back and used in different tasks but rarely used in medical applications. This thesis investigated four main attention mechanisms; Squeeze and Excitation, Dual Attention Network, Pyramid Attention Network, and Attention UNet to be used in medical image segmentation. Also, different hybrid architectures proposed by the author were tested. Methods were tested on a kidney tumor dataset and against UNet architecture as a baseline. One version of Squeeze and Excitation attention outperformed the baseline. Original Dual Attention Network and Pyramid Attention Network showed very poor performance, especially for the tumor class. Attention UNet architecture achieved close results to the baseline but not better. Two more hybrid architectures achieved better results than the baseline. The first is a modified version of Squeeze and Excitation attention. The second is a combination between Dual Attention Networks and UNet architecture. Proposed architectures outperformed the baseline by up to 3% in tumor Dice coefficient. The thesis also shows the difference between 2D architectures and their 3D counterparts. 3D architectures achieved more than 10% higher tumor Dice coefficient than 2D architectures.

Accurate medical image segmentation is crucial for safely and effectively administering radiation therapy in cancer treatment. State of the art methods for automatic segmentation of 3D images are currently based on the U-net machine learning architecture. The current U-net models are large, often containing millions of parameters. However, the size of these machine learning models can be decreased by removing parts of the models, in what is called pruning. One algorithm, called simultaneous training and pruning (STAMP) has shown capable of reducing the model sizes upwards of 80% while keeping similar or higher levels of performance for medical image segmentation tasks.  This thesis investigates the impact of using the STAMP algorithm to reduce model size and inference time for medical image segmentation on 3D images, using one MRI and two CT datasets. Surprisingly, we show that pruning convolutional filters randomly achieves performance comparable, if not better than STAMP, provided that the filters are always removed from the largest parts of the U-net.  Inspired by these results, a modified "Flat U-net" is proposed, where an equal number of convolutional filters are used in all parts of the U-net, similar to what was achieved after pruning with our simplified pruning algorithm. The modified U-net achieves similar levels of test dice score as both a regular U-net and the STAMP pruning algorithm, on multiple datasets while avoiding pruning altogether. In addition to this the proposed modification reduces the model size by more than a factor of 12, and the number of computations by around 35%, compared to a normal U-net with the same number of input-layer convolutional filters.

In recent years semantic segmentation models utilizing Convolutional Neural Networks (CNN) have seen significant success for multiple different segmentation problems. Models such as U-Net have produced promising results within the medical field for both regular 2D and volumetric imaging, rivalling some of the best classical segmentation methods. In this thesis we examined the possibility of using a convolutional neural network-based model to perform segmentation of discrete bone fragments in CT-volumes with segmentation-hints provided by a user. We additionally examined different classical segmentation methods used in a post-processing refinement stage and their effect on the segmentation quality. We compared the performance of our model to similar approaches and provided insight into how the interactive aspect of the model affected the quality of the result. We found that the combined approach of interactive segmentation and deep learning produced results on par with some of the best methods presented, provided there were adequate amount of annotated training data. We additionally found that the number of segmentation hints provided to the model by the user significantly affected the quality of the result, with convergence of the result around 8 provided hints.

The determination of the center slice, defined as a slice through the lateral ventricles in the axial plane in a volume of MR images is important to the segmentation of the image into its anatomical parts. The center or ventricle slice in a set of MR images is recognized by the shape of the ventricles in the axial plane as depicted by the cerebro-spinal fluids in the image. Currently, no technique exists to detect this slice and the purpose of this thesis is to find a slice through the lateral ventricles in the axial plane from a volume of MRI brain scan slices. There are several methodologies which will be discussed in the thesis, the Hough Transform and Object Matching using deformable templates being the primary ones. It is shown, in the test cases used, that these algorithms used together provided results that had almost 80 percent accuracy. However, a simple method to spatially calculate the center slice is also competitive in accuracy.

Medical imaging is an important tool for diagnosis and treatment planning today. However as the demand for efficiency increases at the same time as the data volumes grow immensely, the need for computer assisted analysis, such as image segmentation, to help and guide the practitioner increases. Medical image segmentation could be used for various different tasks, the localization and delineation of pathologies such as cancer tumors is just one example. Numerous problems with noise and image artifacts in the generated images make the segmentation a difficult task, and the developer is forced to choose between speed and performance. In clinical practise, however, this is impossible as both speed and performance are crucial. One solution to this problem might be to involve the user more in the segmentation, using interactivite algorithms where the user might influence the segmentation for an improved result. This thesis has concentrated on finding a fast and interactive segmentation method for liver tumor segmentation. Various different methods were explored, and a few were chosen for implementation and further development. Two methods appeared to be the most promising, Bayesian Region Growing (BRG) and Level Set. An interactive Level Set algorithm emerged as the best alternative for the interactivity of the algorithm, and could be used in combination with both BRG and Level Set. A new data term based on a probability model instead of image edges was also explored for the Level Set-method, and proved to be more promising than the original one. The probability based Level Set and the BRG method both provided good quality results, but the fastest of the two was the BRG-method, which could segment a tumor present in 25 CT image slices in less than 10 seconds when implemented in Matlab and mex-C++ code on an ACPI x64-based PC with two 2.4 GHz Intel(R) Core(TM) 2CPU and 8 GB RAM memory. The interactive Level Set could be succesfully used as an interactive addition to the automatic method, but its usefulness was somewhat reduced by its slow processing time ( 1.5 s/slice) and the relative complexity of the needed user interactions.