|
|
(Plocková veronika) Advanced analysis of specific thrombi features in multiphase CT data
Plocková, Veronika ; Nemčeková, Petra (referee) ; Jakubíček, Roman (advisor)
This diploma thesis focuses on advanced analysis of thrombus in multiphase CT data. The first part presents a literature review on ischemic stroke, its treatment, and the impact of thrombus structure on treatment success. Additionally, the literature review includes brief anatomy of the brain and methods for extracting radiomic features and possible methods for thrombus analysis from imaging data. In the last part of the theoretical work, the extracted features used in the practical part of the work and methods of their calculation are summarized. A brief introduction to statistical methods used in the practical part of the work follows. The practical part focuses on extracting statistical and textural features from CT data. Experiments deal with the analysis of extracted features, discriminatory properties of thrombus and background, and proposing an approach to thrombus segmentation from CT data. The results of the experiments are discussed.
|
|
|
Using unlabeled data for retinal segmentation
Shemshur, Andrii ; Jakubíček, Roman (referee) ; Vičar, Tomáš (advisor)
Tato bakalářská práce se zabývá vývojem a hodnocením pokročilých metod pro segmentaci lékařských snímků v kontextu omezených trénovacích dat. Studie zkoumá techniky učení pod dohledem využívající konvoluční neuronové sítě (CNN), přenosové učení s předtrénovanými modely a strategie učení s částečným dohledem. Jako základní model byl použit model konvoluční neuronové sítě (CNN) s dohledem založený na architektuře U-Net, který dosáhl koeficientu Dice 77,6% a průniku nad sjednocením (IoU) 63,4%. Použití přenosového učení pomocí kodéru ResNet34 předtrénovaného na síti ImageNet vedlo k výraznému zlepšení výkonu s koeficientem Dice 81,9%, IoU 69,3% a přesností 96,7%. Kromě toho byly ke zvýšení výkonu modelu použity strategie učení s částečným dohledem, včetně pseudoznačení a předtrénování denoizace. Přístup pseudoznačení přinesl koeficient Dice 81,7% a IoU 69,1%, čímž prokázal účinnost využití neoznačených dat. Přístup před tréninkem denoizace prokázal robustní výkonnost a dosáhl koeficientu Dice 80,3% a IoU 67,0%, a to i v přítomnosti zašuměných a neoznačených dat. Tyto výsledky podtrhují potenciál transferového učení a poloprovozních metod pro zvýšení přesnosti segmentace při analýze lékařských snímků. Poskytují solidní základ pro budoucí výzkum v této oblasti.
|
|
|
Deep learning-based noise reduction in X-ray images
Říhová, Barbora ; Jakubíček, Roman (referee) ; Zemek, Marek (advisor)
Technologie zobrazování pomocí rentgenových paprsků je základem zkoumání vnitřní struktury velké škály objektů a výsledky mohou být právě kvůli šumu kompromitovány. Tato práce se zabývá odstraňováním šumu v rentgenových projekcích pomocí hlubokého učení, které má schopnost adaptovat se na konkrétní problém. Práce obsahuje teoretickou rešerši zaměřenou na oblasti produkce a detekce rentgenových paprsků, šumu v rentgenových snímcích a neuronových sítí. Speciální kapitola je věnována popisu vybraného řešení, které je provedeno pomocí tvorby datasetu složeného z části z modelovaných rentgenových projekcí s následně implementovaným šumem odpovídající modelu v reálných snímcích a částečně ze sérií rentgenových projekcí získaných ze zařízení Rigaku nano3DX. K implementaci byla vybrána architektura konvoluční neuronové sítě RIDNet, vzhledem k tomu, že poskytuje v oblasti redukce šumu dobré výsledky. Byly natrénovány tři modely s použitím různých částí datasetu. Nejlepší výkon byl pozorován u modelů, u kterých byla při trénování použita reálná data. Jejich účinnost je srovnatelná s tradičními metodami jako BM3D.
|
|
|
Methods of object skeletonization in biomedical images
Matejčík, Martin ; Odstrčilík, Jan (referee) ; Jakubíček, Roman (advisor)
This thesis deals with the skeletonization of binary objects in 2D images. It describes the principles, approaches and uses of skeletonization. The functioning of the Lee, Zhang, Guo-Hall, and Medial Axis Transform methods is also theoretically explained. A custom image dataset consisting of biomedical and non-biomedical images of different topological levels was created. The image dataset was subjected to skeletonization by all of the selected methods. Experiments were designed to investigate the behavior of skeletonization under the influence of boundary noise and rotation. New modified images from the original dataset were created for individual experiments. The results indicated a degraded quality of skeletons from images affected by medium and high boundary noise. The methods showed robustness to rotation only at rotation angles of 90, 180 and 270 degrees, with the best results produced by Medial axis transform method. The calculation time of individual methods was also monitored depending on the increasing number of pixels in the image. The Guo-Hall method produced interesting results, but was also the slowest.
|
|
|
Deep neural network learning methods with limited datasets
Németh, Filip ; Vičar, Tomáš (referee) ; Jakubíček, Roman (advisor)
The master thesis aims to investigate the effectiveness of deep neural networks in image processing with limited training data. As part of the work, the effects of various techniques and approaches on the learning of these networks were analyzed, including transfer learning, data augmentation, and neural style transfer method. Experimental results suggest that transfer learning using pre-trained weights from large datasets such as ImageNet is effective in improving results on limited data, achieving high F1-scores. The use of different forms of data augmentation can lead to variable results, where it provides different advantages and disadvantages that have a significant impact on the success and efficiency of the models. In general, the method using a neural style transfer network does not yield significant improvements and proved less effective for dataset with a large diversity of perspectives and geometric features.
|
|
|
Implementation of thrombi detector in multiphase CT data
Rudol, Filip ; Chmelík, Jiří (referee) ; Jakubíček, Roman (advisor)
Early diagnosis is a key factor for the successful treatment of ischemic stroke. Software for automatic thrombi detection can be a helpful tool for radiologists. The bachelor’s thesis deals with thrombi detection in multiphase CT data. The main objective of this thesis was to detect thrombi and develop a prototype of the detection software. The algorithm also allows the thrombi to be segmented. The proposed approach is fully automatic. The success of the algorithm could be objectively evaluated using the available manual annotations. The practical part was implemented in the MATLAB R2023b programming environment. In the theoretical part, a research on the issue was developed.
|
|
|
Ischemic thrombus analysis in multiphasic brain stroke CT data
Mikešová, Tereza ; Holeček, Tomáš (referee) ; Jakubíček, Roman (advisor)
This master thesis deals with analysis of ischemic thrombus in brain CT scans. In theoretical part, a review of methods, especially thrombus segmentation, is developed. Furthermore, the anatomy of cerebral arteries and acute ischemic stroke is summarized. Selected methods from the field of image processing are briefly described. The practical part results in a comparison of thrombus segmentation methods. The segmentation itself was preceded by data preprocessing, which is described in the theses, and the creation of a manual annotation database. The best implemented method was found to be the adaptive thresholding method, which achieved a Dice score of 0,4555. By combining the methods appropriately, a final Dice score of 0,5145 was achieved. Thrombus parameters were then calculated from the segmented volumes. The median intensity value was 51,55~HU, the median length was 15,16 mm, and the median volume was determined to be 65,34 mm3. Subsequent correlation analysis showed no significant relationship between the derived parameters.
|
|
|
Segmentation of ribs in thoracic CT scans
Kašík, Ondřej ; Kolář, Radim (referee) ; Jakubíček, Roman (advisor)
This thesis deals with design and implementation of an algorithm for segmentation of ribs from thoracic CT data. For the segmentation method of rib centerlines detection is chosen. The first step of this approach is to extract the centerlines of all the bones located in the scan. These centerlines are divided into short primitives, which are subsequently classified into couple of categories, depending on whether they represent the centerline of the rib. Subsequently, the centrelines of ribs are used as the seed points of the region growing algorithm in three-dimensional space, which realizes the final segmentation of the ribs. Within the work, a database of 10 CT scans was manually annotated, which was subsequently used to validate a performance of the proposed segmentation approach. The achieved success rate of primitive classification is 96,7 %, the success rate of rib segmentation (Dice coefficient) is 86,8 %.
|
|
|
Convolutional neural networks for identification of axial 2D slices in CT data
Vavřinová, Pavlína ; Harabiš, Vratislav (referee) ; Jakubíček, Roman (advisor)
This thesis deals with the classification of axial 2D slices in CT patient’s data into six categories. The sphere of convolutional neural networks was used for this purpose. For a better understanding of this issue, the basics of neural networks and then the principles of deep learning including convolutional neural networks are explained at first. The AlexNet network was specifically selected for the intention of this identification, and it was tested on the created data set after being adaptated. The overall classification success rate was 86% ,after the final adjustments, a slight improvement was achieved and the identification success rate was 87%.
|
|
|
Time development analysis of treated lesion in spinal CT data
Nohel, Michal ; Jan, Jiří (referee) ; Jakubíček, Roman (advisor)
This diploma thesis is focused on time-development analysis of treated lesion in CT data. The theoretical part of the thesis deals with the anatomy, physiology, and pathophysiology of the spine and vertebral bodies. It further describes diagnostic and therapeutic options for the detection and treatment of spinal lesions. It contains an overview of the current state of usage of time-development analysis in oncology. The problems of the available databases are discussed and new databases are created for subsequent analysis. Futhermore, the methodology of time-development analysis according to the shape characterization and the size of the vertebral involvement is proposed. The proposed methodological approaches to feature extraction are applied to the created databases. Their choice and suitability is discussed, including their potential for possible usege in clinical practice of monitoring the development and derivation of characteristic dependences of features on the patient's prognosis.
|
guest ::
RSS feed.