|
|
Detection of Diseases Caused by Diabetes in Retinal Images
Zapletal, Michal ; Semerád, Lukáš (referee) ; Kavetskyi, Andrii (advisor)
The goal of this thesis is to design and implement an algorithm for detecting exudates and microaneurysms in colored retinal images. These diseases are the first signs of diabetic retinopathy and early detection is crucial. The proposed algorithm begins with preprocessing, where excess background is removed, contrast is enhanced using CLAHE and histogram stretching, and noise filtering is applied. Optic disc localization is based on iterative background removal and row and column variances. Exudates detection is performed based on gamma correction, thresholding and optic disc removal. Microaneurysm detection is based on morphological operations, hit-or-miss transformation and principal component analysis (PCA). The algorithm was tested on 4 datasets with accuracy 73,1 % for exudates and 73,3 % for microaneurysms. The resulting program could assist in automatic disease detection, which could potentially save time for doctors.
|
|
|
Application for Person Authentication Based on Retinal Images
Moncz, Oliver ; Goldmann, Tomáš (referee) ; Kavetskyi, Andrii (advisor)
This paper deals with the problem of authentication of people based on retinal images. The main objective is to design an algorithm that can match the retinal image with the registered images in the database. The outcome is the determination of identity with a degree of certainty. Another goal was to create an application that allows to perform the mentioned operations through a simple graphical user interface. Lastly, the results are evaluated and compared with existing solutions. The proposed authentication system achieved an average accuracy of 72.46 % for the DRIVE and STARE datasets, and 78.9 % for the FIRE dataset.
|
|
|
Retinal Images Generation with a Limited Amount of Training Data
Senichak, Yahor ; Semerád, Lukáš (referee) ; Kavetskyi, Andrii (advisor)
The purpose of this study is to explore the progress and application of computer vision and generative adversarial networks (GANs3.1) in the diagnosis and study of fundus diseases. Particular attention is paid to the latest advances in the field of medical data synthesis and the development of our own algorithm. Recent advances in the deep learning architecture U-GAT-IT [22], which includes two pairs of deep neural networks (two generators and two discriminators), have been implemented. This implementation was trained for approximately 300,000 iterations, during which positive results were obtained. The dynamics of the training process were recorded and tests were performed to demonstrate the ability to generate high-quality synthetic images of the ocular background independent of the input data
|
|
|
Semantic Segmentation of Pathologies in Retinal Images
Čabala, Roman ; Orság, Filip (referee) ; Kavetskyi, Andrii (advisor)
The thesis aimed to segment pathology visible in the retina images, such as exudates, hemorrhages, and microaneurysms. For that, two well known deep neural networks, named U-Net and SegFormer, were trained. To test the performance of the models, one publicly available dataset was used, named IDRiD. Obtained results were reported after analyzing different factors which affected the performance of the models U-Net and Segformer.
|
|
|
Vein-Artery Segmentation of Blood Vessels in Retinal Images
Sedlář, Radek ; Kanich, Ondřej (referee) ; Kavetskyi, Andrii (advisor)
This work focuses on an introduction to the issue of segmentation of veins and arteries from retinal images. The work contains a comparison of the most used methods with their pros and cons. Furthermore, a proprietary method for segmentation and division into veins and arteries is proposed. The work also contains a detailed description of the implementation of the proposed method and a summary of their results.
|
|
|
Retinal Blood Vessel Segmentation
Nemčeková, Barbora ; Drahanský, Martin (referee) ; Kavetskyi, Andrii (advisor)
The retina is an important part of the human eye. Incident light is processed here and moreover, it plays an essential role in diagnosing various diseases. Its early diagnostics can prevent serious consequences, such as blindness. The most common retinal diseases include diabetic retinopathy, as a consequence of diabetes, and age-related macular degeneration. Automatic retinal vessels segmentation facilitates and speeds up the work of an ophthalmologist. This work focuses on retinal blood vessels segmentation and its further classification into thin and thick vessels. The proposed algorithm is based on morphological operations, k-means clustering, and Frangi's algorithm. Evaluation of the proposed method was performed on two publicly available datasets - Drive and HRF. The results obtained represent 69,89 % for sensitivity, 91,55 % for specificity, and 88,63 % for accuracy. Division of the vessels shows, that on average 21,50 % vessels pixels belong to thick vessels and the rest 78,50 % belong to thin vessels.
|
|
|
Laptop Touchpad Palm Detection with AI/ML
Menzyński, Mark Alexander ; Kavetskyi, Andrii (referee) ; Drahanský, Martin (advisor)
Situace ohledně detekci a odmítnutí dlaně na laptopech je méně než ideální. Většina výzkumů se zabývá odmítnutím dotyků na dotykových obrazovkách, a na laptopy probíhá téměř žádný. Patrně nějaký uzavřený výzkům probíhá uvnitř výrobců laptopů, ale i přes to je technologie pozadu. Tato práce prozkoumává několik metod plytkého a hlubokého strojového učení, a výsledná přesnost byla zjištěna jako více než dostačující. Také implementuje aplikaci v reálném čase na demonstraci modelu.
|
|
|
Pathologies Detection in Retinal Images
Hurta, David ; Drahanský, Martin (referee) ; Kavetskyi, Andrii (advisor)
The main goal of this work is to design and implement an algorithm for the detection of microaneurysms, hard exudates, and soft exudates on color fundus images. An algorithm for detecting objects based on deep learning has been proposed. The Faster R-CNN architecture with a feature pyramid network and a pre-pretrained residual network was used together with various data transformation methods. A total of six retinal image datasets were used to train, validate and test the models. The trained models achieved 0.46 mean average accuracy (mAP) in microaneurysm detection and 0.48 mAP in exudates detection during testing. The resulting models have been compared with published articles and make it possible to detect given pathologies with commendable accuracy.
|
|
|
Semantic Segmentation of Pathologies in Retinal Images
Čabala, Roman ; Orság, Filip (referee) ; Kavetskyi, Andrii (advisor)
The thesis aimed to segment pathology visible in the retina images, such as exudates, hemorrhages, and microaneurysms. For that, two well known deep neural networks, named U-Net and SegFormer, were trained. To test the performance of the models, one publicly available dataset was used, named IDRiD. Obtained results were reported after analyzing different factors which affected the performance of the models U-Net and Segformer.
|
|
|
Automated Retinal Images Quality Assessment Using Machine Learning
Mikheda, Vladislav ; Vaško, Marek (referee) ; Kavetskyi, Andrii (advisor)
This work focuses on solving the problem of retinal image quality assessment. When diagnosing a disease, physicians focus on the quality of individual anatomical structures of the retina, according to which the diagnosis is made. The aim of this work is to design and implement a program for automated quality assessment of retinal images based on anatomical structures using neural networks. Overall six neural networks were developed and implemented to solve the abovementioned problem. Three of them were to segment individual anatomical structures of the retina, and three others were meant to evaluate images based on the quality of the segmented structure. Testing of each neural network separately, as well as testing of the entire program, was performed. The model allows the evaluation of the quality of retinal images based on the anatomical structures.
|
guest ::
