|
| |
|
|
Deep learning methods for vessel and optic disc segmentation in ophthalmologic sequences
Rozhoňová, Andrea ; Odstrčilík, Jan (referee) ; Hesko, Branislav (advisor)
The aim of the following thesis was to study the issue of optical disc and retinal vessels segmentation in ophthalmologic sequences. The theoretical part of the thesis summarizes the principles of different approaches in the field of deep learning, which are used in connection with the given issue. Based on the theoretical part, methods for optical disk segmentation and retinal vessel segmentation based on the convolutional neural networks Linknet, PSPNet, Unet and MaskRCNN are proposed. The practical part of the thesis deals with the description of their implementation and subsequent evaluation.
|
|
|
Text recognition with artificial neural networks
Peřinová, Barbora ; Hesko, Branislav (referee) ; Mézl, Martin (advisor)
This master’s thesis deals with optical character recognition. The first part describes the basic types of optical character recognition tasks and divides algorithm into individual phases. For each phase the most commonly used methods are described in the next part. Within the character recognition phase the problematics of artificial neural networks and their usage in given phase is explained, specifically multilayer perceptron and convolutional neural networks. The second part deals with requirements definition for specific application to be used as feedback for robotic system. Convolution neural networks and CNTK library for deep learning using algorithm implementation in .NET is introduced. Finally, the test results of the individual phases of the proposed solution and the comparison with the open source Tesseract engine are discussed.
|
|
|
Deep Convolutional Networks For Oct Image Classification
Hesko, Branislav
In this work, OCT (optical coherence tomography) images are classified according to the present pathology into four distinct categories. Three different neural network models are used to classify images, each model is recent and we are achieving exceptional results on the testing dataset, which was unknown to the network during the training. Accuracy on the testing set is higher than 98% and only a few of images are classified into the wrong category. This makes our approach perspective for future automatic use. To further improve results, all three models are using transfer learning.
|
|
|
Therapeutic game for reaction time measurement using the BITalino platform
Veselá, Cindy ; Mézl, Martin (referee) ; Hesko, Branislav (advisor)
This master’s thesis focuses on real-time detection of activity in electromyographic signal for reaction time measurement. For patients motivation there was designed and implemented therapeutic car game controlled throught the muscle activity. In this thesis were used three different algorithms for muscles activity detection in EMG signal. The best accuracy of this three methods has designed artificial network with U-Net hierarchy, which is used to segment samples into two categories - samples of signal with activity and samples representing calm. Accuracy of this method is 97 %. Later there were examined differences between groups of probands, different stimulus and the changes of reaction time over time.
|
|
|
Eye movement tracking using the Raspberry Pi platform
Hunkařová, Nikol ; Mézl, Martin (referee) ; Hesko, Branislav (advisor)
This master's thesis deals with eye movement tracking using the Raspberry Pi platform. The theoretical part describes eye anatomy, eye detection and eyetracking. A system in Python programming language was designed in the practical part. This algorithm is able to perform the eye tracking function using the Raspberry Pi platform and the RPi Camera module. The OpenCV library is used for loading and preprocessing images from the camera. A method that detects and evaluates the direction of view after a calibration is available. The accuracy of the program is tested on three vector methods and two target methods for four screen resolutions.
|
|
|
Deep learning methods for vessel and optic disc segmentation in ophthalmologic sequences
Rozhoňová, Andrea ; Odstrčilík, Jan (referee) ; Hesko, Branislav (advisor)
The aim of the following thesis was to study the issue of optical disc and retinal vessels segmentation in ophthalmologic sequences. The theoretical part of the thesis summarizes the principles of different approaches in the field of deep learning, which are used in connection with the given issue. Based on the theoretical part, methods for optical disk segmentation and retinal vessel segmentation based on the convolutional neural networks Linknet, PSPNet, Unet and MaskRCNN are proposed. The practical part of the thesis deals with the description of their implementation and subsequent evaluation.
|
|
|
Adjustment of diagnostic instrument Array Reader C-series
Čičatka, Michal ; Hesko, Branislav (referee) ; Mézl, Martin (advisor)
Thesis deals with actualization of application for adjustment of the chemical analyser Array Reader C-series developed by company BioVendor. At the beginning it defines and sorts out chemical analysers; then describes their physical principles. Following chapters sum up knowledge about microarray analysis and describes it’s chemical and physical principles. Furthermore, the thesis deals with quality and it’s management and machine adjustment. In the thesis is also mentioned .NET Framework – software platform that will be used in the practical part of the thesis. At the end of the theoretical part is described Array Reader C-series. Practical part documents current application for adjusting the analyser. Based on personal experiences with the adjusting application was created and thoroughly described concept of a new adjusting application. Main task of this new concept should be simplification and automatization of the adjusting process. After that implementation of the new concept is described including automatation processes. In the end of this thesis both original and new versions are compared with regards to time consumption and user experience. Newer version of the application was evaluted as better than it's previous version in light of time consumption and user experience. Application was expanded for fluorescent analyzer and for both types of analyzers were developed derivated applications for service adjustment. Each of these applications was implemented to the software of BioVendor company and is used on daily basis.
|
|
| |
|
|
Segmentation of the Common Carotid Artery Intima Media Using Active Contour Models
Hesko, Branislav
Ultrasound measurements of the human carotid artery intima media thickness are conventionally obtained using manual tracing between tissue layers. This article consists of different approach, when intima media thickness is determined automatically. Tissue layers are detected by use of active contour segmentation. This method does not use any type of gradient, only intensity based image is needed. Therefore, robustness to speckle noise is present. A total of 15 carotid artery ultrasound images is analyzed to determine wall thickness. For evaluation, each of the images contains manually added marks by an expert.
|
guest ::
RSS feed.