Návrh modelu pro prostorové a volumetrické modelování efektivity vlnkové transformace pro 1D a 2D medicínská data

Abstract

This diploma thesis deals with the creation of a model and its robust testing for tracking and predicting the optimal setting of the wavelet transformation during filtering and decomposition of medical 1D and 2D data. Procedures for evaluating the quality of filtering on selected datasets of 1D signals and 2D images are created and robustly tested in the work, for a large number of variable settings that the wavelet transformation offers, and a large range of noise intensities with the option of choosing a specific type of noise. The quality of the filtering is evaluated on the basis of the evaluation parameters, which are the root mean square error, the Euclidean distance, the correlation index and the structural similarity index. On the basis of these parameters, the quality, accuracy and consistency of the filtering of individual settings of the wavelet transformation for a specific type of noise are evaluated. For the prediction of quality settings, 2D maps are implemented, which show in color the best settings of the wavelet transformation for the selected data and specific noise intensity. Subsequently, 3D color maps are also processed, on which the best settings of the wavelet transformation for multiple levels of noise intensities can be recognized. Based on the procedures, it is also possible to visualize the maximum achievable correlation for a specific setting for all wavelet families depending on the dynamically changing noise intensity for individual datasets.