A Comparative Study Between DWT and DCT Transformations for Compressed Sensing of Mouse Abdominal MRI
Alexandre Rodrigues Farias1; Hermes Aguiar Magalhães2; Márcio Flávio Dutra Moraes2; Eduardo Mazoni Andrade Marçal Mendes2
1 Centro Federal de Educação Tecnológica de Minas Gerais - CEFET-MG; 2 Universidade Federal de Minas Gerais - UFMG
Baixar PDF doi:10.20906/CPS/CBA2018-0669
Resumo
In this work, two different transformations are used alongside with the well-known compressed sensing methodology to evaluate the quality and time performance of image reconstruction using real magneto-resonance images from a mouse abdomen, namely, the discrete wavelet transform (DWT) and the discrete cosine transformation (DCT). For evaluating image quality, k-space data were subsampled using a pseudo-random variable density model with a reduction factor between 2-10. The quantitative analysis of the reconstructed images was made by using Pearson's correlation coefficient calculated between the reference and reconstructed images. The results of image reconstruction using DWT and DCT presented similar Pearson's coefficient values for all acceleration rates indicating similarity above 0.90 between the reference and reconstructed images up to acceleration rate 6. Using DWT, a reduction between 79% and 85% of time spent on the image reconstruction was achieved. The experimental results show the efficiency of DWT as far as keeping similarity with the reference image in terms of correlation and on accelerating mouse abdominal MRI is concerned.
Palavras-chave: Compressed Sensing; Undersampling Pattern; Small Animal; Magnetic Resonance Imaging (MRI); Variable Density