A multi resolution and multi observation framework for multi modal medical images processing and analysis

Motivated by the need for a framework dedicated to multi modal images processing and analysis, as well as Wavelet and Contourlet transforms properties, we developed a WCHMT (Wavelet-Contourlet Hidden Markov Tree) as a quadtree-structured probabilistic graph that captures the statistical properties of the whole observed data. This combining strategy was applied to exploit advantages of both Wavelet and Contourlet transforms. The proposed framework has been first evaluated in this work for mono modal PET image denoising. The results of denoising, using the proposed WCHMT with compared to Gaussian filtering and local adaptive Bishrink wavelet denoising currently considered as the state-of-the-art filter in PET imaging, are promising. High increases of the SNR were obtained with both methods. However, denoising performed through WCHMT led to lower quantitative bias in absolute intensities and a better structure preservation. Future work will investigate the potential of the WCHMT for multi modal (PET/CT, PET/MR) images segmentation, deconvolution and filtering.