Partial volume correction using an energy multiresolution analysis

Position Emission Tomography (PET) allows the in-vivo monitoring of functional processes in the body. However its limited spatial resolution induces Partial Volume Effect (PVE), which leads to a loss of signal in tissues of size similar to the Point Spread Function (PSF) of the imaging device and induces activity spillover between adjacent structures with different amounts of activity. The aim of Mutual Multiresolulion Approach (MMA) is to introduce the high frequency information of an anatomical image into the functional image in order to achieve a higher resolution PET and to decrease PVE. Nevertheless, the anatomical details (high frequency MRI wavelet coefficients) can not be directly replaced into the functional image since they do not have the same spatial resolution neither the same wavelet coefficient intensities. Thus the process relies on a wavelet-based image merging, and aims at detecting, modifying and incorporating high resolution details of the MRI into the PET. In this work a Partial Volume Correction (PVC) algorithm was implemented based on the decomposition of two co-registered PET and MR images using a three-dimensional Discrete Wavelet Transform (3D DWT) and on both wavelet coefficient energies. To validate the performance of the proposed PVC design, regarding the quantitative measurements, the NEMA Image Quality phantom was used. Finaly, in order to demonstrate the use of the algorithm in clinical context, the technique was also applied to a ¹⁸F-FDG study of a patient with severe leukariosis, having already a T-1 weighted 3D MRI scan. The quantitative results demonstrate that the combination of MMA and the energy image analysis leads to a significant recovery of lost intensity induced by PVE. Results from this work are encouraging, however it presents its own limitations since it introduces an artifact in the final corrected functional image.