Abstract
Maximum Likelihood (ML) estimation based Expectation Maximization (EM) reconstruction algorithm has shown to provide good quality reconstruction for PET. Our previous work introduced the multigrid EM (MGEM) and multiresolution (MREM) and Wavelet based Multiresolution EM (WMREM) algorithm for PET image reconstruction. This paper investigates the use of various wavelets in the new Wavelet based Multiresolution EM (WMREM) algorithm. The wavelets are used to construct a multiresolution data space, which is then used in the estimation process. The beauty of the wavelet transform to provide localized frequency-space representation of the data allows us to perform the estimation using these decomposed components. The advantage of this method lies with the fact that the noise in the acquired data becomes localized in the high-high or diagonal frequency bands and not using these bands for estimation at coarser resolution helps speed up the recovery of various frequency components with reduced noise estimation. Different wavelet bases result in different reconstructions. Custom wavelets are designed for the reconstruction process and these wavelets provide better results than the commonly known wavelets. The WMREM reconstruction algorithm is implemented to reconstruct simulated phantom data and real data.
| Original language | English (US) |
|---|---|
| Pages (from-to) | I/- |
| Journal | Proceedings of SPIE - The International Society for Optical Engineering |
| Volume | 3979 |
| State | Published - 2000 |
| Externally published | Yes |
| Event | Medical Imaging 2000: Image Processing - San Diego, CA, USA Duration: Feb 14 2000 → Feb 17 2000 |
All Science Journal Classification (ASJC) codes
- Electronic, Optical and Magnetic Materials
- Condensed Matter Physics
- Computer Science Applications
- Applied Mathematics
- Electrical and Electronic Engineering