A new computation compression scheme based on a multifrequential approach

Nowadays medical applications such as functional imaging (PET and SPECT) or interventional imaging (CT fluoroscopy) can involve dynamic data. Thus, the computation time must be reduced. Therefore, the authors developed a new fast algorithm for dynamic reconstruction based on computation compression. The computation compression and data compression have the same fundamental steps of “decomposition”, “quantification” and “uncompression”. The authors´ algorithm performs an indirect subband decomposition of f(f=Σf_i) through Rf which is filtered. But as for data compression, they want to preserve the pertinent information only. To obtain this result, they reduce the number of components: Rf_i is filtered and backprojected only if f_i is significant (quantification step). Finally, to estimate signal, the authors uncompress the components (uncompression step). They implement their approach with a subband Fourier decomposition. But alternatively Cosinus, Wavelet or Karhunen Loeve decomposition could also be used. This algorithm allows to control the compression ratio and the signal quality. It yields an easy parallelization and leads to a computation time reduction directly related to the data compression rate