Clinical investigation of a knowledge-based data compression algorithm for dynamic neurologic FDG-PET images

A knowledge-based data compression algorithm for dynamic positron emission tomography (PET) images has been developed previously (D. Feng et al., “An optimal image sampling schedule design for cerebral blood volume and partial volume correction in neurologic FDG-PET studies”, The 29^th Annual Scientific Meeting of A & NZ Society of Nuclear Medicine, April 4-8, 1998, Melbourne, Australia, to appear in Australia and New Zealand Journal of Medicine), and it has been shown that this algorithm can greatly reduce image storage requirements in both spatial and temporal domains, while retaining medically relevant information. However, this algorithm has not yet been validated with real clinical dynamic image data. Here, the authors validated this algorithm in clinical dynamic brain PET studies using the [¹⁸F] 2-fluoro-deoxy-glucose (FDG) tracer. Compression performance and image quality were evaluated. Results demonstrate that the storage requirements for dynamic PET image data can be reduced by more than 95%, without loss in diagnostic quality. Furthermore, use of this compression algorithm greatly reduces the computational complexity of further clinical image processing such as generation of functional images. It could benefit to the current expansion in medical imaging, and image data management