Whole body PET image compression
Magnus L Dahlbom, K.A. Gardner, Arion F. Chatziioannou, Carl K. Hoh · 2002
A major difficulty in providing rapid transfer of whole body PET (WB-PET) images to referring physicians at remote hospitals is the large size of the data set. A WB-PET study contains typically over 400 images, or over 35 Mbyte of data. Using standard high speed modems (14.4 kbps) and fast communication protocols, a transfer time of 7-8 hrs/study is required. To reduce the transfer time, a means of compressing the data is necessary. Due to the relatively high noise levels in WB-PET images, conventional error-free compression algorithms provides little or no compression. The compression can be improved by processing the image data prior to compression. In this work, different pre-processing approaches are evaluated as well as different compression algorithms. By removing all negative values, who carries little or no diagnostic information, the compression is slightly improved. A more significant improvement was achieved by masking the transaxial images such that all pixel elements located outside the body contour are set to zero. Three different masking schemes based on geometrical shapes (circular, rectangular and elliptical) were evaluated. In addition, a masking scheme based on transmission images was also investigated. Using the transmission image mask, with negative value removal, and an efficient error-free compression algorithm, compressions of up to 1:7 were obtained, with no loss in image information.>