Detection and validation of the body edge in low count emission tomography images

Leighton R. Barnden*, John Dickson, Brian F. Hutton

*Corresponding author for this work

    Research output: Contribution to journalArticle

    12 Citations (Scopus)

    Abstract

    Segmentation of the body edge in tomographic images with low count, noisy edges is needed for both PET and SPECT respiratory motion correction and brain SPECT attenuation correction. To reduce noise we re-projected tomographic images and searched for edges in the projection count profiles or their spatial derivatives. Edge location versus projection angle was fitted with cosine basis functions after rejecting outliers and including information about edges of previous sections. We processed 10 s duration FDG PET of the thorax, HMPAO brain, DAT brain and lung perfusion SPECT. Stable edges for all four types of scan were achieved but with different profiles. Edges were validated against edges of coregistered CT or MRI. The best root mean square (rms) accuracy was 8.2 mm (PET) and 5.2 mm (brain SPECT). Inter-scan variability (standard deviation) in the estimated-to-control edge distance for 17 PET scans was 0.4 mm, for 25 ordered subset expectation maximisation (OSEM) reconstructed brain SPECT 1.0 mm and for 18 filtered back-projection (FBP) reconstructed brain SPECT 1.4 mm.

    Original languageEnglish
    Pages (from-to)153-161
    Number of pages9
    JournalComputer Methods and Programs in Biomedicine
    Volume84
    Issue number2-3
    DOIs
    Publication statusPublished - Dec 2006

    Keywords

    • Body edge
    • OSEM
    • PET
    • Projection edge
    • SPECT

    Fingerprint Dive into the research topics of 'Detection and validation of the body edge in low count emission tomography images'. Together they form a unique fingerprint.

  • Cite this