Abstract #3614

Onset Estimation for Dual Input DCE-MRI Liver Data : Information Criteria Used to Determine Statistical Optimality of Global or Pixel-Wise Onset Estimation

Matthew R. Orton¹, Keiko Miyazaki¹, Dow-Mu Koh², David J. Collins¹, David Atkinson³, David J. Hawkes³, Martin O. Leach¹

¹Clinical Magentic Resonance Research Group, Institute of Cancer Research, Sutton, Surrey, UK; ²Academic Department of Radiology, Royal Marsden Hospital, Sutton, Surrey, UK; ³Centre for Medical Image Computing, Univeristy College London, London, UK

The onset time of contrast enhancement is typically estimated pixel-wise when analysing DCE-MRI data. We have developed an algorithm for estimating the onset time as a single global parameter, which has the benefit of reducing the variability in such estimates. This is particularly beneficial when a dual-input liver model is used as the arterial-portal fraction is highly correlated with the onset time. Statistical information criteria are used to determine whether global or pixel-wise onset estimation is optimal when applied to data acquired from patients with liver metastases. This analysis suggests that in this case global estimation is optimal.