Meeting Banner
Abstract #1854

Apparent Diffusion Coefficient Estimation in Prostate DW-MRI Using Maximum Likelihood

Valentin Hamy1, Simon Walker-Samuel2, David Atkinson1, Shonit Punwani1

1Centre for Medical Imaging, UCL, London, United Kingdom; 2Centre for Advanced Biomedical Imaging, UCL, London, United Kingdom

The problem of apparent diffusion coefficient (ADC) estimation from Rician distributed diffusion-weighted magnetic resonance (DW-MR) data is addressed. The least squares (LS) algorithm, widely used in clinical practice, is known to produce biased estimates as it considers the noise as normally distributed. Maximum likelihood (ML) can provide a more robust alternative. In this study based on prostate cancer DW-MR, we compared LS and ML efficiency, for signal to noise ratios typical of the different types of tissue. The ML approach provided significantly less biased estimates than the LS, potentially allowing better accuracy in prostate cancer grading from MR images.

Keywords

absolute account accuracy achieved advanced aggressiveness alternative always among apparent appearing appears application approaches asked assessment benefits better biased bigger biomedical blue cancer cancerous clearly clinical close coefficient combination common concerning conditions confirming consisted contoured correlation corrupted couple covers creation dataset decay detection diagnose diagnosed diffusion disease distributed error errors estimation evaluate example experience experiment exponential expressed finally fitted fitting generally generated grade ground healthy help heterogeneous impacting introduced invasive investigates kingdom larger least likelihood localized magnitude magnitudes maps marker median medical noise patient patients percentage performance peripheral pixels potential predict preference preferred previous produces prostate qualitative quality quantify radical radiologist radiologists radiology random real realistic related reliable repeated representation representations representing respect robust robustness routine sensitivity significantly simulated simulation simulations solid specify squares studied suggesting surrounded taken threshold thresholds tissue tried true truth tumors typical unbiased underestimation usually variations various varying visual walker whereas years yields zone