Reproducibility of magnetic resonance perfusion imaging

Abstract

Dynamic MR biomarkers (T2*-weighted or susceptibility-based and T1-weighted or relaxivity-enhanced) have been applied to assess tumor perfusion and its response to therapies. A significant challenge in the development of reliable biomarkers is a rigorous assessment and optimization of reproducibility. The purpose of this study was to determine the measurement reproducibility of T1-weighted dynamic contrast-enhanced (DCE)-MRI and T2*-weighted dynamic susceptibility contrast (DSC)-MRI with two contrast agents (CA) of different molecular weight (MW): gadopentetate (Gd-DTPA, 0.5 kDa) and Gadomelitol (P792, 6.5 kDa). Each contrast agent was tested with eight mice that had subcutaneous MDA-MB-231 breast xenograft tumors. Each mouse was imaged with a combined DSC-DCE protocol three times within one week to achieve measures of reproducibility. DSC-MRI results were evaluated with a contrast to noise ratio (CNR) efficiency threshold. There was a clear signal drop (>95% probability threshold) in the DSC of normal tissue, while signal changes were minimal or non-existent (<95% probability threshold) in tumors. Mean within-subject coefficient of variation (wCV) of relative blood volume (rBV) in normal tissue was 11.78% for Gd-DTPA and 6.64% for P792. The intra-class correlation coefficient (ICC) of rBV in normal tissue was 0.940 for Gd-DTPA and 0.978 for P792. The inter-subject correlation coefficient was 0.092. Calculated Ktrans from DCE-MRI showed comparable reproducibility (mean wCV, 5.13% for Gd-DTPA, 8.06% for P792). ICC of Ktrans showed high intra-subject reproducibility (ICC = 0.999/0.995) and inter-subject heterogeneity (ICC = 0.774). Histograms of Ktrans distributions for three measurements had high degrees of overlap (sum of difference of the normalized histograms <0.01). These results represent homogeneous intra-subject measurement and heterogeneous inter-subject character of biological population, suggesting that perfusion MRI could be an imaging biomarker to monitor or predict response of disease. © 2014 Zhang et al.