Kelvin Wong PhD, Presenter: Nothing to Disclose

Chi Pan Tam BSC, Abstract Co-Author: Nothing to Disclose

Michael K. Ng PhD, Abstract Co-Author: Nothing to Disclose

Stephen T.C. Wong PhD, Abstract Co-Author: Nothing to Disclose

We propose using Least-Absolute-Deviation (LAD) method with adaptive regularization to reduce the dependence on the shape of the arterial input function (AIF) in model independent MR perfusion deconvolution.

The AIF was simulated using a gamma variate function Ca(t) = C0(t-t0)a·e(-t-t0)/b for t > t0 where C(t) is the contrast concentration time curve. Due to partial volume effect, the value of a and b can take on various values. We simulate a range of values for a = 3, 4, 5 with b = 1.5 as well as a = 10 with b = 0.5, 1.0. The baseline SNR was set to 100, the simulated CBV from 2% to 4%, MTT ranged from 4s, 8s, 12s, 18s and R(t), the tissue residue function, was simulated using exponential and linear residue function models. Monte Carlo simulation was done for each setting averaged 100 times and the results were compared with the circular SVD method.

The estimated CBF for our method was consistently between 85% and 120% of the true CBF with a mean of about 95% while the circular SVD method ranged from 75% to 130%. Both methods had stable performance while our method has very little flow dependence.

This method has the inherent advantage to be able to reduce flow dependence and recover residue function shape better than SVD based methods. Studies are underway to evaluate the performance of this algorithm in cases with imperfect AIF shapes.

The LAD regularization deconvolution method that uses adaptive regularization parameters may reduce the AIF dependence and flow dependence issue in CBF estimation in MR perfusion.

Wong, K, Tam, C, Ng, M, Wong, S, Robust Adaptive Least-Absolute-Deviation Regularization Method to Reduce AIF Shape Dependence in MR Perfusion.  Radiological Society of North America 2009 Scientific Assembly and Annual Meeting, November 29 - December 4, 2009 ,Chicago IL. http://archive.rsna.org/2009/8016647.html