Donglai Huo MS, Presenter: Nothing to Disclose

Dan Xu MS, Abstract Co-Author: Nothing to Disclose

Zhipei Liang PhD, Abstract Co-Author: Nothing to Disclose

David L. Wilson PhD, Abstract Co-Author: Nothing to Disclose

Many methods have been proposed to improve image reconstruction from sensitivity-encoded (SENSE) phased array MRI data. Selection of these methods and the embedded parameters under different circumstances require an effective, quantitative image quality evaluation tool.

The Perceptual Difference Model (PDM) is a computer human vision model that calculates the visual difference between a “test image” and a “gold standard image.”, and it has been shown to correlate well with human observers in a variety of MR experiments including spiral imaging and keyhole imaging. We performed human observer experiments for the verification of PDM for parallel imaging. We generated data sets for four configurations of phased array receiver coils; six SNR levels for the sensitivity maps, and five SNR levels for measured data. Images were acquired using three sensitivity map estimation methods and six reconstruction methods. PDM was used to evaluate these 2160 images and the PDM error scores were statistically analyzed.

High correlation between the PDM error scores and human observer scores showed that PDM is effective in evaluating parallel MR image quality. Not unexpectedly, image quality improves as the number of coils, the SNR of the sensitivity maps, and the SNR of the input images increases. However, increasing the SNR of the sensitivity maps beyond approximately 20dB provides little increase in image quality. PDM gives us a method for making quantitative comparisons as one designs the acquisition scheme. Our studies also show that the wavelet denoising method is effective in improving estimation of the coil sensitivity functions; a new method, “SENSE plus regularization with prior information,” proved to be better than other, more conventional reconstruction methods.

PDM provides an objective, useful tool for the assessment of parallel MR image quality and can greatly aid the design of parallel MR acquisition and signal processing strategies.

Huo, D, Xu, D, Liang, Z, Wilson, D, Quantitative Image Quality Evaluation of SENSE Reconstruction Techniques.  Radiological Society of North America 2004 Scientific Assembly and Annual Meeting, November 28 - December 3, 2004 ,Chicago IL. http://archive.rsna.org/2004/4412488.html