Publication

Abstract : Intravoxel Incoherent Motion (IVIM) based Magnetic Resonance Imaging (MRI) technique allows the quantitative evaluation of perfusion and diffusion without the use of contrast agents. The correctness of the diagnosis depends upon the accuracy and precision of IVIM parameter estimation. To achieve this, several diffusion weighted images must be acquired. The criteria of selection of diffusion weights varies among researchers. The diffusion weights are incorporated by a factor called ‘b value’, which reflects the intensity and duration of the diffusion gradient pulses used for imaging. IVIM imaging takes more time for image acquisition with multiple b values and the selection of the absolute b values as well as the number of b values, is a real challenge. As the b value count increases the scan time increases, which leads to increased patient discomfort and motion artifacts, resulting in poor image quality. Moreover, in most cases, these b values are found using trial and error methods during image acquisition. These issues can be addressed to a large extent by finding, optimum number and range of b values. In this paper, we propose a population based Metaheuristic algorithm for arriving at the optimal b value count and the range of absolute b values for liver, which is an organ with high perfusion. Three separate models are developed to appropriately choose all possible b values ranging between 0 s/mm2 and 850 s/mm2, to observe the effects of diffusion and perfusion as well as to increase the global search space. The effect of low (0 s/mm2 to 50 s/mm2 ), medium (55 s/mm2 to 220 s/mm2) and high (230 s/mm2 to 800 s/mm2) b values has been studied to find the optimal number of b values that can be used for IVIM imaging. In order to define a b value count that minimizes the error in IVIM parameters, simulation experiments are performed for different b value counts specifically 16, 14, 12, 10, 8, 6 and 4. For each of these experiments, repeated observations are made to analyze the parameter uncertainty. The results obtained show that the b value count can be minimized for a better quantitative estimation of IVIM parameters with the least possible errors and the difference in error between each of these observations is found to be less than 0.001. Minimization of b value count results in the reduction of overall image acquisition time and hence patient discomfort. It is also observed that for a very good SNR, b value count can be reduced to 4 although for a reasonable SNR, 8 or 10 b values are to be used for accurate quantitative estimation of IVIM parameters.