نمونه برداري وفقي از تصوير MRI براي استفاده در حسگري فشرده

An Adaptive Method for Under-sampling of MRI Images Based on Compressive Sensing

Compressive sensing (CS) utilizes sparsity of MRI images for accurate reconstruction of under-sampled k-space data. In order to use MRI in CS, k-space image should be sampled and then CS techniques should be applied. Although most common sampling methods in CS framework may have good properties, they are not optimal in image reconstruction due to their finite data. In this paper, a new method will be presented for adaptive sampling consisting of two updating steps: namely as sampling method and image update steps. Given reconstructions are used in sampling update step and fixed sampling method are used in image update step besides convergence in PSNR. Wavelet transform and image blocking are also applied. The blocks used in the adaptive stage are chosen spirally leading to less calculations and maintaining low-frequency image information in the centre of k-space. Simulation results indicated 7.5dB improvement in PSNR reconstruction using adaptive sampling.

Compressive sensing (CS) utilizes sparsity of MRI images for accurate reconstruction of under-sampled k-space data. In order to use MRI in CS, k-space image should be sampled and then CS techniques should be applied. Although most common sampling methods in CS framework may have good properties, they are not optimal in image reconstruction due to their finite data. In this paper, a new method will be presented for adaptive sampling consisting of two updating steps: namely as sampling method and image update steps. Given reconstructions are used in sampling update step and fixed sampling method are used in image update step besides convergence in PSNR. Wavelet transform and image blocking are also applied. The blocks used in the adaptive stage are chosen spirally leading to less calculations and maintaining low-frequency image information in the centre of k-space. Simulation results indicated 7.5dB improvement in PSNR reconstruction using adaptive sampling.