Simulation-based analysis of the effect of hot spot size, image SNR and undersampling scheme on compressed sensing reconstruction of MR temperature images during HIFU treatment

Abstract

Compressed sensing MRI can accelerate MRI temperature monitoring of tissues undergoing the high-intensity focused ultrasound (HIFU) treatment. This paper investigates the relationship between the accuracy of the MRI temperature maps reconstructed using reference image-based compressed sensing and some parameters such as the HIFU hot spot size, the image signal-to-noise ratio (SNR) and the k-space undersampling scheme by a simulation approach. Results indicate that a big HIFU hot spot size almost always helps to reduce the temperature error for Cartesian variable-density undersampling schemes but scarcely influences the reconstruction performance for the radial undersampling scheme. Moreover, the temperature error almost always increases with the image noise level. The radial undersampling scheme outperforms its Cartesian counterparts under the condition of a small HIFU hot spot size, high image SNR and a high degree of undersampling. Our future work will include finding suitable Cartesian undersampling patterns prospectively for our specific HIFU treatment and reducing the reconstruction error at low image SNR as well as the reconstruction time for the radial undersampling scheme.