4856 INVERSION: A robust method for co-registration of MPRAGE and Diffusion MRI images

2014
2image to look like an MPRAGE image. This transformation means that inter-modal image registration methods can be replaced by simpler and more robust methods designed for registering images from the same-modality. In this work, we invert the intensity histogram of the b=0s/mm 2 image, and then register this image to the MPRAGE image using the SSD cost function. In practice, we also apply histogram matchingto the inverted b=0s/mm 2 image and MPRAGE image to further refine the intensity match. In order to study the robustness of the proposed method we acquired an MPRAGE image, a diffusion dataset (single-shot EPI, TE=115ms, TR=10s, 65 diffusion-weighted image with b=2500s/mm 2 , one b=0s/mm 2 , 2x2x2mm) and a B0 inhomogeneity map for a single subject. The diffusion dataset was first corrected for susceptibility induced distortion using the acquired inhomogeneity map 4 . Then the MPRAGE image was aligned to b=0s/mm 2 image using a manually guided rigid registration procedure 5 . This manual result was used as a gold-standard to compare the accuracy of different automatic registration methods. In order to understand the properties of different cost functions, we studied how they changed as a function of mis-registration (translation along the x-axis). Consistency of the solutions obtained by different cost functions was evaluated by applying 36 known rigid transformationsto the MPRAGE image and assessing the registration quality achieved with each cost function using our implementation of CR and NMI. The registration accuracy was quantified using RMS error 3 . All cost functions were optimized using simple gradient decent.
    • Correction
    • Source
    • Cite
    • Save
    1
    References
    0
    Citations
    NaN
    KQI
    []
    Baidu
    map