Room: AAPM ePoster Library
To evaluate the feasibility of super-resolution reconstruction of T2-weighted (T2W) fat saturated (FS) knee MRI using lower-resolution single-plane overlapping images.
Five symptomatic patients with knee pain were scanned on a 3T Siemens MAGNETOM Vida scanner using an 18-channel knee coil. The sequences acquired and their corresponding pulse parameters are provided in Fig.1a. Super-resolution reconstruction was performed using two sagittal SMS2-T2w-FS acquisitions with overlapping their field-of-view (FOVs) by one-half of the slice thickness in the slice-select direction. Schematics of overlapping slice acquisition and the proposed reconstruction formulation are illustrated in Fig.1.b. The inverse problem of super-resolution reconstruction from low-resolution images was an ill-conditioned problem which was solved through Tikhonov regularization.
The SNR and CNR for cartilage, muscle, PCL, bone marrow, and meniscus were used to compare the reconstructed high-resolution images with directly acquired images.
Fig.1c-f show representative sagittal images from direct acquisitions and high-resolution reconstruction. Qualitatively, no noticeable difference was observed in SNR between low-resolution acquisition and reconstructed super-resolution images.
As shown in Fig.g-l the proposed super-resolution method significantly reduced the partial volume effect in sagittal plane as well as image quality and edge definition in reformatted images in coronal and axial planes compared to the standard (0.4x0.4x3mm³) acquisition.
There was no significant difference between the SNR and CNR of the standard sagittal T2W-FS images (0.4x0.4x3mm³) and the reconstructed super-resolution image (0.8x0.8x0.8mm³) (p=0.5)
Using the Simultaneous Multi-Slice (SMS) acquisition the required scan time for super-resolution reconstruction was only 1’ 15” longer than the standard sagittal (0.4x0.4x3mm³) scan.
The proposed super-resolution method generated similar SNR/CNR to the direct thick slice acquisition with the added benefit of being able to be reformatted into orthogonal planes with minimal blurring effect. This is a work in progress and further optimizations of SMS acquisition and improvement of the regularized solution are underway.