PURPOSE

Typical clinical brain diffusion-weighted imaging based on Echo Planar Imaging (DW-EPI) has limited spatial resolution compared to other routine neuro images due to its high sensitivity to B₀ inhomogeneities. DW-EPI with smaller voxel size causes further image distortion. Sensitivity encoding (SENSE) helps to reduce the voxel size without increasing image distortion, but it often suffers from increased noise-like artifacts on the center of the images due to the high geometry factor^1,2. It has been shown recently that DW-EPI with Compressed SENSE enables pseudo-3D (2D multi-slice acquisition with ultra-thin-slice thickness) whole-brain diffusion imaging with voxel size of 1.15 mm³ in a feasible scan time³. Alternatively, single-shot turbo spin-echo (TSE) DWI is a distortion-free imaging approach but it has also limited spatial resolution due to its low SNR.
Recently, CS-AI, integrating artificial intelligence (AI) into the Compressed SENSE reconstruction, based on Adaptive-CS-Net^4,5, has been introduced. We hypothesize that the CS-AI reconstruction will further improve the SNR of both DW-EPI and DW-TSE with high spatial resolution and ultra-thin-slice pseudo-3D acquisition. In this study, we attempt to utilize the CS-AI framework for optimizing the pseudo-3D whole-brain diffusion imaging. The purpose of this study is to demonstrate the feasibility of CS-AI in pseudo-3D DW-EPI and DW-TSE imaging.

MATERIALS AND METHODS

A total of five volunteers/patients were examined on a 3.0T whole-body clinical system (Ingenia Elition X, Philips Healthcare). The study was approved by the local IRB, and written informed consent was obtained from all subjects.
CS-AI DWI is based on single-shot DW-EPI acquisition with a regular sampling pattern after which the CS-AI framework is used for reconstruction. The CS-AI model used in this study is the extension of the Adaptive-CS-Net algorithm. In CS-AI, the iterative optimization procedure in the C-SENSE reconstruction chain is unrolled for a fixed number of Unet type of blocks. The model was trained on more than 700,000 images, including 2D and 3D data, and multiple contrasts and anatomical areas.
CS-AI-DWI images were compared with conventional SENSE and CS-DWI images for image quality, especially for the reduction of image noise.Imaging parameters for pseudo-3D EPI-DWI were as follows: axial acquisition, voxel size=1.15mm³, FOV=230*230mm, 93slices, b-value=0 and 1200s/mm², TR=16196ms, TE=65ms, SENSE/C-SENSE acceleration factor = 4.0, and total acquisition time=5min40s, and imaging parameters for pseudo-3D TSE-DWI were: axial acquisition, voxel size=1.5mm³, FOV=220*220mm, 93slices, b-value=0 and 1200s/mm², TR=5579ms, TE=80ms, SENSE/C-SENSE acceleration factor = 4.0, and total acquisition time=3min16s.

RESULTS AND DISCUSSION

Figure 1 shows representative b=0, 1200s/mm² images and ADC maps using EPI-DWI with SENSE, CS and CS-AI in a volunteer. CS and CS-AI clearly reduced the noise in the center of the SENSE images. CS-AI further improved the contrast among gray matter, white matter and CSF. Figure 2 shows representative 3mm MPR images with three directions obtained by pseudo-3D EPI-DWI. The p3D-DWI with EPICS achieved high-resolution (1.15 mm³) isotropic DWI within clinically feasible scan time.
On the other hand, Figure 3 shows representative b=0, 1200s/mm² images and ADC maps using TSE-DWI with SENSE, CS and CS-AI in a volunteer and 3mm MPR images with three directions obtained by pseudo-3D TSE-DWI are shown in Figure 4. Similar to EPI-DWI, TSE-DWI with CS-AI clearly reduced the noise and further improved the contrast among gray matter, white matter and CSF.
Figure 5 shows representative clinical images of a patient with epidermoid (arrow). CS-AI improved the SNR and contrast especially for the skull base where epidermoid lesion is existing.

CONCLUSION

CS-AI clearly improves the image quality of whole-brain ultra-thin-slice pseudo-3D DWI using single-shot EPI or single-shot TSE acquisition, compared to conventional SENSE and CS DWI. Since this technique is simple and does not require any additional phase correction unlike 3D multi-shot DWI, it would be promising for robust whole-brain high-resolution diffusion imaging within a clinically feasible acquisition time.

Acknowledgements

No acknowledgement found.