Journal: Magma (New York, N.y.)

Article Title: Field-cycling imaging yields repeatable brain R 1 dispersion measurement at fields strengths below 0.2 Tesla with optimal fitting routine

doi: 10.1007/s10334-025-01230-w

Figure Lengend Snippet: R 1 mapping results for a single participant. Top right of figure shows the FLAIR image obtained from 3 T MRI and co-registered to FCI image space. Brain maps consist of quantitative maps of R 1 at 0.2 mT (left) and dispersion slope b (right). Maps are shown for each fitting model F1 and S1–S4, with motion correction and denoising applied before fitting. Image contrast contained within R 1 maps can be seen to differentiate between SVD regions (hypointense) and WM and GM regions (hyperintense). Matching histogram distributions of R 1 at 0.2 mT are shown for regions of WMH (black) and WM (red)

Article Snippet: After motion correction, images were denoised using a pretrained denoising convolutional neural network (dnCNN) approach contained within MATLAB, introduced in R2017b [ ].

Techniques: Dispersion

Journal: Scientific Reports

Article Title: Phase-based fast 3D high-resolution quantitative T 2 MRI in 7 T human brain imaging

doi: 10.1038/s41598-022-17607-z

Figure Lengend Snippet: Human whole-brain T 2 maps with a 0.85 mm isotropic voxel. ( a ) without denoising, ( b ) with denoising, based on DnCNN model for Gaussian noise removal. Arrows point to the cerebellum region, which especially benefits from denoising. Top row, Sagittal and Coronal planes. Bottom two rows, six slices of the Axial plane, at 10 mm intervals.

Article Snippet: To provide even higher robustness following the reduced SNR of the high-resolution datasets, we also incorporated denoising based on a DnCNN deep-learning network (provided in MATLAB, The Mathworks, Natick MA, for Gaussian noise removal).

Techniques:

Journal: Scientific Reports

Article Title: Boosting the signal-to-noise of low-field MRI with deep learning image reconstruction

doi: 10.1038/s41598-021-87482-7

Figure Lengend Snippet: 3D brain image reconstruction at 6.5 mT using AUTOMAP compared to conventional IFFT with or without additional image-based denoising pipelines. ( a – c ) Reconstruction of 3D human head dataset—an 11-min (NA = 50) 3D acquisition dataset was reconstructed with AUTOMAP ( a ) and IFFT ( b ). Shown here are 10 slices from the full 15 slice dataset. For comparison, a 22-min (NA = 100) acquisition reconstructed with IFFT is shown in ( c ). The window level is unchanged in all images. ( d , e ) The two denoising algorithms (DnCNN and BM3D respectively) were applied to the IFFT reconstructed brain image (magnitude only) to compare to the denoising performance of AUTOMAP. ( f – i ) Noise floor comparison—Slice 4 from the NA = 50 reconstructed brain dataset shown above in ( a , b ) is displayed here with two different window levels: a normalized image on the top and a window level chosen to highlight the noise at the bottom. AUTOMAP is shown in ( f ), and IFFT is shown in ( g ). An additional DnCNN or BM3D image denoising step was applied to the image data reconstructed with IFFT ( h , i respectively).

Article Snippet: The first denoiser is a state-of-the-art image-only deep learning denoising approach (DnCNN) , (recently incorporated as a built-in MATLAB function denoiseImage), which utilizes a deep single-scale convolutional neural network in a residual learning context to perform the denoising.

Techniques: Comparison

Journal: Scientific Reports

Article Title: Boosting the signal-to-noise of low-field MRI with deep learning image reconstruction

doi: 10.1038/s41598-021-87482-7

Figure Lengend Snippet: Image metric analysis on the 3D human brain dataset. ( a – d ) Image metric analysis of AUTOMAP and IFFT reconstruction with- and without the DnCNN step or the BM3D denoiser following transformation of the raw k -space data—The mean overall SNR in the whole-head ROI across all the 15 slices is shown in ( a ) for IFFT (filled circle), denoised IFFT with BM3D (filled square) with DnCNN (filled triangle), AUTOMAP (open circle). Three additional metrics are computed: PSNR ( b ), RMSE ( c ), and SSIM ( d ). ( e ) The table summarizes the mean PSNR, SSIM, RMSE, SNR and SNR gain values across all the slices. The SNR gain was calculated with respect to the conventional IFFT.

Article Snippet: The first denoiser is a state-of-the-art image-only deep learning denoising approach (DnCNN) , (recently incorporated as a built-in MATLAB function denoiseImage), which utilizes a deep single-scale convolutional neural network in a residual learning context to perform the denoising.

Techniques: Transformation Assay

Journal: Scientific Reports

Article Title: Boosting the signal-to-noise of low-field MRI with deep learning image reconstruction

doi: 10.1038/s41598-021-87482-7

Figure Lengend Snippet: AUTOMAP reconstruction using the RootBox synthetic roots database versus IFFT reconstruction of a 96 × 96 root dataset. All eight 2D projections reconstructed with AUTOMAP are shown in ( a ), and with IFFT in ( b ). Each of the magnitude image was processed with either DnCNN as shown in the third panel ( c ) or with BM3D in the fourth panel ( d ) The window level for projections 1–7 were set to the same value except for projection 8, where the threshold was lowered on both panels to reveal the noise floor differences. To generate figure ( e ), the SNR was evaluated for AUTOMAP reconstruction using the RootBox training and compared to IFFT reconstruction with and without the denoising pipelines, and the SNR of each of the 8 projections reconstructed with AUTOMAP (open circle), IFFT without the denoising algorithm (filled circle), IFFT with BM3D (filled square), and IFFT with DnCNN (filled triangle) is plotted.

Article Snippet: The first denoiser is a state-of-the-art image-only deep learning denoising approach (DnCNN) , (recently incorporated as a built-in MATLAB function denoiseImage), which utilizes a deep single-scale convolutional neural network in a residual learning context to perform the denoising.

Techniques: