No abstract

BackgroundAlthough susceptibility‐weighted imaging (SWI) is the gold standard for visualizing cerebral microbleeds (CMBs) in the brain, the required phase data are not always available clinically. Having a postprocessing tool for generating SWI contrast from T2*‐weighted magnitude images is therefore advantageous.PurposeTo create synthetic SWI images from clinical T2*‐weighted magnitude images using deep learning and evaluate the resulting images in terms of similarity to conventional SWI images and ability to detect radiation‐associated CMBs.Study TypeRetrospective.PopulationA total of 145 adults (87 males/58 females; 43.9 years old) with radiation‐associated CMBs were used to train (16,093 patches/121 patients), validate (484 patches/4 patients), and test (2420 patches/20 patients) our networks.Field Strength/Sequence3D T2*‐weighted, gradient‐echo acquired at 3 T.AssessmentStructural similarity index (SSIM), peak signal‐to‐noise‐ratio (PSNR), normalized mean‐squared‐error (nMSE), CMB counts, and line profiles were compared among magnitude, original SWI, and synthetic SWI images. Three blinded raters (J.E.V.M., M.A.M., B.B. with 8‐, 6‐, and 4‐years of experience, respectively) independently rated and classified test‐set images.Statistical TestsKruskall–Wallis and Wilcoxon signed‐rank tests were used to compare SSIM, PSNR, nMSE, and CMB counts among magnitude, original SWI, and predicted synthetic SWI images. Intraclass correlation assessed interrater variability. P values <0.005 were considered statistically significant.ResultsSSIM values of the predicted vs. original SWI (0.972, 0.995, 0.9864) were statistically significantly higher than that of the magnitude vs. original SWI (0.970, 0.994, 0.9861) for whole brain, vascular structures, and brain tissue regions, respectively; 67% (19/28) CMBs detected on original SWI images were also detected on the predicted SWI, whereas only 10 (36%) were detected on magnitude images. Overall image quality was similar between the synthetic and original SWI images, with less artifacts on the former.ConclusionsThis study demonstrated that deep learning can increase the susceptibility contrast present in neurovasculature and CMBs on T2*‐weighted magnitude images, without residual susceptibility‐induced artifacts. This may be useful for more accurately estimating CMB burden from magnitude images alone.Evidence Level3.Technical EfficacyStage 2.

Quantitative susceptibility mapping (QSM) has the potential for being a biomarker for various diseases because of its ability to measure tissue susceptibility related to iron deposition, myelin, and hemorrhage from the phase signal of a T2*‐weighted MRI. Despite its promise as a quantitative marker, QSM is faced with many challenges, including its dependence on preprocessing of the raw phase data, the relatively weak tissue signal, and the inherently ill posed relationship between the magnetic dipole and measured phase. The goal of this study was to evaluate the effects of background field removal and dipole inversion algorithms on noise characteristics, image uniformity, and structural contrast for cerebral microbleed (CMB) quantification at both 3T and 7T. We selected four widely used background phase removal and five dipole field inversion algorithms for QSM and applied them to volunteers and patients with CMBs, who were scanned at two different field strengths, with ground truth QSM reference calculated using multiple orientation scans. 7T MRI provided QSM images with lower noise than did 3T MRI. QSIP and VSHARP + iLSQR achieved the highest white matter homogeneity and vein contrast, with QSIP also providing the highest CMB contrast. Compared with ground truth COSMOS QSM images, overall good correlations between susceptibility values of dipole inversion algorithms and the COSMOS reference were observed in basal ganglia regions, with VSHARP + iLSQR achieving the susceptibility values most similar to COSMOS across all regions. This study can provide guidance for selecting the most appropriate QSM processing pipeline based on the application of interest and scanner field strength.