Cerebral microbleeds (CMBs) are established imaging biomarkers of cerebral small vessel disease and are a defining feature of cerebral amyloid angiopathy (CAA), yet their automated detection in susceptibility-weighted imaging (SWI) remains challenging due to high false-positive rates from vessel cross-sections, iron and calcium deposits, and other hypointense mimics. We present a fully automated, three-stage cascade pipeline that combines subject-adaptive unsupervised candidate generation with two successive lightweight 3D ResNet classifiers, trained with only human-in-the-loop quality-assurance (QA) labels (yes/no per candidate) rather than dense voxel-wise segmentation masks. The candidate generation stage is performed by fitting a Gaussian Mixture Model (GMM) to each subject’s SWI intensity histogram to define an adaptive low-intensity threshold, followed by anatomical masking to exclude physiologically irrelevant regions (image edges, ventricles/CSF/choroid plexus, and cerebellum), and filters candidates by size and sphericity. The model was trained and evaluated across a nested 3×5-fold cross-validation on N = 30 subjects from a publicly available labeled microbleed dataset and a CAA cohort (11,424 CMB candidate lesions) with data augmentation during training. Stage A classifies all geometric candidates as CMB or non-CMB and Stage B refines the predicted positives to suppress false positives (cascade AUC = 0.9587, sensitivity = 0.712, specificity = 0.975, PPV = 0.676, F1 = 0.693). The cascade reduces Stage A false positives by 76.8% (888/1,157 false positives eliminated) while retaining competitive sensitivity. Inference was performed on 141 SWI scans, detecting a mean 40.3 CMBs per scan and being preferred for use in 85% of high CMB cases, as evaluated by a blinded neurologist. The inference pipeline outputs binary CMB segmentation NIfTI images and radiologist-ready QA visualizations.
Sam Bogdanov, G. Rudravaram, Adam M. Saunders et al.· bioRxiv· 0 citations
Nonlinear gradients alter the diffusion encoding in brain diffusion tensor imaging (DTI), leading to spatially varying diffusion weighting which bias quantitative measures if uncorrected. Although the overall effects of gradient nonlinearity correction in brain studies are typically minimal and often fall below the detection limits of traditional imaging resolutions and sensitivities, their cumulative impact on clinical outcomes requires further study. This study investigates the significance and effects of correcting gradient nonlinearity in DW-MRI, focusing on the microstructural and macrostructural changes in white matter (WM) and gray matter (GM) across a clinical cohort. Our primary aim is to clarify whether the observed nonlinearity significantly alters the interpretation of aging in clinical settings, particularly in studies comparing healthy individuals to those with neurological conditions. We assess the extent of nonlinear fields impact on individual scans, interscanner observations, and a tract-based analysis. Using data from the Vanderbilt Memory & Aging Project (n = 948 imaging sessions, 933 on Scanner B and 15 on Scanner A acquired with single-shell diffusion tensor imaging protocol), we find 1%, 3.3%, and 5-degree changes in microstructure measures, fractional anisotropy (FA), mean diffusivity (MD), and primary eigen vector (V1) respectively, affecting at least 20% of the brain. Across sessions, head positioning sampled typical clinical variability, with head offsets of approximately 0–10 mm and rotations of 0–10° relative to magnet isocenter. Subcortical regions in the superior regions, occipital lobules, and parietal lobules exhibit relatively higher impacts. Macrostructural measures show changes up to 12% after nonlinear field correction. GNL effects are 5% and 0.33% of FA and MD changes between mild cognitive impairment and controls. A simple power analysis indicates that these subtle effects of gradient nonlinearity correction can become statistically detectable in larger multi-site studies exceeding ~1000 subjects, suggesting that GNL should be considered and, where possible, corrected or at least quantified in such settings.
Praitayini Kanakaraj, Tianyuan Yao, Zhiyuan Li et al.· PLoS ONE· 0 citations
