Quality assurance (QA) in magnetic resonance (MR) imaging is critical but remains a challenging and time-intensive process, particularly when working with large-scale, multi-site imaging datasets. Manual QA methods are subjective, prone to inter-rater variability, and impractical for high-throughput workflows. Existing automated QA methods often lack generalizability to diverse datasets or fail to provide interpretable insights into the causes of poor image quality. To address these limitations, we introduce an unsupervised and interpretable QA framework for multi-contrast MR images that quantifies artifact severity. By assigning a numerical score to each image, our method enables objective, consistent evaluation of image quality and highlights specific levels of artifact presence that can impair downstream analysis. Our framework employs an unsupervised contrastive learning approach, leveraging simulated artifact transformations, including random bias, noise, anisotropy, and ghosting, to train the model without requiring manual labels or preprocessing. A margin-based contrastive loss further enables differentiation between varying levels of artifact severity. We validate our framework using simulated artifacts on a public dataset and real artifacts on a private clinical dataset, demonstrating its robustness and generalizability for automatic MR image QA. By efficiently evaluating image quality and identifying artifacts prior to data processing, our approach streamlines QA workflows and enhances the reliability of subsequent analyses in both research and clinical settings.
Savannah P. Hays, Lianrui Zuo, Blake E. Dewey et al.· Proceedings of machine learn...· 2 citations
Background: Changes in ischemic white matter hyperintensities volume (WMH) on MRI over time are associated with cognitive decline. We investigated whether changes in WMH volume over time exhibit threshold effects of normalized WMH volume on declining cognitive performance and whether these effects on cognition differ between deep white matter hyperintensities (DWMH) and periventricular white matter hyperintensities (PVWMH). Methods: We followed 339 participants longitudinally from GeneSTAR with brain MRI and neuropsychological testing at baseline (2009–2013) and at 13-year follow-up (2023–present) (62% female, and 33% Black, mean baseline age 49.7±9.6). WMH were classified as PVWMH (within 2 mm of ventricles) or DWMH. Two-segment linear spline regression models using adjusted mixed linear regression identified test-specific thresholds longitudinally beyond which cognitive decline accelerated. Cognitive scores from both timepoints were treated as repeated measures, with WMH included as a time-varying predictor. Results: Declines in motor function and processing speed accelerated beyond thresholds of changing PVWMH and DWMH volumes. For Grooved Pegboard tests, changes in volume were associated with minimal effects below a threshold of changing volume (log-transformed ratio of lesion volume to intracranial volume for: PVWMH −9.42 to −9.29; and DWMH −11.8 to −11.7). Substantial declines in cognitive performance were observed above thresholds of increases in volume (slope differences: PVWMH; 14.5–15.1 seconds per log-unit, p < 0.001; and DWMH; 9.54–10.9, p < 0.001). Digit Symbol Substitution Test demonstrated paradoxical positive associations below changing volume thresholds (PVWMH; β=6.68, p=0.001 and DWMH; β=6.98, p < 0.001), reversing to decline above thresholds of increase in volume for PVWMH (Δβ=−11.2, p < 0.001) and DWMH (Δβ=−9.77, p < 0.001). Conclusion: Changes in WMH volume exhibit nonlinear threshold effects on changes in cognitive performance over time and differ by anatomic region. Minimal cognitive impact occurred below thresholds, with accelerated declines above. PVWMH demonstrate larger effects on declining cognitive function than DWMH, particularly for motor and processing speed functions and progress at a faster rate.
Sarvin Sasannia, M. Matsyuk, Shimeng Wang et al.· Stroke· 0 citations
