Denoising Improves Cross‐Scanner and Cross‐Protocol Test–Retest Reproducibility of Diffusion Tensor and Kurtosis Imaging

• Published in: Human brain mapping Vol. 46; no. 4; pp. e70142 - n/a
• Main Authors: Ades‐Aron, Benjamin, Coelho, Santiago, Lemberskiy, Gregory, Veraart, Jelle, Baete, Steven H., Shepherd, Timothy M., Novikov, Dmitry S., Fieremans, Els
• Format: Journal Article
• Language: English
• Published: Hoboken, USA John Wiley & Sons, Inc 01.03.2025
• Subjects: Adult; Alzheimer's disease; Bias; Brain - diagnostic imaging; brain white matter; clinical translation; Cognitive development; Datasets; diffusion kurtosis imaging; Diffusion Magnetic Resonance Imaging - methods; Diffusion Magnetic Resonance Imaging - standards; diffusion tensor imaging; Diffusion Tensor Imaging - methods; Diffusion Tensor Imaging - standards; Evaluation; Female; higher‐order diffusion; Humans; image denoising; Image processing; Image Processing, Computer-Assisted - methods; Image Processing, Computer-Assisted - standards; Imaging techniques; Kurtosis; Magnetic resonance imaging; Male; Medical imaging; Middle Aged; MPPCA; Neural networks; Neuroimaging; Neuroimaging - methods; Neuroimaging - standards; NMR; Noise reduction; Nuclear magnetic resonance; Outliers (statistics); Population studies; Principal components analysis; Reproducibility; Reproducibility of Results; Scanners; Tensors; test–retest reproducibility; Variability; White Matter - diagnostic imaging; Young Adult; clinical translation; test–retest reproducibility; higher‐order diffusion; MPPCA; diffusion kurtosis imaging; diffusion tensor imaging; brain white matter; image denoising
• ISSN: 1065-9471; 1097-0193; 1097-0193
• DOI: 10.1002/hbm.70142

ABSTRACT The clinical translation of diffusion magnetic resonance imaging (dMRI)‐derived quantitative contrasts hinges on robust reproducibility, minimizing both same‐scanner and cross‐scanner variability. As multi‐site data sets, including multi‐shell dMRI, expand in scope, enhancing reproducibility across variable MRI systems and MRI protocols becomes crucial. This study evaluates the reproducibility of diffusion kurtosis imaging (DKI) metrics (beyond conventional diffusion tensor imaging (DTI)), at the voxel and region‐of‐interest (ROI) levels on magnitude and complex‐valued dMRI data, using denoising with and without harmonization. We compared same‐scanner, cross‐scanner, and cross‐protocol variability for a multi‐shell dMRI protocol (2‐mm isotropic resolution, b = 0, 1000, 2000 s/mm2) in 20 subjects. We first evaluated the effectiveness of Marchenko‐Pastur Principal Component Analysis (MPPCA) based denoising strategies for both magnitude and complex data to mitigate noise‐induced bias and variance, to improve dMRI parametric maps and reproducibility. Next, we examined the impact of denoising under different population analysis approaches, specifically comparing voxel‐wise versus region of interest (ROI)‐based methods. We also evaluated the role of denoising when harmonizing dMRI across scanners and protocols. The results indicate that DTI and DKI maps visually improve after MPPCA denoising, with noticeably fewer outliers in kurtosis maps. Denoising, either using magnitude or complex dMRI, enhances voxel‐wise reproducibility, with test–retest variability of kurtosis indices reduced from 15%–20% without denoising to 5%–10% after denoising. Complex dMRI denoising reduces the noise floor by up to 60%. Denoising not only reduced variability across scans and protocols, but also increased statistical power for low SNR voxel‐wise comparisons when comparing cross sectional groups. In conclusion, MPPCA denoising, either over magnitude or complex dMRI data, enhances the reproducibility and precision of higher‐order diffusion metrics across same‐scanner, cross‐scanner, and cross‐protocol assessments. The enhancement in data quality and precision facilitates the broader application and acceptance of these advanced imaging techniques in both clinical practice and large‐scale neuroimaging studies. MPPCA denoising enhances the reproducibility and precision of higher‐order diffusion metrics in dMRI, by reducing variability and noise across same‐scanner, cross‐scanner, and cross‐protocol assessments. This improvement supports broader clinical application and acceptance of advanced imaging techniques.