'Pscore' - A Novel Percentile-Based Metric to Accurately Assess Individual Deviations in Non-Gaussian Distributions of Quantitative MRI Metrics.

BACKGROUND:Quantitative MRI metrics could be used in personalized medicine to assess individuals against normative distributions. However, conventional Zscore analysis is inadequate for measurements with non-Gaussian distributions. PURPOSE or HYPOTHESIS:Demonstrate systematic skewness in diffusion MRI (dMRI) metrics. Propose a novel percentile-based method, 'Pscore' to address this and document its performance on a publicly available dataset. STUDY TYPE:Cohort. POPULATION:961 healthy young adults, the Human Connectome Project (HCP). FIELDSTRENGTH/SEQUENCE:Siemens 3T 'Connectome Skyra' scanner, spin-echo diffusion echo planar imaging (EPI). ASSESSMENT:The dMRI data was preprocessed using the TORTOISE pipeline. Average values within 48 regions of interest (ROIs) were computed from various diffusion tensor (DT) and mean apparent propagator (MAP) metrics. For each ROI, percentile ranks across participants were first computed to generate 'Pscores'- which normalize the difference between the median and a participant's value with the corresponding difference between the median and the 5th/95th percentile values. STATISTICAL TESTS:ROIwise distributions were assessed using 'Log'-transformation, 'Zscore' and the 'Pscore' methods. The distributions and percentages of extreme values (>95th and <5th percentile boundaries) were also compared across all ROIs comprising the overall white matter. Bootstrapping was performed to test the reliability of Pscores in small samples (N=100) using 100 iterations. RESULTS:The dMRI metrics demonstrated systematic skewness, leading to skewed 'Log'-transform and 'Zscore' distributions. Zscores showed extreme value biases, which were strongest for the Propagator Anisotropy. 'Pscore' distributions were symmetric and robustly maintained 5% extreme values in both tails, even for 100 iterations in small, bootstrapped samples. DATA CONCLUSION:The inherent skewness observed for dMRI metrics preclude the use of conventional Zscore analysis. The proposed 'Pscore' method accurately estimates individual deviations in skewed normative data. Although the HCP dMRI data was showcased, Pscores offer a general solution, even for smaller databases with non-Gaussian distributed values of neuroimaging and clinical measurements.