arXiv subjects

Alexander Rauscher

Publications and source records attributed to Alexander Rauscher.

2 recordsLinked to original sources

arXiv2025

Dataset Properties Shape the Success of Neuroimaging-Based Patient Stratification: A Benchmarking Analysis Across Clustering Algorithms

Background: Data driven stratification of patients into biologically informed subtypes holds promise for precision neuropsychiatry, yet neuroimaging-based clustering methods often fail to generalize across cohorts. While algorithmic innovations have focused on model complexity, the role of underlying dataset characteristics remains underexplored. We hypothesized that cluster separation, size imbalance, noise, and the direction and magnitude of disease-related effects in the input data critically determine both within-algorithm accuracy and reproducibility. Methods: We evaluated 4 widely used stratification algorithms, HYDRA, SuStaIn, SmileGAN, and SurrealGAN, on a suite of synthetic brain-morphometry cohorts derived from the Human Connectome Project Young Adult dataset. Three global transformation patterns were applied to 600 pseudo-patients against 508 controls, followed by 4 within-dataset variations varying cluster count (k=2-6), overlap, and effect magnitude. Algorithm performance was quantified by accuracy in recovering the known ground-truth clusters. Results: Across 122 synthetic scenarios, data complexity consistently outweighed algorithm choice in predicting stratification success. Well-separated clusters yielded high accuracy for all methods, whereas overlapping, unequal-sized, or subtle effects reduced accuracy by up to 50%. SuStaIn could not scale beyond 17 features, HYDRA's accuracy varied unpredictably with data heterogeneity. SmileGAN and SurrealGAN maintained robust pattern detection but did not assign discrete cluster labels to individuals. Conclusions: The study results demonstrate the impact of statistical properties of input data across algorithms and highlight the need for using realistic dataset distributions when new algorithms are being developed and suggest greater focus on data-centric strategies that actively shape and standardize the input distributions.

cs.LG↗

arXiv2021

In vivo investigation of the multi-exponential T2 decay in human white matter at 7T: Implications for myelin water imaging at UHF

Multicomponent T2-mapping using a gradient and spin-echo (GraSE) acquisition has become standard for myelin water imaging at 3T. Higher magnetic field strengths promise SNR benefits but face specific absorption rate limits and shortened T2 times. This study investigates compartmental T2 times in vivo and addresses advantages and challenges of multi-component T2-mapping at 7T. We acquired 3D multi-echo GraSE data in seven healthy adults at 7T, with three subjects scanned also at 3T. Stimulated echoes arising from B+1 inhomogeneities were accounted for by the extended phase graph (EPG) algorithm. We used the computed T2 distributions to determine T2 times that identify different water pools and assessed signal-to-noise and fit-to-noise characteristics of the signal estimation. We compared short T2 fractions and T2 properties of the intermediate water pool at 3T and 7T. Flip angle mapping confirmed that EPG accurately determined the larger inhomogeneity at 7T. Multi-component T2 analysis demonstrated shortened T2 times at 7T compared to 3T. Fit-to-noise and signal-to-noise ratios were improved at 7T but depended on B1 homogeneity. Lowering the shortest T2 to 8 ms and adjusting the T2 threshold that separates different water compartments to 20 ms, yielded short T2 fractions at 7T that conformed to 3T data. Short T2 fractions in myelin-rich white matter regions were lower at 7T than at 3T, and higher in iron-rich structures. Adjusting the T2 compartment boundaries was required due to the shorter T2 relaxation times at 7T. Shorter echo spacing would better sample the fast decaying signal but would increase peripheral nerve stimulation. We used a multi-echo 3D-GraSE sequence to characterize the multi-exponential T2 decay at 7T. We adapted T2 parameters for evaluation of the short T2 fraction. Obtained 7T multicomponent T2-maps were in good agreement with 3T data.

physics.med-ph↗