A New Approach to Spatial Covariance Modeling of Functional Brain Imaging Data: Ordinal Trend Analysis

Habeck, Christian; Krakauer, John W.; Ghez, Claude; Sackeïm, Harold A.; Eidelberg, David; Stern, Yaakov; Moeller, James R. · 2005 · OpenAlex-citations

DOI: 10.1162/0899766053723023

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper introduces Ordinal Trend Analysis (OrT), a novel spatial covariance modeling technique designed to identify sustained patterns of regional functional connectivity in neuroimaging data. The authors address a limitation in existing network analyses, such as Partial Least Squares (PLS), which often struggle to detect connectivity patterns that remain constant while their expression scales monotonically with increasing task difficulty. OrT is specifically engineered to recover latent activation patterns that increase in expression as experimental task parameters increase, while maintaining stable correlative relationships between brain regions. This approach allows for the testing of cognitive theories that predict sustained functional connectivity across parametric changes, such as increasing memory load, without requiring a priori quantitative models of brain-behavior relationships. The methodology involves a five-step computational algorithm applied to neuroimaging data matrices. First, a projection operator eliminates task-independent effects. Second, a unique high-dimensional design matrix is applied to enhance the salience of patterns expressing ordinal trends, differentiating them from patterns with inconsistent subject trends or non-monotonic changes. Third, singular value decomposition (equivalent to Principal Component Analysis) is performed on the transformed data. Fourth, the resulting eigenimages are tested for the presence of an ordinal trend using linear regression and a "number-of-exceptions" criterion to assess statistical significance. Finally, bootstrap resampling estimates the reliability of voxel weights using the inverse coefficient of variation. The authors validated the feasibility and statistical specificity of OrT using Monte Carlo simulations, demonstrating that it outperforms conventional low-dimensional Canonical Variates Analysis and standard PCA in recovering target ordinal patterns while maintaining low Type I error rates. The paper demonstrates the application of OrT to two datasets: an event-related fMRI study of verbal working memory and an H2 15 O-PET study of visuomotor learning. In the fMRI study, eighteen subjects performed a delayed-match-to-sample task with varying memory loads (one, three, or six letters). OrT successfully identified activation patterns that expressed positive ordinal trends on a subject-by-subject basis, consistent with theoretical predictions of sustained connectivity in working memory networks. The analysis provided estimates of pattern expression for each subject and task condition, quantifying the statistical significance of these trends. The significance of this work lies in providing a robust statistical tool for testing the assumption of sustained functional connectivity in cognitive neuroscience. By focusing on within-group models that identify patterns scaling with task parameters, OrT complements existing methods like PLS. The authors conclude that OrT has broad potential applications, extending beyond studies of young adults to include research on normal aging and neurological or psychiatric diseases, where understanding how functional connectivity scales with cognitive demand is critical.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• cognitive capacity variation