canonical correlation analysis  (Stiefel)

Journal: bioRxiv

Article Title: A cortical signature of very preterm birth across development and its association with neurodevelopmental outcomes

doi: 10.1101/2025.10.17.683086

Figure Lengend Snippet: a) the significant canonical correlation mode identified by sparse canonical correlation analysis (sCCA) between the VPT cortical signature and clinical risk factors. b) clinical loadings on the significant correlation mode. GA: gestational age; BW: birth weight; VENT: duration of mechanical ventilation; CPAP: duration of continuous positive airway pressure; TPN: duration of total parenteral nutrition. c) cortical loadings on the significant correlation mode. d) associations between the brain score and neurodevelopmental outcomes at 20 months.

Article Snippet: Third, in a fourth infant cohort of 414 VPT neonates who were recruited into the Evaluation of Preterm Imaging Study (ePrime), we examined the relationships between cortical alterations and neonatal clinical factors using sparse canonical correlation analysis (sCCA), which is a multivariate method for identifying maximal relationships between two sets of data, particularly in high-dimensional neuroimaging data, using regularization to achieve sparsity [ – ].

Techniques:

Journal: bioRxiv

Article Title: Neural subspaces of imagined movements in parietal cortex remain stable over several years in humans

doi: 10.1101/2023.07.05.547767

Figure Lengend Snippet: Methods. A) The neural signal was recorded on all days of the task using the same microelectrode arrays (Blackrock Neurotech). Arrays were implanted in the Anterior Intraparietal Area (AIP) and Brodmann Area 5 (BA5) of participant 1, and the junction of the postcentral and intraparietal sulcus (PC-IP) in participant 2. B) Data were arranged in a tensor of multi-unit activity (MUA) on each channel, of each trial completed during the duration of the study. All trials within a day were grouped and a pair of days was selected for further analysis (e.g., day N or M). For each day, the latent neural data was calculated using principal component analysis (PCA). The latent data were aligned by canonical correlation analysis (CCA), and the magnitude of the alignment was calculated as the principal subspace angle (PSA).

Article Snippet: The latent data from each day in the pair was then aligned using Canonical Correlations Analysis (CCA) (‘canoncorr’ function, Matlab 2021b).

Techniques: Activity Assay