Performance of canonical correlation analysis in language tests by functional MRI
(English)Manuscript (preprint) (Other academic)
In this paper a new method based on constraitwd canonical correlation analysis (CCA) for the analysis of fMRI data is evaluated. In particular the method benefits from a powerful way of choosing temporal basis functions in additiou to an adaptive spatial filtering scheme. A modified receiver operating characteristic (ROC) method was used to quantify the results and to compare it with traditionally used statistics in an objective way. The evaluation was performed using real fMRI data form a language test. It was shown that the CCA based method offers a significant gain in detection power.
Medical and Health Sciences
IdentifiersURN: urn:nbn:se:liu:diva-97715OAI: oai:DiVA.org:liu-97715DiVA: diva2:650285