Abstract

Magnetoencephalography (MEG) neuroimaging technique has provided a non-invasive tool for facilitating early detection, diagnosis and treatment of brain conditions. A key limitation of current diagnostic tests such as the Crawford-Garthwaite test and the Anderson-Darling test used in neuroscience is the assumption that case-control groups consist of normally distributed and homogenous subjects. This makes these tests not efficient in the single-case study when a case is compared to non-normally distributed and heterogeneity subjects. To meet this challenge, we propose a flexible mixture-model-based likelihood procedure for predictive testing for hidden brain injury with MEG data.