fMRI study of Ultra-High Risk for Psychosis
fMRI study using SPM
- Pre-processed 55.6G structural and functional MRI data from 66 undergraduates using MATLAB packages (SPM and FreeSurfer).
- Used hypothesis testing (one sample t-test, paired t-test, etc.) to find statistically significant activated cerebral areas(Dorsolateral Prefrontal Lobe), serving as input features of binary classifiers.
- Used machine learning methods (SVM, Bayesian classifier, GBDT, and ResNet) to design models of a binary classifier, predicting whether a given MRI image is from a person at high risk of schizophrenia.