Functional magnetic resonance imaging (fMRI) is a prevalent technology in brain research of cognition,emotion,development,and brain disorders.The traditional fMRI analysis is based on vol-ume-based preprocessing pipelines and algorithms,which means that the brain MRI data is to be registered to a 3-dimensional(3D) coordinate [1].However,the relatively low spatial resolution of fMRI may lead to partial-volume-effect (e.g.,a 3D region may contain signals from grey matter,white matter and even cere-brospinal fluid).Given the human brain function is organized in a brain surface mesh manner,therefore,a growing number of stud-ies conducted surface-based preprocessing pipelines and algo-rithms.Surface-based methods reconstructed the brain grey matter into 2-dimensional cortical surface which better represent the curving structure of the brain.Surface-based method is supe-rior to volume-based method on brain registration,signal-noise ratio and reproducibility of algorithms [2].Specifically,the tradi-tional volume-based approach was reported with a spatial localiza-tion that is only 35% of the best surface-based method [2].