A new correction approach of partial volume effects in functional MRI for orange computing

Abstract

Functional magnetic resonance images are widely known as an effective scanning tool in both clinical diagnosis and academic studies, to investigate soft tissues of specific regions in human beings, and it receives vast popularity because of its prominent merits in zero radiation, high spatial resolution and affordable scanning price. However, several critical issues still need to be carefully considered and tackled before acquired raw functional images are fed into the post-processing cycle, and the correction problem of partial volume effects is one of them. In this study, one special imaging modality of functional images, i.e., the arterial spin labeling, is emphasized, and its challenging correction problem of partial volume effects is to be solved. Significantly different from several contemporary correction studies in arterial spin labeling images, which mainly rely on neighboring pixels for adding additional information to facilitate the correction procedure, the new approach solely counts on the single pixel itself for its own correction problem. The superiority of the new approach can be suggested by its more clear corrected image outcomes without much blurring and significant tissues information loss, which are commonly witnessed in contemporary correction approaches for arterial spin labeling. Experiments based on a database composed of 360 demented patients and comprehensive analysis from the statistical perspective also substantiates that.