Segmentation of magnetic resonance images to construct human head model for diffuse optical imaging

The brain activation image obtained by diffuse optical tomography (DOT) is obtained by solving inverse problem using the spatial sensitivity profile (SSP). The SSP can be obtained from the analysis of the light propagation using three-dimensional head models. The head model is based upon segmented magnetic resonance (MR) image and there are several types of software based on binarization for segmentation of MR head images. We segmented superficial tissues which effect the light propagation in human head from MR images acquired with FATSAT and FIESTA pulse sequences by using region growing algorithm and morphological operation to facilitate the construction of the individual head models for DOT. The pixel intensity distribution of these images has appropriate characteristics to extract the superficial tissues by using algorithm based on binarization. The result of extraction was compared with the extraction from T2-weighted image which is commonly used to extract superficial tissues. The result of extraction from FATSAT or FIESTA image agree well with ground truth determined by manual segmentation.