Automatic Segmentation of Human Brain, Grey and White Matter in MRI: A Robust and Accurate Algorithm Based on the Tissues' Features Analysis

Introduction The recognition and segmentation of brain tissues is usually performed with manual or semiautomatic softwares. Semiautomatic softwares have an overall accuracy of about 9% compared with results obtained through manual segmentation, while fully automatic softwares actually provide inferior accuracies [1]. Moreover, we noticed that most of the commercial softwares do not offer built-in evaluations of the precision of the obtained results. All in all, this is very time consuming and implies high subjectivity of the results. These remarks emphasize the usefulness of an algorithm for automatic analysis of brain tissues, specifically for grey and white matter segmentation from raw images, without the introduction of subjective parameters and able to give informations about the precision of the results.