Segmentation of diploë in a T1 weighted 3D MR image of human head

This paper presents the method to automatically segment the diploë in a T1 weighted three dimensional magnetic resonance image of human head. The method performs segmentation using a series of mathematical operations. A sphere is generated for interpolation of manually marked control points. The inner and outer boundaries of diploë are determined using thresholding and the deviated points are corrected via smoothing algorithm. As manual segmentation takes 6-8 hrs, this automatic tool will speed up the procedure manifolds and bring the segmentation time to 20 minutes. The results of automatic segmentation of ten different MR scans were compared against manual segmentation using different similarity measures. Dice coefficients´ values range between 97 percent to 99 percent and sensitivity and positive predictive values are more than 99 percent which shows very good segmentation.