Automatic Extraction of High Texture Features used for Patient Realignment in Medical Imaging of the Head

In order to develop a technique for automatic patient realignment in medical imaging particularly in magnetic resonance imaging (MRI) of the brain it is essential to extract key features automatically from the various slices of the head with as great an accuracy as possible. The brain is a key feature which is vital for the accurate realignment of the patient in all the 3 planes (Sagittal, Transverse, Coronal). The MRI scans(Tl spin-lattice relaxation time) which produce fast images and are therefore used in patient realignment, provide highly textured images.These are difficult to segment using conventional thresholding or edge enhancement techniques due to their "grainy" appearance, making it difficult to isolate the brain from the other components found in the slice. In this paper a method is developed for the accurate extraction of the brain profile in the sagittal,transverse and coronal scans of the head by enhancing the original image to remove the high texture content.This results in a good working image which isolates the brain from other features, in a particular slice. The results obtained by implementing the algorithm on scans, taken of the same patient over a period of time, show the very high accuracy of the brain extraction algorithm.