Optimisation strategies for the registration of Computed Tomography images of electropalatography

Abstract Electropalatography is a technique that employs a custom-made artificial palate to measure the contact established between the tongue and the hard palate. This technique is widely used in treatment of articulation disorders and studies of speech. In order to evaluate the accuracy of the electropalate, the device needs to be separated from the volume that usually contains electropalate worn on hard palate. This is done with the use of segmentation techniques. Prior to the segmentation, the registration of the two volumes, one containing the electroplate worn on hard palate, and one containing only hard palate, needs to be done. The registration is a technique of aligning multiple images by geometrical transform. Over the years, many methods for registration have been developed. The following paper describes the method of registration based on sensitivity analysis. Sensitivity analysis is a technique that evaluates the change in the number of pixels with different intensity with a shift of the volumes in different dimensions. Then based on the found optimal shift value, the shift in different dimension of the matrix is made. The technique successfully improves the alignments between two data sets, reducing the number of non-matching pixels. The sensitivity analysis-based registration should be useful in the future improvement of image processing tools that are crucial for the medical imaging.