3D Modeling Of Nasal Tract On Magnetic Resonance Image

Magnetic resonance image is widely used in speech production for a long time because it does not harm the human body during scan.The subjects can continue to complete acoustic data collection after MRI scanning,so that morphological data and acoustic data are complete.However,the image quality is often criticized,especially in the study of nasal tract,which restricts the further research.There are some previous work on 3D modeling of vocal tract on MRI and 3D modeling of nasal tract on CT image.3D modeling of nasal tract on MRI is an innovative work.The experiment builds a database containing four sets of static MRI images of the nasal cavity.It is necessary to finish the image registration to make sure that the corresponding points of different image sets are on the same position.The paper uses the concept of mutual information,calculating the information redundancy between two images to register images.Threshold segmentation method finishes the nasal tract segmentation after defining the scope of the region of interest.Morphological correction reduces image noise.The evaluation of these three sets of images and manual operation is necessary after the output of result.The paper uses a semi-automatic method with algorithm and manual operation,together modeling the nasal tract in 3D space.The result of the morphological analysis is as expected as the common sense of anatomy.Further research could collect data on the MRI of the nasal tract and corresponding nasal sound to build model of whole vocal tract.The research on nasal could make progress.