Automatic Tongue Region Segmentation from head MRI to aid for SAS Diagnosis

Abstract

This research proposes the automatic segmentation method of tongue region from head MRI to aid for Sleep Apnea Syndrome( SAS ) diagnose. This method calculates five feature values from the image which is trimmed around tongue from head MRI, and then that values are learned by Support Vector Machine( SVM ). As a result, accuracy rate which distinguished five images is about 55.7%, and the highest accuracy rate is about 74.08%. In order to improve accuracy, We utilize combined feature values, seek other efficient feature values in future.