Automatic segmentation of glands in infrared meibomian gland image

Abstract

Meibomian gland dysfunction (MGD) is the most common cause of dry eye disease. Ophthalmologists conduct qualitative evaluation of meibomian glands(MGs) of patients by observing infrared meibomian gland images. But it is subjective to make a diagnosis only with the naked eye. Automatic segmentation of MGs could be challenging and play a key role in MGD morphology analysis and diagnosis. In this paper, an automatic gland segmentation method based on UNet++ and a meibography image dataset are proposed. Data augmentation is used to expand training samples. Infrared meibomian gland images are fed into the preserved model for accurate segmentation. The experiments including comparison with the latest methods show that the presented method effectively segment the MGs and outperform other methods with an average accuracy of 94.28%.