AUTHOR=Deng Xianyu , Tian Lei , Zhang Yinghuai , Li Ao , Cai Shangyu , Zhou Yongjin , Jie Ying 

TITLE=Is histogram manipulation always beneficial when trying to improve model performance across devices? Experiments using a Meibomian gland segmentation model

JOURNAL=Frontiers in Cell and Developmental Biology

VOLUME=Volume 10 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/cell-and-developmental-biology/articles/10.3389/fcell.2022.1067914

DOI=10.3389/fcell.2022.1067914

ISSN=2296-634X

ABSTRACT=Meibomian gland dysfunction (MGD) is caused by abnormalities of the meibomian glands (MG) and is one of the causes of evaporative dry eye disease (DED). Precise MG segmentation is crucial for MGD-related DED diagnosis because the morphological parameters of these glands are of importance. Deep learning has achieved state-of-the-art performance in medical image segmentation tasks, especially when training and test data come from the same distribution. However, in practice, MG images can be acquired from different devices or hospitals. When testing image data from different distributions, deep learning models that have been trained on a specific distribution may show poor performance. Histogram specification (HS) has been reported to be an effective method for contrast enhancement and improving model performance on images of different modalities. Additionally, contrast-limited adaptive histogram equalization (CLAHE) was used as a preprocessing method to enhance the contrast of MG images.
In this study, we developed and evaluated the automatic segmentation method of the eyelid and MG areas based on the convolutional neural network (CNN) and the automatically calculated MG atrophy rate. This method was evaluated in internal and external testing sets from two meibography devices. In addition, to assess whether HS and CLAHE improve segmentation results, we trained the network model using images from one device (internal testing set) and tested it on images from another device (external testing set). High DSC (0.84 for MG region, 0.92 for eyelid region) for the internal test set and lower DSC (0.69~0.71 for MG region, 0.89~0.91 for eyelid region) for the external testing set were obtained. Also, HS and CLAHE were reported to cause no statistically significant improvement in the segmentation results of MG in this experiment.