Meibomian glands segmentation in infrared images with limited annotation

●AIM:To investigate a pioneering framework for the segmentation of meibomian glands(MGs),using limited annotations to reduce the workload on ophthalmologists and enhance the efficiency of clinical diagnosis.●METHODS:Totally 203 infrared meibomian images from 138 patients with dry eye disease,accompanied by corresponding annotations,were gathered for the study.A rectified scribble-supervised gland segmentation(RSSGS)model,incorporating temporal ensemble prediction,uncertainty estimation,and a transformation equivariance constraint,was introduced to address constraints imposed by limited supervision information inherent in scribble annotations.The viability and efficacy of the proposed model were assessed based on accuracy,intersection over union(IoU),and dice coefficient.●RESULTS:Using manual labels as the gold standard,RSSGS demonstrated outcomes with an accuracy of 93.54%,a dice coefficient of 78.02%,and an IoU of 64.18%.Notably,these performance metrics exceed the current weakly supervised state-of-the-art methods by 0.76%,2.06%,and 2.69%,respectively.Furthermore,despite achieving a substantial 80%reduction in annotation costs,it only lags behind fully annotated methods by 0.72%,1.51%,and 2.04%.●CONCLUSION:An innovative automatic segmentation model is developed for MGs in infrared eyelid images,using scribble annotation for training.This model maintains an exceptionally high level of segmentation accuracy while substantially reducing training costs.It holds substantial utility for calculating clinical parameters,thereby greatly enhancing the diagnostic efficiency of ophthalmologists in evaluating meibomian gland dysfunction.

基于深度学习的睑板腺腺体分割方法研究

目的:探讨应用深度学习技术解决睑板腺腺体自动分割问题的效果与价值。方法:采集并筛选出193幅红外睑板腺图像构建图像数据库,由3名临床医师对图像进行人工标记;引入UNet++网络与自动数据增广策略构建睑板腺腺体自动分割模型,采用精确率、敏感性、特异性、准确率和交并比分析该模型的可行性与有效性。结果:以人工标注结果为金标准,基于UNet++的睑板腺腺体自动分割模型取得94.31%的准确率,敏感性、特异性分别为82.15%和96.13%,腺体分割表现具有较好的稳定性,模型处理单张图像的平均用时仅为0.11s。结论:引入深度学习技术实现睑板腺腺体的自动分割,具有良好的准确性、稳定性和高效性,可服务于睑板腺功能障碍患者腺体形态参数的计算,辅助相关疾病的临床诊断和筛查,提高诊断效率。

Correlation of c-Myc and PIK3R3 expression with invasion and migration genes in melanoma tissue

Objective:To study the correlation of c-Myc and PIK3R3 expression with invasion and migration genes in melanoma tissue.Methods: Surgical removed melanoma tissue and normal skin tissue left in full-thickness skin grafting in 95 Clinical Department, Fuzhou General Hospital between May 2015 and March 2017 were selected, the RNA in the tissue was separated and extracted, and then fluorescence quantitative PCR kit was used to determine the mRNA expression of c-Myc and PIK3R3 as well as invasion and migration genes.Results:c-Myc, PIK3R3, CXCL12, CXCR4, CXCR7, NRP1, MMP2, MMP9 and CatK mRNA expression in melanoma tissue were significantly higher than those in normal skin tissue while E-cadherin, TIMP1 and KISS1 mRNA expression were significantly lower than those in normal skin tissue;c-Myc mRNA expression in melanoma tissue was positively correlated with PIK3R3 mRNA expression;CXCL12, CXCR4, CXCR7, NRP1, MMP2, MMP9 and CatK mRNA expression in melanoma tissue with higher c-Myc expression were significantly higher than those in melanoma tissue with lower c-Myc expression while E-cadherin, TIMP1 and KISS1 mRNA expression were significantly lower than those in melanoma tissue with lower c-Myc expression.Conclusion:The high expression of c-Myc and PIK3R3 in melanoma tissue can promote the invasion and migration of cancer cells.