Multi-Feature Algorithm Based Meibomian Gland Dysfunction Assisted Grading System

Meibomian gland dysfunction is a chronic and diffuse common disease,but it has not reached a global consensus on classification and diagnosis.The detection rate of meibomian gland dysfunction in China is 6.1 %-52.4 %.However,at present,this disease in China still adopts the traditional diagnostic method of human observation.This method brings a lot of repetitive work to the primary screening doctors,and is highly dependent on the subjective judgment of professional doctors.It is inefficient,and it is difficult for a large group of patients to be diagnosed quickly and effectively.Therefore,it is urgent to develop an auxiliary grading system for meibomian gland dysfunction.With the development of deep learning and image processing technology,convolutional neural network has made great progress.In this paper,U-Net network is used to train the image of eyelid and meibomian gland of patients,and then the biological parameters of meibomian gland are extracted.The multi-feature auxiliary grading standard based on the area loss rate and bending curvature of meibomian gland is confirmed,and the auxiliary grading system of meibomian gland dysfunction based on multi-feature algorithm is built.The main research contents and innovations of this paper include :1、The eyelid and meibomian gland of the patient image were extracted based on the U-Net network.The patient images of the provincial hospital were sorted and classified,and the data set was enhanced,including balancing the brightness of the image,weakening the background and enhancing the gland area.The U-Net network was used to train the data set and extract the eyelid and meibomian gland areas.2、The auxiliary classification of patients with meibomian gland dysfunction based on multi-feature algorithm.The biological parameters of eyelid and meibomian gland were extracted,and then the area loss rate and bending curvature of meibomian gland were extracted to confirm the multi-feature auxiliary grading standard characterized by the area loss rate and bending curvature of meibomian gland.3、The auxiliary grading system of meibomian gland dysfunction was built based on multi-feature algorithm.Through systematic research,improve the multi-feature algorithm of meibomian gland dysfunction auxiliary grading,and build a patientassisted grading system.In this paper,based on the area and curvature characteristics,the multi-feature auxiliary grading algorithm of meibomian gland dysfunction is determined.Based on the multi-feature algorithm,the auxiliary grading system of meibomian gland dysfunction is developed to realize the automatic auxiliary grading of patients,reduce the workload of doctors ’ manual analysis of images,increase objectivity,reduce the influence of subjective factors,and improve the rate of diagnosis of meibomian gland dysfunction.