Study On Eye Movement As A Biomarker For Amblyopia Diagnosis

Amblyopia is one of the main causes of visual impairment in modern people,and it s incidence can reach 4%.The key age for amblyopia treatment is before the age of 6-8.Early diagnosis and early treatment are of great significance for the control of amblyo pia.The current diagnostic methods for amblyopia have problems such as high require ments for patient cooperation,single indicators,and high detection rate.Young children have a certain risk of misdiagnosis.In addition,at present,the classification of amblyop ia is based on its risk factors which doesn’t work well to treat different types of amblyop ia.Targeting the defect of eye movement in amblyopia,eye tracker,as a noninvasive cognitive behavioral research device,not only requires a low degree of coopera tion,but also extract various eye movement indicators related to perception and cognitio n.There is good potential for research on eye movement defects.This study investigated whether eye movement indicators can be used as diagnostic biomarkers for amblyopia and discussed its value in different treatment of amblyopia.T he paper consist of the following topics: 1)A gaze experiment and a saccade experiment were designed,and the eye movement data of various types of amblyopia and healthy c ontrols under different viewing conditions were collected;2)The eye movement data w ere calculated according to the eye movement defect of amblyopia.Eye movement para meters such as bivariate contour ellipse（BCEA）,velocity sequence fuzzy entropy,and s accade initiation time in the fixation region were investigated,and their contributions to the diagnosis of amblyopia were discussed.The results showed that parameters such as BCEA were significantly varied among different types of amblyopia and treatment diffi culty.So is fuzzy entropy among different severity levels 3)Combining the above three indicators and various eye movement data time series features as input,the machine lear ning algorithm was used to classify and measure tasks such as amblyopia,amblyopia ty pe,severity,and treatment difficulty,and found that except for the severity prediction ta sk,the Support Vector Machine（SVM）and logistic regression models both achieved ov er 95% prediction accuracy in each task.In conclusion,this study verifies the value of three quantitative indicators,BCEA,velocity sequence fuzzy entropy,and saccade initiation time,in distinguishing different types of amblyopia and the difficulty of treatment,and combines various eye movement features in various amblyopia classification tasks.Good prediction results were obtaine d,which proved that eye movement indicators can be used as biomarkers for the diagno sis of amblyopia.