Artificial Intelligence Diagnosis And Related Clinical Study Of Lens Grading Of Age-related Cataract

Objective: To establish an artificial intelligence diagnosis syst em of lens grading of age-related cataract;To study the association betw een the cortical and nuclear opacity of cataract with age,gender,high m yopia and diabetes mellitus.Methods: A total of patients(>50 years ol d,938 eyes)with cataract were enrolled in Sichuan Provincial People’s H ospital from January 2016 to August 2016,which included 326 eyes of ca taract with high myopia and 235 eyes of cataract with diabetes.Images by Slit-lamp anterior segment camera system were collected,and then the Lens Opacities Classification System(LOCS III,proposed by Chlylack et al.,1993,USA)was used to classify the type and grade of lens op acity in each eye.After the cataract images are processed by feature extr action,etc.Inputting the ture degree of cataract into the computer for lear ning.Finally an artificial intelligence diagnosis system for distinguishing the degree of cataract is established.Conducting a retrospective study bas ed on case in the hospital,and analyzing statistically the data that the i mages of cataract are graded,and then to find the association between th e cortical and nuclear opacity of cataract with age,gender,high myopia and diabetes mellitus by rank sum test and multivariate regression.Resul ts: At present,we have almost completed the establishment of an artifici al intelligence diagnosis system of lens grading of nuclear cataract and c ortical cataract.The N3 and N4 are the best prediction results,there are a few abnormal values exist.The error analysis of the grade of nuclear cataract showed that the correct rate was 69.7%,and the ratio of error w hich not exceeding one grade was 96.7%.The C1 was the best prediction results,there are a few abnormal values exist.The error analysis of the grade of cortical cataract showed that the correct rate was 48%,and the ratio of error which not exceeding one grade was 76.6%.The clinical r esearch results showed that the grade of cortical opacity in patients with high myopic was significantly lower than that in patients without high m yopia(P<0.001),and the grade of nuclear opacity is higher(P=0.001)in pati ents with high myopic.The grade of nuclear opacity in diabetic patients(P=0.003)and the elderly patients(P<0.001)was significantly higher.Men have higher grades of nuclear opacity(P=0.024)and cortical opacity(P<0.001)than women.Conclusion: The system can reduce the threshold of cataract screening and improve the overall prevention level of cataract.However,The diagno ses are one-sided,and its accuracy needs to be improved.The combinatio n of artificial intelligence and doctors is the development direction of me dical in the future.The important factors on nuclear cataract include age,diabetes mellitus,high myopia and gender;The development of nuclear c ataract can be effectively delayed by blood sugar control and myopia pre vention.Both in cortical cataract and nuclear cataract,the development of lens opacity may be faster in men than in women.