| Objective To explore the basic parameters which affected the myopic control effection of night-wear orthokeratology(OK)lenses by colecting clinical datus,and establish an artificial intelligence model to predict the efficacy of OK lenses.Method 1.A retrospective study to find the basic parameters affecting the efficacy of OK lens myopia prevention and control.A total of 110 adolescents(210 eyes)who wore OK lens at night were included.According to the different AL growth,it is divided into:low growth group(106 eyes)and high growth group(104 eyes).The differences of parameters before OK lens shaping were compared between the two groups(OK lens brand,patient gender,eye gender,age,spherical lens,cylinder lens,cylinder axis position,SE,Ks,Kf,Es,Em).Differences in parameters of OK lenses after shaping(the size of the optical center and the difference between the nasal,temporal,superior and inferior curvatures and the central point of the cornea with a diameter of 3mm,5mm,and 7mm from the center of the cornea)were compared.The composition ratios of all OK lenses wearing eyes,whether the defocus ring is complete or not after shaping,and the difference in 1-year AL growth are compared;for those with complete defocus rings,the 1-year AL in different horizontal and vertical deviation directions is described and counted respectively.difference in growth.Do a correlation analysis at the same time.To verify whether some meaningful parameters have exact effects on the curative effect of OK lens,the second part of the prospective study was conducted.2.Prospective study:validating the effective impact parameters of the first part.In the first part of the study,the parameters(age,SE,Em)that were found to be statistically significant before wearing OK lenses after statistical analysis were selected,and they were grouped according to the mean of the AL low growth group.A total of 124 adolescents with the same brand OK were included,with124 eyes(right eye),and were divided into high age group(Y-1),low age group(Y-2)and high SE group(SE-1 Group),low SE group(SE-2 group),high Em group(Em-1 group),and low Em group(Em-2 group),to observe and compare the growth of AL in the high and low groups after 1 year later.3.The relevant data of prospective and retrospective studies on OK mirrors(334cases)were aggregated into a total data set,and the total data set was randomly divided into training set,test set and validated set according to the ratio of 6:2:2set.Among them,the training set and test set completed the establishment of the four-layer artificial neural network(ANNs)model at the same time,to evaluate the speed of the eye axis growth controlled by the OK lens,and the validation set was used to verify the model.At the same time,the established AI model is compared with the logistic regression model.Result 1.Annual growth of low growth group was 0.04±0.12mm;high group was 0.39±0.16mm,with the significant difference(p<0.001).For OK lens wearers,pre-shaping parameters:low growth age(12.58±2.35 years)was higher than high growth(10.18±1.77 years),low growth SE(-3.56±1.53D)less than high growth group(-2.83±1.23D),and low growth Em(0.53±0.84)less than high growth(0.56±0.81).Pearson’s correlation analysis showed the age high negative correlation(r=-0.63),SE middle positive correlation(r=0.31),Em low positive correlation(r=0.23),all were statistically significant.Results of logistic regression analysis of pre-shaping parameters:age correlation 0.54(0.44-0.65)and SE correlation 1.72(1.33-2.21),all were statistically significant.After shaping parameters:N3(1.32±2.45D),T5(0.34±1.52D),N5(3.03±1.90D)and S5(3.93±1.84D)were significantly higher than N3(0.08±0.89D),T5(-0.06±0.62D),N5(2.12±1.35D),S5(2.95±1.36D)in the low growth group than high growth group.In 210 The eyes,intact in 187 and incomplete in 23eyes,showed no statistically significant difference in annual AL growth between the two groups.The 187 eyes with intact defocus rings,biased 0.64±0.37mm,were negatively associated with lower annual AL growth,r=-0.16,p=0.03.After shaping,the horizontal deviation amount(-0.45±0.44mm)is greater than the vertical direction(-0.13±0.36mm).Inferior vertical deviation(0.17±0.20mm)and middle(0.24±0.27mm)were lower than superior(0.31±0.17mm).There was no statistical difference in the annual growth amount of AL lost in different directions and biased in different directions.The average age of 12 years,SE mean-3.50D and Em mean 0.53 were used for the prospective study group limits in Part 2 to further explore the influence of these three factors on the annual growth of AL.2.Prospective studies show the AL growth(0.04±0.26mm)in Y-1 group was significantly lower than that in Y-2 group(0.22±0.23mm),(p<0.001)after 1 year of wearing the OK lens.There were no significant differences in other baseline characters between two groups.The AL growth(0.02±0.23mm)in SE-1 group was significantly lower than in Y-2 group(0.20±0.25mm),which p value lower than 0.001.There were no significant differences in other baseline characters between two groups.There was no statistical difference between AL growth(0.12±0.28mm)between Em-1 and Em-2(0.15±0.25mm)(p=0.51).3.ANNs prediction model,the sensitivity was 87.50%,the specificity was86.21%,the Kappa value was 0.73(χ~2=37.16,p<0.001),and the area under the curve(AUC)was 0.87.which is higher than logistic model(sensitivity was62.91%,the specificity was 72.13%,the Kappa value was 0.35(χ~2=41.30,p<0.001),and the AUC was 0.76).Conclusion Myopic adolescents wear night-wearing OK lenses,while older adolescents(>12years)and adolescents with slightly higher myopia(<-3.50D)wear OK lenses,which can better control myopia,which may be related to the increase in the curvature difference in the 5mm area of the cornea and 3mm nasal area after shaping.Wearing night-wearing OK lenses,a small number of cases have an incomplete defocus ring,which is has little effect on the myopia prevention.The horizontal deviation of the lens is often greater than the vertical deviation,and it is more common in the temporal side in the horizontal direction and down in the vertical direction.During the shaping process,and the slight horizontal deviation has little effect on the myopia prevention and control effect of the OK lens.In the vertical direction,the upward deviation should be avoided as much as possible,which is conducive to better prevention and control of myopia.The ANNs model based on the deep learning of the multi-neural network system can better predict the myopia prevention effect of the OK lens,and is expected to become a new tool for the development of the OK lens fitting work to provide efficacy prediction. |
PDF Full Text Request