An auxiliary analysis model for predicting corneal edema after phacoemulsification

Abstract

The present invention discloses an auxiliary analysis model for predicting corneal edema after phacoemulsification surgery for cataracts, which belongs to the field of medical technology. The model uses the best corrected visual acuity (BCVA) before phacoemulsification surgery and the cumulative released energy (CDE) during the phacoemulsification process during the surgery as indicators to construct a nomogram; wherein, the nomogram includes five line segments, each line segment is marked with a fixed scale, and the relative position of each line segment is fixed, and the five line segments are, from top to bottom, BCVA and CDE score line segments, BCVA value line segments, CDE value line segments, BCVA and CDE score total line segments, and corneal edema probability line segments; the prediction model of the present invention can complete relatively accurate predictions using only two common clinical data, has low data requirements for use, is convenient and practical, and has a high accuracy rate, can help ophthalmologists quickly and accurately judge the risk of corneal edema after phacoemulsification surgery, and assist ophthalmologists in making decisions in clinical work.

Description

Technical Field

[0001] The present invention relates to the field of medical technology, and in particular to an auxiliary analysis model for predicting corneal edema after cataract phacoemulsification. Background Art

[0002] Phacoemulsification is currently the leading treatment for cataracts. Corneal edema is a common complication after phacoemulsification, affecting both the patient's subjective experience and the surgical outcome. Preventing postoperative corneal edema through proactive clinical intervention is crucial for improving the efficacy of phacoemulsification. However, there is currently a lack of clinically accurate methods for predicting the occurrence of corneal edema.

[0003] Although there are similar prediction research reports, such as the "Prediction model for corneal edema after surgery in diabetic cataract patients" reported by Tian Jing et al., the model was constructed through Logistic regression, but the model has the following shortcomings: (1) The coverage of the patient group is limited, only for the diabetic cataract patient group; (2) The model is verified by its own data, and the model training set and test set are the same, so the prediction performance is not convincing; (3) The model uses many clinical indicators, which makes the model application complicated and cumbersome. For example, the "Prediction model for corneal edema after cataract surgery" reported by Pan Shangang et al., the model was constructed through χ 2 A prediction model was constructed by using the univariate logistic regression test. However, the model has the following shortcomings: (1) the model only included 9 indicators for univariate analysis, and some risk factors that have an important impact on the occurrence of postoperative corneal edema, such as preoperative BCVA, patient axial length, and CDE, were not included. There was no specific standard for anterior chamber depth / shallowness; (2) the multivariate logistic regression analysis only selected meaningful data from the univariate analysis, which lacked a theoretical basis; (3) the final model used too many clinical indicators, which was not convenient for application. Summary of the Invention

[0004] The purpose of the present invention is to provide a medical auxiliary analysis model for predicting the risk of corneal edema, so as to solve the problem that ophthalmologists cannot accurately predict the occurrence of corneal edema after phacoemulsification surgery, thereby being unable to prevent and treat patients at high risk of corneal edema.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: an auxiliary analysis model for predicting corneal edema after phacoemulsification surgery for cataracts, wherein the model uses the best corrected visual acuity before phacoemulsification surgery and the cumulative released energy during the phacoemulsification process as indicators to construct a nomogram, namely;

[0006] Among them, the nomogram includes five line segments, each line segment is marked with a fixed scale and the relative position of each line segment is fixed. The five line segments are, from top to bottom: BCVA and CDE score segment, BCVA value segment, CDE value segment, total BCVA and CDE score segment, and probability segment of corneal edema.

[0007] As a preferred technical solution: the scale of the BCVA and CDE scoring line segment is 0-100, the scale of the BCVA value line segment is 0-1, the scale of the CDE value line segment is 0-20, the scale of the total BCVA and CDE scoring line segment is 0-110, and the scale of the probability line segment of corneal edema is 0.001-0.999.

[0008] As a preferred technical solution: the starting points of the BCVA and CDE scoring line segment, BCVA value line segment, CDE value line segment, and the total line segment of BCVA and CDE scoring are aligned to the left; the end points of the BCVA and CDE scoring line segment, CDE value line segment, and the total line segment of BCVA and CDE scoring are aligned to the right; the end point of the BCVA value line segment is aligned with the scale "13.22" of the BCVA and CDE scoring line segment; the scale "2" of the CDE value line segment is aligned with the scale "10" of the BCVA and CDE scoring line segment; the starting point of the probability line segment of corneal edema is aligned with the scale "10" of the total line segment of BCVA and CDE scoring, and the end point is aligned with the scale "61" of the BCVA and CDE scoring line segment.

[0009] As a preferred technical solution: the best corrected visual acuity before the phacoemulsification surgery and the cumulative released energy index during the phacoemulsification surgery are determined by variable screening of copula entropy and multivariate logistic regression.

[0010] The "best corrected visual acuity" (BCVA) of the present invention is the best corrected visual acuity measured by an ophthalmologist or technician in a hospital before cataract surgery.

[0011] The cumulative dissipated energy (CDE) during phacoemulsification is the cumulative energy released by the surgical equipment during phacoemulsification of the lens during cataract phacoemulsification surgery.

[0012] The inventors of this application obtained some clinical data indicators that may be related to corneal edema after phacoemulsification by screening variables using copula entropy for multivariate logistic regression;

[0013] The results of multivariate logistic regression showed that the best corrected visual acuity before phacoemulsification surgery and the cumulative energy released during phacoemulsification were significantly correlated with the occurrence of corneal edema after phacoemulsification. A nomogram was generated based on the graphical analysis of the results of these two indicators in the multivariate logistic regression.

[0014] Based on the specific clinical data of BCVA and CDE and the results of multivariate logistic regression, two line segments and scales of BCVA and CDE were drawn;

[0015] The BCVA and CDE indicators are scored separately in the nomogram, and the scores correspond to the Points segment in a fixed ratio;

[0016] The sum of the scores of BCVA and CDE is plotted in a fixed ratio as the Total Points segment and its corresponding Risk of CE segment representing the risk of corneal edema.

[0017] Compared with the prior art, the advantages of the present invention are:

[0018] (1) For the first time, this study provides an auxiliary analysis model for accurately predicting corneal edema after phacoemulsification, filling a gap in the clinical work of ophthalmologists;

[0019] (2) The prediction model of the present invention can help ophthalmologists quickly and accurately determine the risk of corneal edema after phacoemulsification, and assist ophthalmologists in making decisions in clinical work;

[0020] (3) The prediction model of the present invention can complete relatively accurate predictions using only two common clinical data. The usage conditions have low data requirements, are convenient and practical, and have high accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 Schematic diagram of the prediction model structure of the present invention;

[0022] Figure 2 The decision curve diagram.

[0023] Figure 1 In the figure, BCVA is the best corrected visual acuity of the target cataract patient before phacoemulsification; CDE is the cumulative released energy during the phacoemulsification process during the operation; Points is the score of BCVA and CDE; Total Points is the sum of the scores of BCVA and CDE; Risk of CE is the probability of corneal edema. DETAILED DESCRIPTION

[0024] The present invention will be further described below with reference to the accompanying drawings.

[0025] Example 1:

[0026] An auxiliary analysis model for predicting corneal edema after phacoemulsification is constructed by:

[0027] The clinical data used in this example were derived from 178 phacoemulsification cases within the framework of a completed randomized controlled clinical trial (AGSPC trial); specifically, the clinical data of these 178 phacoemulsification cases were derived from cataract patients diagnosed with age-related cataract and who underwent phacoemulsification at the First Medical Center of PLA General Hospital in Beijing, China, from March 2021 to March 2022.

[0028] The collected clinical data included 17 clinical indicators that may affect the occurrence of corneal edema after surgery: gender, age, history of hypertension, history of diabetes, preoperative best-corrected visual acuity, preoperative intraocular pressure, lens nuclear hardness, axial length, anterior chamber depth, lens thickness, central corneal thickness, corneal endothelial cell density, type of fluid flow system used during surgery, cumulative released energy, ultrasound time, irrigation fluid volume, and total aspiration time. Variable screening was performed by calculating the copula entropy between these clinical indicators and the occurrence of corneal edema. Eight clinical indicators with values greater than the median of the 17 copula entropies were selected for multivariate logistic regression. The results of the multivariate logistic regression showed that "cumulative released energy" (CDE) was significantly correlated with the patient's preoperative "best-corrected visual acuity" (BCVA) and the occurrence of corneal edema after surgery.

[0029] All calculations were performed using R software (4.1.3). The copula entropy between the 17 variables and corneal edema was calculated using the copent package in R. The specific code is as follows:

[0030]

[0031]

[0032]

[0033] Eight clinical indicators with a copula entropy median greater than 17 were selected for multivariate logistic regression as a variable screening process. The results of multivariate logistic regression showed that "cumulative released energy" (CDE) was significantly correlated with the patient's preoperative "best corrected visual acuity" (BCVA) and the occurrence of corneal edema after surgery. Based on this result, a nomogram was generated in R software. The specific code is as follows:

[0034] library(rms)

[0035] library(readr)

[0036] mydata<-read.csv(″shuizhong.csv″)

[0037] mydata$edema<-as.factor(mydata$edema)

[0038] mydata$gender<-as.factor(mydata$gender)

[0039] mydata$NH<-as.factor(mydata$NH)

[0040] mydata$hbp<-as.fact or(mydata$hbp)

[0041] mydata$dia<-as.factor(mydata$dia)

[0042] mydata$fludics<-as.factor(mydata$fludics)

[0043] View(mydata)

[0044] summary(mydata)

[0045] dd<-datadist(mydata)

[0046] options(datadist=″dd″)

[0047] #Logistic regression

[0048] fit<-lrm(edema~age+bcva+NH+1t+cde+us+tat+efu,data=mydata,x=T,y=T)

[0049] fit

[0050] #result:bcva\cde

[0051] fitt<-lrm(edema~bcva+cde,data=mydata,x=T,y=T)#,family=binomial

[0052] fitt

[0053] summary(fitt)

[0054] nom<-nomogram(fitt, fun=plogis, fun.at=c(.001, .01, seq(.1, .9, by=.2), .99, .999), lp=F, funlabel="CE+Logistic")

[0055] plot(nom)

[0056] Compared to the nomogram generated by directly performing multivariate logistic regression on all 17 variables and corneal edema without screening with copula entropy variables, the nomogram of the present invention, which was screened with copula entropy variables, reduced the clinical information used to predict the occurrence of corneal edema from the patient's diabetes status, LT, BCVA, and CDE to BCVA and CDE, without significantly changing the area under the receiver operating characteristic curve (AUC). That is, without reducing the predictive performance of the nomogram, the type and amount of clinical information required for prediction were reduced, thereby increasing clinical practicality. The specific comparison code for the AUC value in R language is as follows:

[0057] library(readr)

[0058] library(rms)

[0059] library(pROC)

[0060] data<-read_csv(″shuizhong.csv″)

[0061] #data<-as.data.frame(data)

[0062] View(data)

[0063] data$edema<-factor(data$edema)

[0064] data$gender<-factor(data$gender)

[0065] data$NH<-factor(data$NH)

[0066] data$hbp<-factor(data$hbp)

[0067] data$dia<-factor(data$dia)

[0068] data$fludics <- factor(data$fludics)

[0069] summary(data)

[0070] # Build the model

[0071] # Logistic

[0072] log.form <- as.formula(edema ~ bcva + lt + dia + cde)

[0073] log.model <- glm(log.form, data = data, family = "binomial")

[0074] celog.fom <- as.formula(edema ~ bcva + cde)

[0075] celog.model <- glm(celog.form, data = data, family = "binomial")

[0076] # AUC calculation and

[0077] # Logistic

[0078] data$logpre <- predict(log.model, type = "response")

[0079] logROC <- roc(data$edema, data$logpre) # Plot roc

[0080] auc(logROC)

[0081] ci(auc(logROC))

[0082] # ceLogistic

[0083] data$celogpre <- predict(celog.model, type = "response")

[0084] celogROC <- roc(data$edema, data$celogpre) # Plot roc

[0085] auc(celogROC)

[0086] ci(auc(celogROC))

[0087] #Draw ROC

[0088] plot(logROC, col="red", legacy.axes=T)

[0089] plot(celogROC, add=TRUE, col="blue")

[0090] #AUC statistical test

[0091] roc.test(logROC, celogROC).

[0092] Based on the results of multivariate logistic regression, cumulative delivered energy (CDE) and best corrected visual acuity (BCVA) were selected to construct a nomogram as a prediction model for the risk of corneal edema after phacoemulsification.

[0093] The nomogram is as follows Figure 1 As shown in the figure, there are five line segments and scales, representing: the specific clinical data of the two indicators BCVA and CDE, the scores of the two clinical indicators, the sum of the scores of the two clinical indicators, and the risk of corneal edema. The scales, proportions, and relative positions between the five line segments are fixed and cannot be changed. The data of the target phacoemulsification patients are finally output under the prediction model as the probability of corneal edema on the first day after surgery;

[0094] In this embodiment, the first four line segments are uniformly fixed, and the scale of the fifth line segment is uneven but fixed, and is fixed as follows: taking the fourth line segment as a reference, the scale corresponding to the fifth line is "fifth line scale - fourth line scale": 0.001-9.80; 0.01-18.45; 0.1-27.31; 0.3-32.41; 0.5-35.43; 0.7-38.7; 0.9-43.63; 0.99-52.65; 0.999-61.16.

[0095] When in use, the specific clinical data of the best corrected visual acuity before surgery and the cumulative energy released during surgery are extracted from the clinical data of the target cataract patients after phacoemulsification, and marked in the corresponding BVCA segment and CDE segment respectively, and the BCVA and CDE scores are determined in the Points segment along the vertical direction of the markings; a new marking is made according to the position of the sum of the two scores on the TotalPoints segment, and the specific probability of corneal edema is predicted in the Risk of CE segment along the vertical direction of the new markings.

[0096] Case 1: A patient's preoperative BCVA was 0.5, corresponding to a 7-point value on the Points segment. Their intraoperative CDE was 4.0, corresponding to a 20-point value on the Points segment, for a total of 27 points. In the Risk of CE segment, the probability corresponding to 27 points on the Total Points segment is approximately 0.1, indicating a 10% probability of corneal edema during phacoemulsification.

[0097] Case 2: A patient's preoperative BCVA was 0.15, corresponding to a 1.5-point value on the Points segment. Their intraoperative CDE was 7.8, corresponding to a 42-point value on the Points segment, for a total of 43.5 points. In the Risk of CE segment, the probability corresponding to 43.5 points on the Total Points segment is approximately 0.9, indicating a 90% probability of corneal edema after phacoemulsification.

[0098] The user of the nomogram can arbitrarily set the threshold probability according to the specific clinical work needs. When the specific probability obtained by the nomogram is greater than the set threshold probability, the prediction result highly suggests that clinical intervention should be carried out in time to prevent the occurrence of corneal edema.

[0099] Example 2

[0100] Model accuracy verification:

[0101] The nomogram was validated using the "leave one out" method, which divided the data from 178 surgeries into a training set and a test set at a ratio of 177:1. This process was repeated 178 times, ensuring that the test set was unique. The result of each validation was whether corneal edema occurred. The "leave one out" method ultimately outputted a nomogram with a prediction accuracy of 90.98%.

[0102] The present invention adopts the "leave one out" method for verification. Compared with the existing technology that uses its own data verification and the same verification method for model training set and test set, the prediction results are more credible.

[0103] Decision curve analysis:

[0104] like Figure 2 As shown in the decision curve, the horizontal axis represents the threshold probability, and the vertical axis represents the net benefit. The solid line represents the assumption that all patients develop corneal edema, while the dashed line represents the assumption that no patient develops corneal edema. The decision curve shows that for any patient threshold probability between 0 and 1, using this model to predict the risk of corneal edema consistently yields a net benefit.

[0105] The analysis results showed that the model showed a net benefit in predicting postoperative corneal edema at almost any threshold probability, suggesting that the model has good clinical application value; the nomogram user can arbitrarily set the threshold probability according to specific clinical work needs. When the nomogram shows that the probability of corneal edema in patients after phacoemulsification is greater than the set threshold probability, it can be used as an important reference for clinical intervention to prevent the occurrence of corneal edema.

[0106] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An auxiliary analysis model for predicting corneal edema after phacoemulsification, characterized by: The model uses the best corrected visual acuity (BCVA) before phacoemulsification surgery and the cumulative released energy (CDE) during phacoemulsification surgery as indicators to construct a nomogram, namely; The nomogram includes five line segments, each of which is marked with a fixed scale and has a fixed relative position. The five line segments are, from top to bottom, the BCVA and CDE score segment, the BCVA value segment, the CDE value segment, the total BCVA and CDE score segment, and the probability segment of corneal edema. The scale of the BCVA and CDE score line segment is 0-100, the scale of the BCVA value line segment is 0-1, the scale of the CDE value line segment is 0-20, the scale of the total BCVA and CDE score line segment is 0-110, and the scale of the probability line segment of corneal edema is 0.001-0.999; The starting points of the BCVA and CDE scoring line segment, BCVA value line segment, CDE value line segment, and the total BCVA and CDE scoring line segment are aligned to the left; the end points of the BCVA and CDE scoring line segment, CDE value line segment, and the total BCVA and CDE scoring line segment are aligned to the right; the end point of the BCVA value line segment is aligned with the scale "13.22" of the BCVA and CDE scoring line segment; the scale "2" of the CDE value line segment is aligned with the scale "10" of the BCVA and CDE scoring line segment; the starting point of the probability line segment of corneal edema is aligned with the scale "10" of the total BCVA and CDE scoring line segment, and the end point is aligned with the scale "61" of the total BCVA and CDE scoring line segment.

Images