Post-ciliary muscle paralysis spherical power prediction method based on feature conversion and screening

Abstract

The invention relates to a post-ciliary muscle paralysis spherical power prediction method based on feature conversion and screening, and belongs to the technical field of electronic information. The method comprises the following steps: acquiring a clinical data sample, performing feature conversion to obtain parameter features related to a crystalline lens, and performing feature screening by using LASSO; an XGBoost model is utilized to train and predict the spherical power S after ciliary muscle paralysis and the cylindrical power C before ciliary muscle paralysis, and the performance of the model is evaluated according to MSE, RMSE, MAE and R2; calculating the equivalent spherical power SE after ciliary muscle paralysis according to the S and the C predicted by the XGBoost model; the anterior segment optical coherence tomography scanner is used for shooting crystalline lens images under different adjustments to clinically verify the change of the crystalline lens diopter caused by the change of the anterior surface diopter of the crystalline lens, namely the physical model used in the invention. According to the method, high-accuracy S prediction after ciliary muscle paralysis is realized, and side effects caused by ciliary muscle paralysis examination are avoided.

Description

Technical Field

[0001] The present invention belongs to the field of electronic information technology and relates to a method for predicting the spherical lens power after cycloplegia based on feature transformation and screening. Background Art

[0002] Myopia is one of the most common refractive errors, an irreversible refractive error eye disease with a high incidence and rapid progression among children and adolescents worldwide. In recent years, the prevalence of myopia has also been gradually increasing. Epidemiological investigation and research show that globally, the prevalence rates of myopia and high myopia are 28.3% and 4% respectively. It is predicted that by 2050, the prevalence rates of myopia and high myopia will increase to 49.8% and 9.8% respectively. Therefore, early detection, early diagnosis, and early treatment of myopia are very important.

[0003] Research shows that the difference in refractive power before and after cycloplegia increases with decreasing age. Therefore, for the low - age population, if only the refractive power before cycloplegia is used as a reference, the proportion of myopia may be overestimated while the proportions of emmetropia and hyperopia may be underestimated. Therefore, it is inaccurate to only use the pre - cycloplegia refraction for the adolescent and child population. Although the refractive power examination after cycloplegia is regarded as the gold standard, the cycloplegia process takes at least half an hour and at most several weeks, which is relatively time - consuming to implement. It not only easily causes congestion of outpatients but also is not easy to implement in the screening work. In addition, it has side effects such as photophobia caused by pupil dilation, blurred near vision caused by accommodation paralysis, and even acute intraocular pressure elevation. Some parents are overly worried about these side effects of cycloplegia and even directly reject the cycloplegia process.

[0004] To solve this problem, considering that the change in the cylindrical lens power before and after cycloplegia is much smaller than the commonly used refractive power measure, i.e., 0.25D, it is considered to use the pre - cycloplegia eye parameters to predict the spherical lens power after cycloplegia. Finally, the equivalent spherical lens power (SE) is calculated through a formula using the predicted spherical lens power (S) and the pre - cycloplegia cylindrical lens power (C). Since the number of pre - cycloplegia eye parameters is large and the relationships are complex, we use feature screening and feature transformation methods to establish a prediction model using pre - cycloplegia eye parameters to predict the spherical lens power after cycloplegia.

[0005] There have been many previous studies that established prediction models to predict the refractive power or refractive state after cycloplegia through parameters such as gender, age, visual acuity, refractive error, axial length of the eye, axial ratio, amount of accommodative lag, and other optometric biometric parameters. Although these studies have achieved good results, they did not take into account that the difference in refractive power before and after cycloplegia is mainly caused by changes in the lens. Research has shown that the change in refractive power caused by cycloplegia is due to the lens, and the change in lens refractive power is mainly caused by the change in the refractive power of the anterior surface of the lens. The present invention will use this change as a physical model for subsequent work. In clinical practice, this change can be roughly observed through a slit lamp microscope (visible Figure 3 ). To more accurately verify the physical model used in the present invention, the present invention uses an anterior segment optical coherence tomography scanner (AS-OCT) to capture lens images under different accommodations to clinically verify this change. Summary of the Invention

[0006] In view of this, the purpose of the present invention is to provide a method for predicting spherical lens power after cycloplegia based on feature transformation and screening, which uses the eye parameters before cycloplegia to predict the spherical lens power after cycloplegia and obtain the refractive power. Since the number of eye parameters before cycloplegia is large and the relationships are complex, therefore, feature screening and feature transformation methods are used to establish a prediction model using the eye parameters before cycloplegia to predict the spherical lens power after cycloplegia. The XGBoost model is used to predict the spherical lens power after cycloplegia, and the model performance is evaluated according to MSE, RMSE, MAE, and R2. It is obtained that a prediction model is established with crystal-related parameters as the focus, which can more accurately predict the spherical lens power after cycloplegia.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A method for predicting spherical lens power after cycloplegia based on feature transformation and screening, the method comprising the following steps:

[0009] S1: Obtain the clinical data of the prediction object and perform preprocessing;

[0010] S2: Perform feature transformation on the preprocessed clinical data;

[0011] S3: Perform feature screening on the data after feature transformation using LASSO regression;

[0012] S4: Establish an XGBoost model based on the screened features and perform prediction;

[0013] S5: Use MSE, RSME, MAE, R2 The evaluation model evaluates the performance of the established XGBoost model;

[0014] S6: Verify the physical model using the clinical verification method.

[0015] Optionally, in the said S1:

[0016] The acquisition criteria for clinical data are: no history of eye diseases and surgeries;

[0017] The exclusion criteria for clinical data are: patients under 4 years old and over 15 years old;

[0018] Clinical data includes: patient age, gender, corrected visual acuity, uncorrected visual acuity, spherical lens in pre-cycloplegic refraction, cylindrical lens in pre-cycloplegic refraction, equivalent spherical lens degree in pre-cycloplegic refraction, intraocular pressure, axial length of the eye, anterior chamber depth, lens thickness, accommodation lag, and corneal curvature. Extract X characteristic indicators, where X = 13.

[0019] Optionally, in the said S1, the preprocessing of clinical data specifically is: cleaning the clinical data, including handling missing values.

[0020] Optionally, the handling of missing values includes mean imputation and median imputation;

[0021] The mean imputation is to use the average value to replace the missing value;

[0022] The median imputation is to use the median to replace the missing value.

[0023] Optionally, in the said S2, the diopter of the lens is replaced by the intraocular lens (IOL) degree, and the corresponding anterior surface diopter F of the lens and the anterior surface curvature radius r of the lens are calculated through the thick lens formula and the diopter formula. A and the anterior surface curvature radius r of the lens A .

[0024] Optionally, in the said S3, LASSO is used for feature selection. By introducing the L1 regularization term, the coefficients of some irrelevant or redundant features are set to zero, and unimportant features are eliminated; Y features are selected, where Y is a positive integer and Y ≤ X.

[0025] Optionally, in the said S4, the machine learning method of extreme gradient boosting XGBoost is used to model the selected features, and two spherical lens degree prediction models are established:

[0026] One is the control group model established by the feature set that does not include lens-related features;

[0027] The other is the experimental group model established by the feature set that includes lens-related features.

[0028] Optionally, in S5, MSE, RSME, MAE, R are used 2 Evaluate the performance of the two models in the control group and the experimental group, and verify the contribution of lens-related features to the prediction model after cycloplegia by comparing the R of the two models 2 , MSE, RMSE, and MAE.

[0029] Optionally, in S6, in the clinical verification method, AS-OCT is used to photograph the parameters of the anterior and posterior surfaces of the lens under different regulations, and the lens refractive power under each anterior and posterior surface parameter is calculated using the thick lens formula; compare the contributions of the changes in the anterior and posterior surfaces of the lens to the change in the lens refractive power respectively.

[0030] Optionally, in S2, the calculation method of the IOL value is as follows:

[0031] Calculate the IOL using the conventional formula (1) given by the SRK / T formula:

[0032]

[0033] LOPT = Al + 0.65696 - 0.02029×Al (2)

[0034] where n a is the refractive index of the aqueous humor and vitreous humor, n a is 1.336; n c m1 is the corneal refractive index minus 1, n c m1 is 0.333, r is the corneal curvature radius, ACD is the anterior chamber depth, AL is the axial length of the eye, and LOPT is the corrected axial length;

[0035] The near-vision formula (3) with reserved degrees given by the SRK / T formula:

[0036]

[0037] Calculate the IOL value of the subject at their original refractive power, that is, IOL M ; where SE is the equivalent spherical refractive power, obtained from formula (4):

[0038] SE = S + 1 / 2×C (4)

[0039] where S is the spherical refractive power and C is the cylindrical refractive power;

[0040] Convert the subject's original equivalent spherical refractive power through the contact lens conversion formula (5), add it to the subject's corneal refractive power, and then calculate the corneal curvature radius r at this time according to formula (6) CL, and then put it into formula (1) to calculate the IOL value obtained by adding the converted original diopter of the subject through the contact lens to the corneal diopter, that is, IOL CL ;

[0041]

[0042]

[0043] IOL CL represents the IOL value under the corneal curvature radius. The diopter of the lens is determined by the refractive index of the intraocular lens;

[0044] The corresponding anterior surface diopter F of the lens is calculated through the thick lens formula (7) and the diopter formula (8) respectively A and the anterior surface curvature radius r of the lens A :

[0045]

[0046]

[0047] where F P is the posterior surface diopter of the lens, and its value is taken from the Bennett and Rabbetts model eye;

[0048] The SE after cycloplegia is obtained by using the established XGBoost model to predict the S after cycloplegia and the C before cycloplegia and passing them through formula (4).

[0049] The beneficial effects of the present invention are as follows: By using the methods of feature transformation and feature screening, a prediction model is established with the pre-cycloplegia eye parameters to predict the spherical lens degree after cycloplegia, achieving efficient and highly accurate prediction, avoiding the side effects such as time-consuming and acute intraocular pressure elevation brought by cycloplegia examination to people, preventing the congestion of outpatient patients caused by time-consuming, and the difficulty of implementing cycloplegia examination in screening work, improving the hospital examination efficiency, reducing the examination cost. At the same time, when selecting the established model, considering from multiple aspects, the established model can more accurately predict the diopter after cycloplegia, which has important research significance and application value for simplifying cycloplegia examination.

[0050] Other advantages, objectives and features of the present invention will be elaborated to some extent in the subsequent specification, and to some extent, will be obvious to those skilled in the art based on the study of the following text, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the following specification. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] To make the objectives, technical solutions and advantages of the present invention more clear, the following will provide a preferred and detailed description of the present invention with reference to the accompanying drawings, where:

[0052] Figure 1 is the flow chart of the present invention;

[0053] Figure 2 is the schematic diagram of the physical model used in the present invention;

[0054] Figure 3 is the change of the lens during accommodation photographed by a slit lamp microscope in clinical practice;

[0055] Figure 4 is the linear regression graph obtained from the true values and predicted values in the model validation set;

[0056] Figure 5 is the comparison graph of the change in the anterior and posterior surface curvature radii of the lens with accommodation obtained by AS-OCT. Specific Embodiments

[0057] The following uses specific specific examples to illustrate the embodiments of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the drawings provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner. Without conflict, the following embodiments and the features in the embodiments can be combined with each other.

[0058] Among them, the drawings are only for illustrative purposes, showing only schematic diagrams, not physical diagrams, and should not be construed as a limitation to the present invention; in order to better illustrate the embodiments of the present invention, some components in the drawings will be omitted, enlarged or reduced, which do not represent the dimensions of the actual product; for those skilled in the art, it is understandable that some well-known structures and their descriptions in the drawings may be omitted.

[0059] In the drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components; in the description of the present invention, it should be understood that if there are terms such as "upper", "lower", "left", "right", "front", "rear", etc. indicating the orientation or positional relationship, they are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, the terms describing the positional relationship in the drawings are only for illustrative purposes and should not be construed as a limitation to the present invention. For those of ordinary skill in the art, the specific meanings of the above terms can be understood according to specific circumstances.

[0060] As Figure 1 shown, the present invention provides a method for predicting the spherical lens power after cycloplegia to obtain the refractive power, including the following steps:

[0061] S1: Obtain the clinical relevant data of the prediction object, collect data according to the inclusion criteria and exclusion criteria for collecting the clinical data of the prediction object, and obtain X features, where X is a positive integer. And preprocess the collected data, mainly for the processing of missing values.

[0062] S2: Perform feature transformation on the data, calculate the IOL value through a formula to replace the refractive power of the lens, and calculate the corresponding anterior surface refractive power F of the lens and the anterior surface curvature radius r of the lens through the thick lens formula and the refractive power formula. A and the anterior surface curvature radius r of the lens A .

[0063] S3: Perform feature screening on the data using LASSO regression, eliminate unimportant features, and obtain Y features, where Y is a positive integer and Y ≤ X;

[0064] S4: Use the screened features to establish an XGBoost model and make predictions;

[0065] S5: Use MSE, RSME, MAE, R 2 to evaluate the model performance, and verify the contribution of the lens-related features to the prediction model after cycloplegia by comparing the R 2 , MSE, RMSE, and MAE of the two models.

[0066] S6: Verify the physical model by the clinical verification method.

[0067] Obtain the data information of the patient to be retrieved, input the data information of the patient to be retrieved into the prediction model, and the prediction model makes a prediction on the refractive power according to the data information of the patient to be retrieved.

[0068] In S1, obtain the clinical relevant data of the prediction object, collect data according to the inclusion criteria and exclusion criteria for collecting the clinical data of the prediction object. The inclusion criteria for collecting clinical data are: no eye disease history and surgery history. The exclusion criteria for collecting clinical data are: patients under 4 years old and over 15 years old.

[0069] In S1, preprocess the data, perform basic cleaning on the data. The basic cleaning of the data includes the processing of missing values. The processing of missing values is mean imputation: using the average value to replace the missing value; median imputation: using the median to replace the missing value.

[0070] In S2, perform feature transformation analysis on the data to obtain the pre-cycloplegic eye parameters. The lens refractive power cannot be directly obtained through routine examinations. Therefore, the refractive power of the lens is replaced by the IOL value, and the corresponding anterior surface refractive power FA and rA of the lens are calculated using the thick lens formula and the refractive power formula.

[0071] In S3, use LASSO to screen the features. By introducing the L1 regularization term, the coefficients of some irrelevant or redundant features are set to zero, and the features with low importance are eliminated; Y features are selected, where Y is a positive integer and Y ≤ X.

[0072] In S4, use the method of feature screening and feature transformation with extreme gradient boosting XGBoost to model the screened features. Two spherical lens power prediction models are established: a control group model established with a feature set that does not include lens-related features and an experimental group model established with a feature set that includes lens-related features; at the same time, use the established models to predict the obtained S after cycloplegia and C before cycloplegia, and calculate SE after cycloplegia through a formula.

[0073] In S5, use MSE, RSME, MAE, R 2 Evaluate the performance of the two models in the control group and the experimental group, and verify the contribution of the lens-related features to the prediction model after cycloplegia by comparing the R 2 , MSE, RMSE, and MAE of the two models.

[0074] The physical model used in the present invention is: the change in refractive power caused by accommodation is mainly caused by the change in the anterior surface of the lens, as Figure 2 shown. In S6, use the clinical verification method to verify the physical model. Use AS-OCT to photograph the parameters of the anterior and posterior surfaces of the lens under different accommodations, and use the thick lens formula to calculate the refractive power of the lens under each anterior and posterior surface parameter. Then compare the contributions of the changes in the anterior and posterior surfaces of the lens to the change in the refractive power of the lens respectively. Figure 3 It is the change of the lens during accommodation photographed by a slit lamp microscope in clinical practice. The left side is photographed under relaxed accommodation, and the right side is photographed under tense accommodation.

[0075] The present invention provides a machine learning device for predicting spherical lens power based on feature screening and feature transformation. The device includes a processor and a memory. The memory is used to store instructions, and the processor is used to execute the instructions to implement the feature screening and feature transformation method as described in any one of the above claims.

[0076] In S5, for the performance evaluation of the two refractive power prediction models constructed by the XGBoost method, MSE, RMSE, MAE, R 2 The comparison results are shown in Table 1.

[0077] Table 1

[0078]

[0079] As can be seen from Table 1, the MSE, RMSE, MAE, and R of the control group 2 are 0.3012, 0.5309, 0.3502, and 0.9331 respectively; as can be seen from Table 1, the MSE, RMSE, MAE, and R of the three experimental groups 2 are 0.2516, 0.4824, 0.3155, and 0.9640 respectively; 0.2420, 0.4733, 0.3121, and 0.9629; 0.2413, 0.4717, 0.3073, and 0.9639.

[0080] From Table 1, the proportion of samples with the difference between the predicted value and the post-paralysis value within 0.50 D can be obtained; for the control group, it is 87.1%, for Model 1 and Model 2, it is 88.71%, and for Model 3, it is 90%. From the linear regression graphs obtained from the true values and predicted values of the four model validation sets, it can be seen Figure 4 .

[0081] The mean-square error (MSE) is a measure reflecting the degree of difference between the estimator and the estimated quantity. Let t be an estimator of the population parameter θ determined from the subsample, and the mathematical expectation of (θ - t) 2 is called the mean-square error of the estimator t. It is equal to σ 2 + b 2 , where σ 2 and b are the variance and bias of t respectively.

[0082] The root mean square error, also known as the standard error, is defined as i = 1, 2, 3,... n. In a finite number of measurements, the root mean square error is often expressed by the following formula: √[∑di 2 / n] = Re, where: n is the number of measurements; di is the deviation of a set of measured values from the true value.

[0083] The mean absolute error (MAE) is another loss function used for regression models. MAE is the sum of the absolute values of the differences between the target value and the predicted value, and it only measures the average magnitude of the prediction error without considering the direction.

[0084] R 2 (R squared, Coefficient of determination), translated into Chinese as "coefficient of determination" or "goodness of fit", reflects the accuracy of the model fitting the data. Generally, R 2The range is from 0 to 1. The closer to 1, the stronger the explanatory power of the variables of the equation for y, and the better the model fits the data. The closer to 0, the worse the model fits.

[0085] In S6, by processing the images of the lens under different accommodations captured by AS-OCT, with the change amount of the anterior and posterior surface curvature radii of the lens as the abscissa and the accommodation amount as the ordinate, a Figure 5 . From Figure 5 it can be seen that at the same accommodation amount, the change amount of the anterior surface curvature radius of the lens is greater than that of the posterior surface curvature radius. Therefore, it can be concluded that the physical model used in the present invention is correct.

[0086] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the purpose and scope of the present technical solution, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A method for predicting the spherical lens power after cycloplegia based on feature transformation and screening, characterized in that: The method includes the following steps: S1: Obtain the clinical data of the prediction object and perform preprocessing; S2: Perform feature transformation on the preprocessed clinical data; S3: Use LASSO regression to perform feature screening on the data after feature transformation; S4: Establish an XGBoost model based on the screened features and perform prediction; S5: Use MSE, RSME, MAE, R 2 Evaluate the model to perform performance evaluation on the established XGBoost model; S6: Use the clinical verification method to verify the physical model.

2. The method for predicting the spherical lens power after cycloplegia based on feature transformation and screening according to claim 1, wherein: In the said S1: The acquisition criteria for clinical data are: no eye disease history and surgical history; The exclusion criteria for clinical data are: patients under 4 years old and over 15 years old; Clinical data includes: patient age, gender, corrected visual acuity, uncorrected visual acuity, spherical lens in pre-cycloplegic refraction, cylindrical lens in pre-cycloplegic refraction, equivalent spherical lens degree in pre-cycloplegic refraction, intraocular pressure, axial length of the eye, anterior chamber depth, lens thickness, accommodative lag, and corneal curvature. Extract X feature indicators, where X = 13.

3. The method for predicting the spherical lens power after cycloplegia based on feature transformation and screening according to claim 1, wherein: In the said S1, the preprocessing of clinical data is specifically: cleaning the clinical data, including missing value processing.

4. The method for predicting the spherical lens power after cycloplegia based on feature transformation and screening according to claim 3, wherein: The missing value processing includes mean imputation and median imputation; The mean imputation is to use the average value to replace the missing value; The median imputation is to use the median to replace the missing value.

5. The method for predicting the spherical lens power after cycloplegia based on feature transformation and screening according to claim 1, wherein: In S2, the diopter of the lens is replaced by the intraocular lens (IOL) power. The corresponding diopter F of the anterior surface of the lens is calculated using the thick lens formula and the diopter formula. A and the radius of curvature r of the anterior surface of the lens A .

6. The method for predicting the spherical lens power after cycloplegia based on feature transformation and screening according to claim 2, wherein: In the said S3, LASSO is used for feature screening. By introducing the L1 regularization term, the coefficients of some irrelevant or redundant features are set to zero to eliminate unimportant features; Y features are screened out, where Y is a positive integer and Y ≤ X.

7. The method for predicting the spherical lens power after cycloplegia based on feature transformation and screening according to claim 1, wherein: In the said S4, the machine learning extreme gradient boosting XGBoost method is used to model the screened features, and two spherical lens degree prediction models are established: One is the control group model established by the feature set that does not include lens-related features; The other is the experimental group model established by the feature set that includes lens-related features.

8. The method for predicting the spherical lens power after cycloplegia based on feature transformation and screening according to claim 1, wherein: In S5, MSE, RSME, MAE, and R are used 2 to evaluate the performance of the two models in the control group and the experimental group, and by comparing the R 2 , MSE, RMSE, and MAE of the two models to verify the contribution of lens-related features to the prediction model after cycloplegia.

9. The method for predicting the spherical lens power after cycloplegia based on feature transformation and screening according to claim 1, wherein: In the said S6, the clinical verification method uses an anterior segment optical coherence tomography (AS-OCT) to capture a set of parameters of the anterior and posterior surfaces of the lens under different accommodations, and uses the thick lens formula to calculate the lens refractive power under each anterior and posterior surface parameter; compare the contributions of the changes in the anterior and posterior surfaces of the lens to the changes in the lens refractive power respectively.

10. The method for predicting the spherical lens power after cycloplegia based on feature transformation and screening according to claim 5, wherein: In the said S2, the calculation method of the IOL value is: The IOL is calculated by the conventional formula (1) given by the SRK / T formula: LOPT = AL + 0.65696 - 0.02029×AL (2) where n a is the refractive index of the aqueous humor and the vitreous humor, and n a is 1.336; n c m1 is the corneal refractive index minus 1, and n c m1 is 0.333, r is the corneal curvature radius, ACD is the anterior chamber depth, AL is the axial length of the eye, and LOPT is the corrected axial length of the eye; The near-using formula (3) for retaining the degree given by the SRK / T formula: Calculate the IOL value of the subject at their original refractive power, i.e., IOL M ; where SE is the spherical equivalent, obtained from formula (4): SE = S + 1 / 2×C (4) Where S is the spherical lens degree and C is the cylindrical lens degree; The original equivalent spherical refractive power of the subject is converted by the contact lens conversion formula (5) and added to the corneal refractive power of the subject. Then, the corneal curvature radius r at this time is calculated according to formula (6). CL Then, it is put into formula (1) to calculate the IOL value obtained by adding the original refractive power of the subject after contact lens conversion to the corneal refractive power, that is, IOL CL ; IOL CL Indicates the IOL value at the corneal radius of curvature. The refractive power of the lens is determined by the refractive index of the intraocular lens; The diopter F of the anterior surface of the lens is calculated respectively through the thick lens formula (7) and the diopter formula (8). A and the radius of curvature r of the anterior surface of the lens A : where F P is the posterior surface diopter of the lens, and its value is taken from the Bennett and Rabbetts model eye; The spherical lens degree S after cycloplegia and the C before cycloplegia obtained by prediction using the established XGBoost model are used to obtain the SE after cycloplegia through formula (4).

Images