An intraocular lens selection and rotational implantation prediction method and electronic device

Abstract

This application discloses a method and electronic device for predicting intraocular lens (IOL) selection and rotation implantation, relating to the field of computer technology. This application acquires the user's ocular biological measurement parameters and constructs an intraocular geometric constraint region. The IOL is abstracted as a rectangular geometries, generating a set of candidate size-axis positions. Each candidate option satisfies the geometric constraints of the patient's intraocular anatomy in terms of implantation position and rotation angle. Based on this, by combining the geometric fit indices of the candidate options with individualized ocular biological parameters, the postoperative arch height, the probability of abnormal arch height, and postoperative refractive results are predicted. After post-processing and calibration, outputs with confidence intervals or risk probabilities are provided. This eliminates reliance on experience or rough estimates for surgical planning, effectively reducing the risk of arch height deviation or postoperative refractive errors, providing the optimal IOL option, and achieving individualized, quantifiable, and predictable implantation planning.

Description

Technical Field

[0001] This application relates to the field of geometric modeling technology, and in particular to a method and electronic device for predicting the selection and rotational implantation of an intraocular lens. Background Technology

[0002] ICL (Implantable Collamer Lens) is a type of implantable contact lens, also called a phakic intraocular lens, or simply an artificial lens or "endoscopic lens." Currently, ICL selection mainly relies on empirical formulas and manufacturer rule tables, using a limited number of parameters such as corneal horizontal diameter (white to white, WTW), anterior chamber depth (ACD), axial length, and comprehensive refraction results for a rough estimate. This method has the following problems:

[0003] First, individual differences are not adequately modeled: the horizontal ciliary sulcus diameter (h-STS) and the vertical ciliary sulcus diameter (v-STS) of patients may be significantly different. Traditional methods that rely on WTW to estimate the implanted intraocular lens often fail to reflect the actual elliptical space and orientation within the implanted eye.

[0004] Second, insufficient consideration of the rotation factor of the artificial lens: The shape of the ICL is approximately rectangular, and it may rotate after implantation. The final axis has an important impact on the fit and stability.

[0005] Third, ICLs have only a limited number of size specifications. For example, Staar Surgical's ICLs are available in only four sizes. If only horizontal or vertical implantation is used, the ICL size cannot be continuously matched with the STS. In the past, horizontal or vertical implantation methods could result in the arch height being too high after increasing the size, causing high intraocular pressure or even glaucoma, or the arch height being too low after decreasing the size, causing cataracts. Or, there could be a dilemma where the arch height is too high after horizontal ICL implantation and too low after vertical ICL implantation, affecting the safety and effectiveness of the surgery.

[0006] Fourth, the separation between ICL selection and post-implantation outcome prediction: Existing methods select ICL size based on WTW before surgery and determine the refractive power of the implanted ICL based on comprehensive refraction, but it is only after implantation that the appropriate arch height and target refractive power can be evaluated, which cannot complete the closed loop of "candidate space generation + post-operative outcome prediction".

[0007] Fifth, insufficient identification of clinical abnormal risks: clinicians are most concerned about the risks of excessively low or high arch height, as well as residual refractive errors after surgery, but the existing procedures cannot provide probabilistic and calibrable risk information.

[0008] Sixth, real-world data contains noise and inconsistent standards: preoperative / postoperative assessment equipment varies across different institutions using different technologies, and the skill levels of equipment operators differ, resulting in issues such as missing data, equipment accuracy errors, human error, and inconsistent follow-up time points. Traditional static formulas are difficult to robustly handle these problems. Summary of the Invention

[0009] This application provides a method and electronic device for predicting the selection and rotation of intraocular lenses (IOLs) for implantation, which at least solves the problem in related technologies that the size and rotation angle of IOLs cannot be determined accurately and individually, and that it is impossible to predict before surgery whether the appropriate arch height and target refractive power will be achieved to determine a reasonable IOL selection and rotation angle.

[0010] This application provides a method for predicting intraocular lens selection and rotational implantation, including:

[0011] Obtain the user's ocular biological measurement parameters, and construct an intraocular geometric constraint region based on the ocular biological measurement parameters;

[0012] The artificial lens is abstracted as a rectangular occupier geometry, and a candidate size-axis set is generated based on the size of the artificial lens and the rotation angle around its center.

[0013] Based on the candidate size-axis set, multiple candidate schemes are formed by determining the geometric adaptation index that satisfies the requirement that the rectangular occupant geometry is located within the intraocular geometric constraint region.

[0014] Based on the ocular biological measurement parameters and the geometric fit index of the candidate schemes, the postoperative arch height, the probability of abnormal arch height, and the postoperative refractive results of each candidate scheme are predicted, and post-processing calibration is performed to output the prediction results of each candidate scheme with confidence intervals or risk probabilities.

[0015] Based on the prediction results, each candidate solution is ranked, and at least one recommended solution is output according to the ranking order and the geometric fit index of each candidate solution.

[0016] This application also provides an electronic device, comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of any of the above-described intraocular lens selection and rotation implantation prediction methods.

[0017] Obtain the user's ocular biological measurement parameters, and construct an intraocular geometric constraint region based on the ocular biological measurement parameters;

[0018] The artificial lens is abstracted as a rectangular occupier geometry, and a candidate size-axis set is generated based on the size of the artificial lens and the rotation angle around its center.

[0019] Based on the candidate size-axis set, multiple candidate schemes are formed by determining the geometric adaptation index that satisfies the requirement that the rectangular occupant geometry is located within the intraocular geometric constraint region.

[0020] Based on the ocular biological measurement parameters and the geometric fit index of the candidate schemes, the postoperative arch height, the probability of abnormal arch height, and the postoperative refractive results of each candidate scheme are predicted, and post-processing calibration is performed to output the prediction results of each candidate scheme with confidence intervals or risk probabilities.

[0021] Based on the prediction results, each candidate solution is ranked, and at least one recommended solution is output according to the ranking order and the geometric fit index of each candidate solution.

[0022] This application utilizes the acquisition of the user's ocular biological measurement parameters and the construction of an intraocular geometric constraint region. The intraocular lens (IOL) is abstracted into a rectangular geometries, generating a candidate size-axis set. This ensures that each candidate option meets the geometric constraints of the patient's intraocular anatomy in terms of implantation position and rotation angle. Based on this, by combining the geometric fit indices of the candidate options with individualized ocular biological parameters, postoperative arch height, the probability of abnormal arch height, and postoperative refractive results are predicted. Post-processing calibration provides outputs with confidence intervals or risk probabilities, eliminating reliance on experience or rough estimations for surgical planning. This method cleverly leverages the infinite number of elliptical diameters of varying lengths in the ciliary sulcus to achieve a perfect match with the finite lens size. It not only allows for the preoperative determination of the optimal IOL size and rotation angle but also accurately assesses the postoperative arch height and refractive results of each candidate option. This effectively reduces the risk of arch height deviation or substandard postoperative refractive results, providing the optimal IOL option and achieving individualized, quantifiable, and predictable implantation planning. This solves the problems of traditional methods, such as the inability to accurately determine the size and angle of IOL selection and rotational implantation, and the inability to accurately predict postoperative arch height and refractive status preoperatively. Attached Figure Description

[0023] To more clearly illustrate the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a flowchart illustrating the method for predicting intraocular lens selection and rotational implantation in one embodiment of this application;

[0025] Figure 2 This is a schematic diagram of the structure when the difference between the candidate angle of the artificial lens relative to the eyeball and the horizontal angle is 0° in one embodiment of this application;

[0026] Figure 3 This is a schematic diagram of the structure in one embodiment of this application when the difference between the candidate angle of the artificial lens relative to the eyeball and the horizontal angle is not 0°;

[0027] Figure 4 This is a structural block diagram of an intraocular lens selection and rotation implantation prediction device in one embodiment of this application;

[0028] Figure 5 This is an internal structural diagram of a computer device in one embodiment of this application. Detailed Implementation

[0029] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of this application.

[0030] It should be noted that, in the description of this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. The terms "first," "second," etc., in this application are used to distinguish similar objects and are not used to describe a specific order or sequence.

[0031] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0032] like Figure 1 As shown, embodiments of this application provide a method for predicting intraocular lens selection and rotational implantation, including the following steps:

[0033] Step S1: Obtain the user's ocular biological measurement parameters, and construct an intraocular geometric constraint region based on the ocular biological measurement parameters;

[0034] Step S2: Abstract the artificial lens into a rectangular occupier geometry, and generate a candidate size-axis set based on the size of the artificial lens and the rotation angle around its center;

[0035] Step S3: Based on the candidate size-axis set, determine the geometric fit index that satisfies the requirement that the rectangular occupant geometry is located within the intraocular geometric constraint region to form multiple candidate schemes;

[0036] Step S4: Based on the ocular biological measurement parameters and the geometric fit index of the candidate schemes, predict the postoperative arch height, the probability of abnormal arch height, and the postoperative refractive results of each candidate scheme, and perform post-processing calibration to output the prediction results of each candidate scheme with confidence intervals or risk probabilities.

[0037] Step S5: Sort the candidate solutions according to the prediction results, and output at least one recommended solution according to the sorting order and the geometric fit index of each candidate solution.

[0038] This method involves acquiring the user's ocular biological measurement parameters and constructing an intraocular geometric constraint region (the ciliary sulcus). The intraocular lens (IOL) is abstracted as a rectangular geometries, generating a set of candidate sizes and axes. This ensures that each candidate option meets the geometric constraints of the patient's intraocular anatomy in terms of implantation position and rotation angle. Based on this, by combining the geometric fit indices of the candidate options with individualized ocular biological parameters, the method predicts postoperative arch height, the probability of abnormal arch height, and postoperative refractive outcomes. Post-processing and calibration provide outputs with confidence intervals or risk probabilities, freeing surgical planning from reliance on experience or rough estimates. This method not only rationally determines the optimal size and rotation angle of the IOL preoperatively but also accurately assesses the postoperative arch height and refractive outcomes of each candidate option. This effectively reduces the risk of arch height deviation or postoperative refractive errors, providing the optimal IOL option and achieving individualized, quantifiable, and predictable implantation planning. It solves the problems of traditional methods where IOL selection and rotation cannot accurately determine size and angle, and postoperative arch height and refractive status cannot be predicted preoperatively.

[0039] In this embodiment, obtaining the user's ocular biological measurement parameters and constructing an intraocular geometrically constrained region based on the ocular biological measurement parameters includes:

[0040] The user's ocular biological measurement parameters include one or more of the following: horizontal ciliary sulcus diameter (h-STS), vertical ciliary sulcus diameter (v-STS), horizontal corneal diameter (white to white, WTW), anterior chamber depth (ACD), axial length (AL), lens thickness, corneal thickness, corneal curvature, axial length, pupil diameter, and refractive data.

[0041] Identify whether there is an abnormal region within the groove of the user's eyeball; if so, represent the abnormal region within the groove by clock position or angle range.

[0042] Determine whether the data for the horizontal ciliary sulcus diameter (h-STS) and the vertical ciliary sulcus diameter (v-STS) are missing;

[0043] If the data for the horizontal ciliary sulcus diameter (h-STS) and the vertical ciliary sulcus diameter (v-STS) are not missing, then the horizontal ciliary sulcus diameter (h-STS) and the vertical ciliary sulcus diameter (v-STS) are respectively used as the major axis or minor axis of the ellipse to construct the intraocular geometric constraint region.

[0044] If either the horizontal ciliary sulcus diameter (h-STS) or the vertical ciliary sulcus diameter (v-STS) is missing, the data of the vertical ciliary sulcus diameter (v-STS) or the horizontal ciliary sulcus diameter (h-STS) is used as a substitute, and the intraocular geometric constraint region is constructed using the substitute data.

[0045] If the data for both the horizontal ciliary sulcus diameter (h-STS) and the vertical ciliary sulcus diameter (v-STS) are missing, the intraocular geometric constraint region is calculated based on the horizontal corneal diameter, wherein the length of the major or minor axis of the ellipse is not limited.

[0046] The refractive data includes spherical power, cylindrical power, and cylindrical axis.

[0047] In some patients, cysts, adhesions or other focal abnormalities may be present in the groove. Doctors can manually mark one or more "forbidden zone" directions (e.g., sector ranges represented by clock positions) before or during the operation, map the forbidden zone to angle intervals to form abnormal areas in the groove, and shield or penalize abnormal areas in the groove during angle scanning and feasible interval output.

[0048] Preferably, the abnormal region within the trench can be buffered and expanded, for example, by expanding by a preset angle on both sides of the boundary of the abnormal region within the trench, in order to improve the safety margin; when the abnormal region within the trench causes the original feasible interval to be divided, multiple feasible angle sub-intervals after division are output.

[0049] In this embodiment, the artificial lens is abstracted as a rectangular placeholder geometry, the diagonal length of which is equal to the nominal size of the artificial lens. A candidate size-axis set is generated based on the size of the artificial lens and the rotation angle around its center, including:

[0050] The artificial lens is abstracted as a rectangular occupier geometry, and the diagonal length of the rectangular occupier geometry is set to be equal to the nominal size of the artificial lens;

[0051] The rectangular placeholder geometry is rotated around its center, and multiple candidate placement angles within the range of 0°–180° are scanned. Each candidate angle is associated with the size of the rectangular placeholder geometry to form a candidate size-axis set.

[0052] In this embodiment, when associating each candidate angle with the size of the rectangular occupier geometry to form a candidate size-axis set, the method further includes:

[0053] A rotation drift interval is set for each candidate angle, and the geometric adaptation margin of rectangular occupant geometries of multiple sizes is sampled within the rotation drift interval. The geometric adaptation margin of the rectangular occupant geometries of each size at each candidate angle is statistically analyzed.

[0054] The rotational drift range after intraocular lens surgery is simulated based on the geometric fit margin of the rectangular occupant geometry of each size at each candidate angle.

[0055] The rotational drift range is added when forming the candidate size-axis set.

[0056] Different rotational drift ranges can be used for astigmatic and non-astigmatic patients. The worst-case fit margin, median fit margin, etc., within the rotational drift range are calculated to measure the robustness of the scheme to postoperative rotation.

[0057] Specifically, when sampling geometric fit margins for rectangular occupant geometries of multiple sizes within the rotational drift interval, random perturbations with a specified standard deviation are injected into the horizontal ciliary sulcus diameter (h-STS) and the vertical ciliary sulcus diameter (v-STS). Monte Carlo sampling is used to construct multiple sets of possible true ciliary sulcus diameters. Then, the worst vertex margin is calculated for each rectangular occupant geometry sample, each angle, and each size, thereby obtaining the fit probability of the candidate scheme under measurement uncertainty.

[0058] In this embodiment, the step of determining multiple candidate schemes based on the candidate size-axis set to satisfy the geometric adaptation index that the rectangular occupant geometry is located within the intraocular geometric constraint region includes:

[0059] An elliptical equation is constructed corresponding to the intraocular geometric constraint region, and the normalized position in the elliptical equation is obtained.

[0060] For each size of the rectangular occupant geometry of any candidate angle in the candidate size-axis set, determine whether the four vertices of the rectangle are all in the normalized position of the ellipse equation.

[0061] When all four vertices of the rectangle are located in the normalized position in the ellipse equation, it is determined that the rectangular occupant geometry of the corresponding size at the candidate angle can be located within the intraocular geometric constraint region.

[0062] Obtain geometric adaptation indicators for the rectangular occupant geometry that can be located within the intraocular geometric constraint region, and form multiple candidate schemes.

[0063] The method further includes:

[0064] By retrospectively calibrating the hyperparameters for forming multiple candidate protocols using the actual implantation size of historical cases and the target follow-up arch height more than 7 days post-operation.

[0065] Evaluation metrics include: whether the actual implantation size is covered by the candidate space; whether the actual implantation axis is within the feasible range; the average number of candidates retained; the candidate space reduction ratio; and the difference in fit probability between good and bad arch heights.

[0066] Retrospective calibration supports grid search, subsampling, parallel geometric computation, and an efficient calibration mechanism that "caches geometry first and then quickly scans the threshold".

[0067] In this embodiment, the step of predicting the postoperative arch height, the probability of abnormal arch height, and the postoperative refractive result of each candidate scheme based on the ocular biological measurement parameters and the geometric fit index of the candidate schemes, and performing post-processing calibration to output the prediction results of each candidate scheme with confidence intervals or risk probabilities, includes:

[0068] Based on historical data of intraocular lens implantation, a prediction model is integrated by stacking multi-base learners, and the prediction model is trained by adding missing indicator variables;

[0069] Based on the ocular biological measurement parameters and the geometric fit index of the candidate schemes, the prediction model is used to predict the postoperative arch height, the probability of abnormal arch height, and the postoperative refractive results of each candidate scheme.

[0070] The risk of the arch height deviating from the safe value is determined based on the predicted postoperative arch height of each candidate treatment, and then converted into the probability of abnormal arch height.

[0071] The postoperative refractive results are vectorized and modeled to obtain postoperative refractive values, wherein the postoperative refractive values ​​include spherical equivalent (SE), astigmatism vector J0, and astigmatism vector J45;

[0072] Based on the postoperative refractive value and the expected postoperative refractive value, a residual target is constructed, and the refractive residual is predicted by the residual target to determine the postoperative refractive result;

[0073] The predicted postoperative arch height, abnormal arch height risk probability, and postoperative refractive results of each candidate scheme are post-processed and calibrated, and the prediction results of each candidate scheme with confidence interval or risk probability are output.

[0074] When training the prediction model by adding missing indicator variables, different missing value handling strategies can be adopted in different prediction models, for example:

[0075] For models that support missing values, NaN is directly retained;

[0076] Median imputation is performed on models that do not support missing values;

[0077] In ensemble models, training set statistics are used to fill in gaps for some base learners.

[0078] To improve clinical relevance and interpretability, the system constructs multiple high-level features, including but not limited to:

[0079] Higher-order interaction features: such as ACD × ICL_size × WTW; where ICL_size is the size of the artificial crystal;

[0080] Biomechanical characteristics: such as compression ratio, lens crowding index, anterior segment ratio, and STS asymmetry;

[0081] Time dynamic characteristics: such as exponential decay characteristics and square root time characteristics;

[0082] Individual normalization characteristics: such as the normalization of the size of the intraocular lens relative to WTW, and the normalization of ACD relative to AL;

[0083] Surgical fit index: such as the difference between ICL_size and the average ciliary sulcus diameter;

[0084] The published empirical formula output is used as a priori feature.

[0085] Among them, a prediction model is formed by stacking multi-base learners based on historical data of intraocular lens implantation, with the following multi-model combination being preferred:

[0086] Quantile regression model: outputs p10 / p50 / p90 quantile predictions, used for interval and uncertainty estimation;

[0087] Point estimation model: Output expected arch height;

[0088] Multi-model stacked ensemble model: The OOF (Out of Fold) predictions of multiple base learners such as CatBoost, XGBoost, LightGBM, Random Forest, ExtraTrees, HistGBM, and TabPFN are used as input, and then a meta-learner outputs a more stable point estimate.

[0089] Among them, the probability of abnormal arch height includes:

[0090] Low arching poses a high risk (e.g., <250 μm);

[0091] High arches pose a high risk (e.g., >750 μm).

[0092] To this end, in addition to calculating probabilities using quantile intervals, the system can also train a dedicated safety classifier to directly output low / high risk probabilities and prioritize the use of classifier probabilities during inference to improve risk identification capabilities.

[0093] Furthermore, a mechanism of "empirical formula + machine learning residual correction" is introduced. For example, the predicted value of a known clinical formula (such as the NK formula) is first calculated, and then the residual of "true arch height - formula predicted value" is learned to improve the model's ability to fit complex individual differences. When the coverage of the empirical formula reaches the preset requirement, this residual learning route is adopted first.

[0094] First, the expected postoperative value is calculated based on the preoperative refractive and intraocular lens plan. Then, the residual target (actual value - expected value) is constructed. The residual is predicted by machine learning. Finally, the residual is added back to the expected value to obtain the final predicted value.

[0095] Since the absolute values ​​of J0 / J45 are prone to over-dispersion or scaling bias, a linear shrinkage model of "predicted absolute value → actual absolute value" is further established on the calibration set to obtain slope, intercept, and residual standard deviation. This model is then used to correct point prediction and reconstruct the prediction interval, thereby improving the usability of the astigmatic vector results.

[0096] Post-processing calibration was performed on the quantile prediction results on an independent calibration set to correct the linear bias of the prediction center. Normalized CQR (Conformal style) parameters were used to expand or shrink the prediction interval so that the interval coverage was closer to the clinical target confidence level.

[0097] For critical tail risk samples such as those with low or high camber, data augmentation under medical constraints can be performed, including:

[0098] Gaussian perturbation based on device accuracy;

[0099] Oversampling of tail samples;

[0100] Mixup interpolation generates intermediate cases;

[0101] Mild noise regularization for regular samples.

[0102] In this embodiment, ranking the candidate schemes according to the prediction results includes:

[0103] For each size of the rectangular occupant geometry, the probability of satisfying the upper and lower limits of the margin, the median worst-case margin, the rotation sensitivity, the asymmetry of the four corner vertices of the rectangular occupant geometry, and the width of the feasible angle interval are calculated to form scoring parameters.

[0104] The candidate solutions are graded according to the scoring parameters, sorted according to the grading results, and the feasible size, the feasible angle range corresponding to each size, the optimal recommended angle, the adaptation probability and stability score are output.

[0105] In this embodiment, the step of classifying each candidate solution according to the scoring parameters includes:

[0106] Calculate the geometric score, arch height safety score, and refractive accuracy score based on the scoring parameters.

[0107] An uncertainty penalty is set up, where a reward or no penalty is given when the difference between the candidate angle and the horizontal (0°) or vertical (90°) angle is less than a first angle; when the difference between the candidate angle and the horizontal (0°) or vertical (90°) angle is greater than or equal to the first angle, the angle is adjusted by rotation and penalized according to the angle deviation from the horizontal or vertical angle; when there is an abnormal area in the user's eyeball, a penalty is imposed based on the boundary distance between the rectangular occupier geometry and the abnormal area in the eyeball.

[0108] Set placement orientation preferences;

[0109] The comprehensive score of each candidate scheme is determined based on the geometric score, the arch height safety score, the refractive accuracy score, the uncertainty penalty item, and the placement orientation preference item, and each candidate scheme is classified according to the comprehensive score.

[0110] Among them, such as Figure 2 , Figure 3 As shown, the diameter of the horizontal ciliary sulcus and the diameter of the vertical ciliary sulcus are respectively used as the major axis or minor axis of an ellipse to construct an elliptical intraocular geometric constraint region, in which a rectangular artificial lens is implanted. Figure 2 This is a schematic diagram of the structure when the difference between the candidate angle of the artificial lens relative to the eyeball and the horizontal angle is 0°. At this time, the horizontal axis of symmetry and the vertical axis of symmetry of the artificial lens are horizontal and vertical, respectively. That is, the horizontal axis of symmetry and the vertical axis of symmetry of the artificial lens coincide with the horizontal ciliary sulcus diameter (h-STS) and the vertical ciliary sulcus diameter (v-STS), respectively. Figure 3 This is a schematic diagram of the structure when the difference between the candidate angle of the artificial lens relative to the eyeball and the horizontal angle is not 0°. At this time, the horizontal and vertical axes of symmetry of the artificial lens form an angle with the horizontal ciliary sulcus diameter (h-STS) or the vertical ciliary sulcus diameter (v-STS), respectively. This angle is used as the rotation angle of the artificial lens relative to the eyeball. Within the range of less than 45°, the larger the angle, the more the artificial lens implantation angle deviates from the horizontal or vertical angle.

[0111] In this embodiment, the step of outputting at least one recommended solution based on the sorting order and the geometric fit index of each candidate solution includes:

[0112] Based on the sorting order, the recommended scheme with horizontal or vertical angle implantation is selected as the optimal recommended scheme.

[0113] Based on the sorting order, the first candidate solution is the one with the same rectangular placeholder geometry size as the optimal recommended solution and whose candidate angle is the smallest rotation angle from the horizontal or vertical direction.

[0114] Based on the sorting order, the recommended solutions with candidate angles of horizontal or vertical implantation and adjacent to the rectangular occupant geometry of the optimal recommended solution are selected as the second alternative solutions.

[0115] Based on the sorting order, the recommended solution with the smallest rotation angle from the horizontal or vertical direction is selected as the third alternative.

[0116] The output should include at least one recommended treatment plan, and the postoperative results output should include at least the following formats:

[0117] Postoperative arch height point estimation and uncertainty interval (e.g., median and upper and lower quantiles).

[0118] Low arch high risk and high arch high risk probability (corresponding to preset thresholds).

[0119] Postoperative refractive outcome prediction should include at least point estimates and uncertainties for SE, J0, and J45;

[0120] The calibrated probability or calibrated interval of the predicted result (used for clinically interpretable risk indication).

[0121] Each candidate solution is first screened under hard constraints, and then ranked based on a comprehensive score.

[0122] Hard constraints (configurable) include:

[0123] (a) Geometric feasibility constraints: p_fit is not lower than a preset threshold, and the candidate angle does not fall into the manually restricted area;

[0124] (b) Safety constraints: The probabilities of low Vault risk and high Vault risk do not exceed a preset upper limit, or the prediction range and the target Vault range have sufficient overlap;

[0125] (c) Refractive constraint (optional): Predict that the residual SE / astigmatism amplitude does not exceed a preset threshold.

[0126] The rules for generating the optimal recommended solution, the first alternative solution, and the second alternative solution are as follows:

[0127] Optimal Recommendation Solution: Among all candidate solutions that meet the hard constraints, prioritize the set of solutions that can be placed horizontally / vertically (candidate angles belong to {0°, 90°} or their neighborhood window), and select the one with the highest comprehensive score within this set as the optimal recommendation solution;

[0128] Alternative solutions: From all candidate solutions that satisfy the hard constraints, select several alternative solutions in descending order of comprehensive score.

[0129] Preferably, the alternative solutions include at least:

[0130] Alternative A: Same size as the optimal recommended solution but with a different feasible angle (if it exists);

[0131] Alternative B: The size is adjacent to the optimal recommended solution (one size larger or smaller) and can be placed horizontally or vertically (if it exists);

[0132] Option C: When none of the above options are available, select the option with the highest overall score from the rotation angle options as the rotation alternative;

[0133] Diversity constraint (optional): Alternative options must differ in size and angle to avoid all options being too similar;

[0134] Output content: The optimal recommended solution and each alternative solution are given the reasons for the recommendation, including: the predicted arch height range and risk probability, the predicted refractive result, the geometric fit probability, and whether to use rotation and restricted area avoidance information.

[0135] When implementing the intraocular lens selection and rotation implantation prediction method, follow these steps:

[0136] Step S11: Input the patient's preoperative parameters, the candidate ICL inventory set (size / power / astigmatism parameters), and the optional manual restricted areas;

[0137] Step S12: Based on h-STS / v-STS, wtw, etc., establish an elliptical groove diameter model, and approximate the ICL as a rectangular occupier with a diagonal length equal to the nominal size of the ICL. Scan the angle from 0° to 180° and consider rotational drift, and output the feasible angle range, optimal angle, p_fit, rotational sensitivity and vertex asymmetry and other geometric indicators corresponding to each size.

[0138] Step S13: Subtract the restricted area from the feasible angle range and perform boundary buffering to obtain the final geometric candidate space;

[0139] Step S14: For each "size-angle" candidate scheme in the geometric candidate space, predict the postoperative vault, low / high vault risk, SE / J0 / J45, and calibrate the probabilities and intervals;

[0140] Step S15: Apply hard constraint screening (geometric + safety + refractive) and calculate the comprehensive score;

[0141] Step S16: Output the first recommended solution and several alternative solutions according to the strategy of "horizontal / vertical priority - rotation alternative - adjacent size rollback".

[0142] The system employs a combined adjustment strategy of size and rotation (selection logic for smaller / larger / rotation). To ensure the decision-making process is interpretable and reproducible, the system adopts a hierarchical adjustment strategy of "angle first, size second."

[0143] (1) Prioritize same size: For a candidate size, prioritize horizontal / vertical placement (0° / 90° and its neighborhood window); if the Vault target range or risk constraint is not met, search for rotation angles that can improve the Vault deviation risk within the feasible angle range of the size;

[0144] (2) Retreat from adjacent dimensions: If the constraint cannot be satisfied even if rotated within the same dimension, select the adjacent dimension (one size smaller or one size larger) and repeat (1).

[0145] When the predicted Vault is too high or the risk of a high Vault is too high, prioritize trying a "smaller" size and prioritize horizontal / vertical; if this is still not satisfied, then allow rotation within its feasible range;

[0146] When the predicted Vault is too low or the risk of low Vault is high, prioritize trying a "larger" size and prioritize horizontal / vertical; if this is still not satisfied, then allow rotation within its feasible range.

[0147] (3) Rotation as an alternative: The system will output rotation as an explicit alternative strategy only when 1) no horizontal / vertical solution that satisfies the constraints can be obtained in the target size and adjacent sizes. 2) The astigmatic axis of the artificial lens in the inventory is inconsistent with the astigmatic axis of the required artificial lens. In order to match the astigmatic axis of the artificial lens in the inventory with the astigmatism that needs to be corrected, it is necessary to rotate the artificial lens for implantation. As long as the size of the artificial lens matches the ellipse diameter of the intraocular geometric constraint area where the major axis is located after rotation, it can also be used as an alternative solution.

[0148] The selection and ranking rules for alternative solutions include:

[0149] Rule R1 (Feasibility): Alternative solutions must meet hard constraints, especially that the predicted Vault is within the target range or its probability of abnormal risk does not exceed the upper limit; at the same time, the candidate angles do not fall into forbidden areas;

[0150] Rule R2 (Preference): Under the premise of satisfying R1, prioritize outputting solutions that can be placed horizontally / vertically; solutions with rotation angles are considered as lower-level candidates.

[0151] Rule R3 (Sorting): Output candidates in descending order of overall score (e.g., output 2–5). The overall score must include at least: geometric p_fit, Vault bias risk probability, refractive residual, uncertainty penalty, vertex asymmetry, and deviation penalty from {0°, 90°};

[0152] Rule R4 (Diversity): Alternative options should be as different as possible in size and angle (e.g., include at least one "same size, different angle" and one "adjacent size" alternative) so that doctors still have viable alternatives when inventory / surgical preferences change.

[0153] Among them, when implanting a rotating intraocular lens, as long as the size of the intraocular lens matches the ellipse diameter of the intraocular geometric constraint region where the major axis is located after rotation, it can also be used as an alternative.

[0154] The aforementioned method for predicting intraocular lens (IOL) selection and rotation involves acquiring the user's ocular biological measurement parameters and constructing an intraocular geometric constraint region. The IOL is then abstracted into a rectangular geometries, generating a set of candidate sizes and axes. This ensures that each candidate option meets the geometric constraints of the patient's intraocular anatomy in terms of implantation position and rotation angle. Based on this, by combining the geometric fit indices of the candidate options with individualized ocular biological parameters, the method predicts postoperative arch height, the probability of abnormal arch height, and postoperative refractive results. Post-processing and calibration provide outputs with confidence intervals or risk probabilities, eliminating reliance on experience or rough estimates for surgical planning. This method not only rationally determines the optimal size and rotation angle of the IOL preoperatively but also accurately assesses the postoperative arch height and refractive results of each candidate option. This effectively reduces the risk of arch height deviation or postoperative refractive errors, providing the optimal IOL selection option and achieving individualized, quantifiable, and predictable implantation planning. It solves the problems of traditional methods, such as the inability to accurately determine the size and angle of IOL selection and rotation, and inaccurate postoperative feedback.

[0155] Compared with the prior art, the present invention has at least the following advantages:

[0156] Decoupling geometric adaptation from outcome prediction: First, use a geometric model to quickly narrow down the candidate space, and then use a data-driven model for fine screening, which improves efficiency and reduces the unreliability of pure black-box decision-making.

[0157] Explicitly considering the diameter of the elliptical ciliary sulcus and rotational drift: Instead of simplifying the intraocular space to a single linear size, h-STS and v-STS are incorporated together, and rotational bandwidth is taken into account, which can more closely reflect the actual implantation situation.

[0158] Outputting a set of candidates instead of rigid single-point suggestions allows clinicians to make flexible choices based on inventory, surgical habits, and special circumstances.

[0159] It can output probabilistic risks and uncertainties: through quantile prediction, calibration and safety classifier, it can directly tell doctors "how much risk is associated with low arches" and "how much risk is associated with high arches".

[0160] It can handle missing and noise in real-world clinical data: it improves robustness through mechanisms such as missing data indication, native NaN support, median imputation, and medical constraint enhancement.

[0161] Directly deployed in clinical software systems: Existing implementations have adopted engineering mechanisms such as task-based pipelines, checkpoint saving, and GPU-isolated subprocesses, making them suitable for deployment in hospital or enterprise products.

[0162] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method.

[0163] In one embodiment, such as Figure 4 As shown, an intraocular lens selection and rotation implantation prediction device 10 is provided, including: an input module 1, a geometric candidate generation module 2, a postoperative result prediction module 3, a calibration module 4, and a recommendation output module 5.

[0164] The input module 1 is used to obtain the user's ocular biological measurement parameters and construct an intraocular geometric constraint region based on the ocular biological measurement parameters.

[0165] The geometric candidate generation module 2 is used to abstract the artificial crystal into a rectangular occupier geometry and generate a candidate size-axis set based on the size of the artificial crystal and the rotation angle around its center.

[0166] The postoperative outcome prediction module 3 is used to determine multiple candidate schemes based on the candidate size-axis set to satisfy the geometric fit index that the rectangular occupant geometry is located within the intraocular geometric constraint area.

[0167] The calibration module 4 is used to predict the postoperative arch height, the probability of abnormal arch height, and the postoperative refractive results of each candidate scheme based on the ocular biological measurement parameters and the geometric adaptation index of the candidate schemes, and to perform post-processing calibration to output the prediction results of each candidate scheme with confidence intervals or risk probabilities.

[0168] The recommendation output module 5 is used to sort the candidate solutions according to the prediction results, and output at least one recommended solution according to the sorting order and the geometric fit index of each candidate solution.

[0169] In this embodiment, obtaining the user's ocular biological measurement parameters and constructing an intraocular geometrically constrained region based on the ocular biological measurement parameters includes:

[0170] The user's ocular biological measurement parameters include one or more of the following: horizontal ciliary sulcus diameter (h-STS), vertical ciliary sulcus diameter (v-STS), horizontal corneal diameter (white to white, WTW), anterior chamber depth (ACD), axial length (AL), lens thickness, corneal thickness, corneal curvature, axial length, pupil diameter, and refractive data.

[0171] Identify whether there is an abnormal region within the groove of the user's eyeball; if so, represent the abnormal region within the groove by clock position or angle range.

[0172] Determine whether the data for the horizontal ciliary sulcus diameter (h-STS) and the vertical ciliary sulcus diameter (v-STS) are missing;

[0173] If the data for the horizontal ciliary sulcus diameter (h-STS) and the vertical ciliary sulcus diameter (v-STS) are not missing, then the horizontal ciliary sulcus diameter (h-STS) and the vertical ciliary sulcus diameter (v-STS) are respectively used as the major axis or minor axis of the ellipse to construct the intraocular geometric constraint region.

[0174] If either the horizontal ciliary sulcus diameter (h-STS) or the vertical ciliary sulcus diameter (v-STS) is missing, the data of the vertical ciliary sulcus diameter (v-STS) or the horizontal ciliary sulcus diameter (h-STS) is used as a substitute, and the intraocular geometric constraint region is constructed using the substitute data.

[0175] If the data for both the horizontal ciliary sulcus diameter (h-STS) and the vertical ciliary sulcus diameter (v-STS) are missing, the intraocular geometric constraint region is calculated based on the horizontal corneal diameter, wherein the length of the major or minor axis of the ellipse is not limited.

[0176] In this embodiment, the artificial lens is abstracted as a rectangular placeholder geometry, the diagonal length of which is equal to the nominal size of the artificial lens. A candidate size-axis set is generated based on the size of the artificial lens and the rotation angle around its center, including:

[0177] The artificial lens is abstracted as a rectangular occupier geometry, and the diagonal length of the rectangular occupier geometry is set to be equal to the nominal size of the artificial lens;

[0178] The rectangular placeholder geometry is rotated around its center, and multiple candidate placement angles within the range of 0°–180° are scanned. Each candidate angle is associated with the size of the rectangular placeholder geometry to form a candidate size-axis set.

[0179] In this embodiment, when associating each candidate angle with the size of the rectangular occupier geometry to form a candidate size-axis set, the method further includes:

[0180] A rotation drift interval is set for each candidate angle, and the geometric adaptation margin of rectangular occupant geometries of multiple sizes is sampled within the rotation drift interval. The geometric adaptation margin of the rectangular occupant geometries of each size at each candidate angle is statistically analyzed.

[0181] The rotational drift range after intraocular lens surgery is simulated based on the geometric fit margin of the rectangular occupant geometry of each size at each candidate angle.

[0182] The rotational drift range is added when forming the candidate size-axis set.

[0183] In this embodiment, the step of determining multiple candidate schemes based on the candidate size-axis set to satisfy the geometric adaptation index that the rectangular occupant geometry is located within the intraocular geometric constraint region includes:

[0184] An elliptical equation is constructed corresponding to the intraocular geometric constraint region, and the normalized position in the elliptical equation is obtained.

[0185] For each size of the rectangular occupant geometry of any candidate angle in the candidate size-axis set, determine whether the four vertices of the rectangle are all in the normalized position of the ellipse equation.

[0186] When all four vertices of the rectangle are located in the normalized position in the ellipse equation, it is determined that the rectangular occupant geometry of the corresponding size at the candidate angle can be located within the intraocular geometric constraint region.

[0187] Obtain geometric adaptation indicators for the rectangular occupant geometry that can be located within the intraocular geometric constraint region, and form multiple candidate schemes.

[0188] In this embodiment, the step of predicting the postoperative arch height, the probability of abnormal arch height, and the postoperative refractive result of each candidate scheme based on the ocular biological measurement parameters and the geometric fit index of the candidate schemes, and performing post-processing calibration to output the prediction results of each candidate scheme with confidence intervals or risk probabilities, includes:

[0189] Based on historical data of intraocular lens implantation, a prediction model is integrated by stacking multi-base learners, and the prediction model is trained by adding missing indicator variables;

[0190] Based on the ocular biological measurement parameters and the geometric fit index of the candidate schemes, the prediction model is used to predict the postoperative arch height, the probability of abnormal arch height, and the postoperative refractive results of each candidate scheme.

[0191] The risk of the arch height deviating from the safe value is determined based on the predicted postoperative arch height of each candidate treatment, and then converted into the probability of abnormal arch height.

[0192] The postoperative refractive results are vectorized and modeled to obtain postoperative refractive values, wherein the postoperative refractive values ​​include spherical equivalent (SE), astigmatism vector J0, and astigmatism vector J45;

[0193] Based on the postoperative refractive value and the expected postoperative refractive value, a residual target is constructed, and the refractive residual is predicted by the residual target to determine the postoperative refractive result;

[0194] The predicted postoperative arch height, abnormal arch height risk probability, and postoperative refractive results of each candidate scheme are post-processed and calibrated, and the prediction results of each candidate scheme with confidence interval or risk probability are output.

[0195] In this embodiment, ranking the candidate schemes according to the prediction results includes:

[0196] For each size of the rectangular occupant geometry, the probability of satisfying the upper and lower limits of the margin, the median worst-case margin, the rotation sensitivity, the asymmetry of the four corner vertices of the rectangular occupant geometry, and the width of the feasible angle interval are calculated to form scoring parameters.

[0197] The candidate solutions are graded according to the scoring parameters, sorted according to the grading results, and the feasible size, the feasible angle range corresponding to each size, the optimal recommended angle, the adaptation probability and stability score are output.

[0198] In this embodiment, the step of classifying each candidate solution according to the scoring parameters includes:

[0199] Calculate the geometric score, arch height safety score, and refractive accuracy score based on the scoring parameters.

[0200] An uncertainty penalty is set up, where a reward or no penalty is given when the difference between the candidate angle and the horizontal (0°) or vertical (90°) angle is less than a first angle; when the difference between the candidate angle and the horizontal (0°) or vertical (90°) angle is greater than or equal to the first angle, the angle is adjusted by rotation and penalized according to the angle deviation from the horizontal or vertical angle; when there is an abnormal area in the user's eyeball, a penalty is imposed based on the boundary distance between the rectangular occupier geometry and the abnormal area in the eyeball.

[0201] Set placement orientation preferences;

[0202] The comprehensive score of each candidate scheme is determined based on the geometric score, the arch height safety score, the refractive accuracy score, the uncertainty penalty item, and the placement orientation preference item, and each candidate scheme is classified according to the comprehensive score.

[0203] In this embodiment, the step of outputting at least one recommended solution based on the sorting order and the geometric fit index of each candidate solution includes:

[0204] Based on the sorting order, the recommended scheme with horizontal or vertical angle implantation is selected as the optimal recommended scheme.

[0205] Based on the sorting order, the first candidate solution is the one with the same rectangular placeholder geometry size as the optimal recommended solution and whose candidate angle is the smallest rotation angle from the horizontal or vertical direction.

[0206] Based on the sorting order, the recommended solutions with candidate angles of horizontal or vertical implantation and adjacent to the rectangular occupant geometry of the optimal recommended solution are selected as the second alternative solutions.

[0207] Based on the sorting order, the recommended solution with the smallest rotation angle from the horizontal or vertical direction is selected as the third alternative.

[0208] The aforementioned intraocular lens (IOL) selection and rotation implantation prediction device acquires the user's ocular biological measurement parameters and constructs an intraocular geometric constraint region. It then abstracts the IOL as a rectangular geometries and generates a set of candidate sizes and axes, ensuring that each candidate option meets the geometric constraints of the patient's intraocular anatomy in terms of implantation position and rotation angle. Based on this, by combining the geometric fit indices of the candidate options with individualized ocular biological parameters, it predicts postoperative arch height, the probability of abnormal arch height, and postoperative refractive results. After post-processing and calibration, it provides outputs with confidence intervals or risk probabilities, freeing surgical planning from reliance on experience or rough estimates. This method not only rationally determines the optimal size and rotation angle of the IOL preoperatively but also accurately assesses the postoperative arch height and refractive results of each candidate option, effectively reducing the risk of arch height deviation or substandard postoperative refractive results. It provides the optimal IOL option, achieving individualized, quantifiable, and predictable implantation planning. This solves the problems of traditional methods in accurately determining IOL size and angle during selection and rotation implantation, as well as inaccurate postoperative feedback.

[0209] For a description of the features in the embodiment of the intraocular lens selection and rotation implantation prediction device, please refer to the relevant description of the embodiment of the intraocular lens selection and rotation implantation prediction method, which will not be repeated here.

[0210] Embodiments of this application also provide an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to perform the steps in any of the above embodiments of the intraocular lens selection and rotational implantation prediction method.

[0211] In one embodiment, the electronic device may be a server, and its internal structure diagram may be as follows: Figure 5 As shown, the electronic device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computational and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores intraocular lens (IOL) selection and rotation implantation prediction data. The network interface communicates with external terminals via a network connection. When the computer program is executed by the processor, it implements an IOL selection and rotation implantation prediction method.

[0212] Embodiments of this application also provide a computer-readable storage medium storing a computer program, wherein the computer program is configured to execute the steps in any of the above embodiments of the intraocular lens selection and rotational implantation prediction method when running:

[0213] Obtain the user's ocular biological measurement parameters, and construct an intraocular geometric constraint region based on the ocular biological measurement parameters;

[0214] The artificial lens is abstracted as a rectangular occupier geometry, and a candidate size-axis set is generated based on the size of the artificial lens and the rotation angle around its center.

[0215] Based on the candidate size-axis set, multiple candidate schemes are formed by determining the geometric adaptation index that satisfies the requirement that the rectangular occupant geometry is located within the intraocular geometric constraint region.

[0216] Based on the ocular biological measurement parameters and the geometric fit index of the candidate schemes, the postoperative arch height, the probability of abnormal arch height, and the postoperative refractive results of each candidate scheme are predicted, and post-processing calibration is performed to output the prediction results of each candidate scheme with confidence intervals or risk probabilities.

[0217] Based on the prediction results, each candidate solution is ranked, and at least one recommended solution is output according to the ranking order and the geometric fit index of each candidate solution.

[0218] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard disk, magnetic disk, or optical disk.

[0219] Embodiments of this application also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the above embodiments of the intraocular lens selection and rotational implantation prediction method:

[0220] Obtain the user's ocular biological measurement parameters, and construct an intraocular geometric constraint region based on the ocular biological measurement parameters;

[0221] The artificial lens is abstracted as a rectangular occupier geometry, and a candidate size-axis set is generated based on the size of the artificial lens and the rotation angle around its center.

[0222] Based on the candidate size-axis set, multiple candidate schemes are formed by determining the geometric adaptation index that satisfies the requirement that the rectangular occupant geometry is located within the intraocular geometric constraint region.

[0223] Based on the ocular biological measurement parameters and the geometric fit index of the candidate schemes, the postoperative arch height, the probability of abnormal arch height, and the postoperative refractive results of each candidate scheme are predicted, and post-processing calibration is performed to output the prediction results of each candidate scheme with confidence intervals or risk probabilities.

[0224] Based on the prediction results, each candidate solution is ranked, and at least one recommended solution is output according to the ranking order and the geometric fit index of each candidate solution.

[0225] Embodiments of this application also provide another computer program product, including a non-volatile computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the steps in any of the above embodiments of the intraocular lens selection and rotational implantation prediction method:

[0226] Obtain the user's ocular biological measurement parameters, and construct an intraocular geometric constraint region based on the ocular biological measurement parameters;

[0227] The artificial lens is abstracted as a rectangular occupier geometry, and a candidate size-axis set is generated based on the size of the artificial lens and the rotation angle around its center.

[0228] Based on the candidate size-axis set, multiple candidate schemes are formed by determining the geometric adaptation index that satisfies the requirement that the rectangular occupant geometry is located within the intraocular geometric constraint region.

[0229] Based on the ocular biological measurement parameters and the geometric fit index of the candidate schemes, the postoperative arch height, the probability of abnormal arch height, and the postoperative refractive results of each candidate scheme are predicted, and post-processing calibration is performed to output the prediction results of each candidate scheme with confidence intervals or risk probabilities.

[0230] Based on the prediction results, each candidate solution is ranked, and at least one recommended solution is output according to the ranking order and the geometric fit index of each candidate solution.

[0231] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0232] The above provides a detailed description of the intraocular lens selection and rotational implantation prediction method and electronic device provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only intended to help understand the method and core ideas of this application. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of the claims of this application.

Claims

1. A method for predicting the selection and rotational implantation of an intraocular lens, characterized in that, include: Obtain the user's ocular biological measurement parameters, and construct an intraocular geometric constraint region based on the ocular biological measurement parameters; The artificial lens is abstracted as a rectangular occupier geometry, and a candidate size-axis set is generated based on the size of the artificial lens and the rotation angle around its center. Based on the candidate size-axis set, multiple candidate schemes are formed by determining the geometric adaptation index that satisfies the requirement that the rectangular occupant geometry is located within the intraocular geometric constraint region. Based on the ocular biological measurement parameters and the geometric fit index of the candidate schemes, the postoperative arch height, the probability of abnormal arch height, and the postoperative refractive results of each candidate scheme are predicted, and post-processing calibration is performed to output the prediction results of each candidate scheme with confidence intervals or risk probabilities. Based on the prediction results, each candidate solution is ranked, and at least one recommended solution is output according to the ranking order and the geometric fit index of each candidate solution. Specifically, when generating the candidate size-axis set, a rotation drift interval is set for each candidate angle, and geometric adaptation margins are sampled for rectangular occupant geometries of multiple sizes within the rotation drift interval; The step of ranking the candidate solutions based on the prediction results includes: For each size of the rectangular occupant geometry, the probability of satisfying the upper and lower limits of the margin, the median worst-case margin, the rotation sensitivity, the asymmetry of the four corner vertices of the rectangular occupant geometry, and the width of the feasible angle interval are calculated to form scoring parameters. The candidate solutions are graded according to the scoring parameters, and the candidate solutions are sorted according to the grading results. The feasible size, the feasible angle range corresponding to each size, the optimal recommended angle, the adaptation probability and stability score are output. The step of classifying each candidate solution according to the scoring parameters includes: Calculate the geometric score, arch height safety score, and refractive accuracy score based on the scoring parameters. An uncertainty penalty is set up, in which a reward or no penalty is given when the difference between the candidate angle and the horizontal or vertical angle is less than the first angle; when the difference between the candidate angle and the horizontal or vertical angle is greater than or equal to the first angle, the angle is adjusted by rotation and a penalty is imposed according to the amount of deviation from the horizontal or vertical angle; when there is an abnormal area in the groove of the user's eyeball, a penalty is imposed according to the boundary distance between the rectangular occupier geometry and the abnormal area in the groove. Set placement orientation preferences; The comprehensive score of each candidate scheme is determined based on the geometric score, the arch height safety score, the refractive accuracy score, the uncertainty penalty item, and the placement orientation preference item, and each candidate scheme is classified according to the comprehensive score.

2. The method for predicting intraocular lens selection and rotational implantation according to claim 1, characterized in that, The step of acquiring the user's ocular biological measurement parameters and constructing an intraocular geometrically constrained region based on the ocular biological measurement parameters includes: The user's ocular biological measurement parameters include one or more of the following: horizontal ciliary sulcus diameter, vertical ciliary sulcus diameter, horizontal corneal diameter, anterior chamber depth, axial length, lens thickness, corneal thickness, corneal curvature, axial length, pupil diameter, and refractive data. Identify whether there is an abnormal region within the groove of the user's eyeball; if so, represent the abnormal region within the groove by clock position or angle range. Determine whether the data for the horizontal ciliary sulcus diameter and the vertical ciliary sulcus diameter are missing; If the data for the horizontal ciliary sulcus diameter and the vertical ciliary sulcus diameter are not missing, then the horizontal ciliary sulcus diameter and the vertical ciliary sulcus diameter are respectively used as the major axis or minor axis of the ellipse to construct the intraocular geometric constraint region; If either the horizontal ciliary sulcus diameter or the vertical ciliary sulcus diameter is missing, the data of the vertical ciliary sulcus diameter or the horizontal ciliary sulcus diameter shall be used as a substitute, and the substitute data shall be used to construct the intraocular geometric constraint region. If the data for both the horizontal ciliary sulcus diameter and the vertical ciliary sulcus diameter are missing, the intraocular geometric constraint region is calculated based on the horizontal corneal diameter, wherein the length of the major or minor axis of the ellipse is not limited.

3. The method for predicting intraocular lens selection and rotational implantation according to claim 1, characterized in that, The process involves abstracting the intraocular lens as a rectangular occupier geometry, where the diagonal length of the rectangular occupier geometry is equal to the nominal size of the intraocular lens. A candidate size-axis set is generated based on the size of the intraocular lens and the rotation angle around its center, including: The artificial lens is abstracted as a rectangular occupier geometry, and the diagonal length of the rectangular occupier geometry is set to be equal to the nominal size of the artificial lens; The rectangular placeholder geometry is rotated around its center, and multiple candidate placement angles within the range of 0°–180° are scanned. Each candidate angle is associated with the size of the rectangular placeholder geometry to form a candidate size-axis set.

4. The method for predicting intraocular lens selection and rotational implantation according to claim 3, characterized in that, When associating each candidate angle with the dimensions of the rectangular occupier geometry to form a candidate size-axis set, the method further includes: The geometric fit margins of the rectangular occupant geometry of each size at each candidate angle are statistically analyzed. The rotational drift range after intraocular lens surgery is simulated based on the geometric fit margin of the rectangular occupant geometry of each size at each candidate angle. The rotational drift range is added when forming the candidate size-axis set.

5. The method for predicting intraocular lens selection and rotational implantation according to claim 1, characterized in that, The process involves determining geometric adaptation indices based on the candidate size-axis set to ensure the rectangular occupant geometry is located within the intraocular geometric constraint region, forming multiple candidate schemes, including: An elliptical equation is constructed corresponding to the intraocular geometric constraint region, and the normalized position in the elliptical equation is obtained. For each size of the rectangular occupant geometry of any candidate angle in the candidate size-axis set, determine whether the four vertices of the rectangle are all in the normalized position of the ellipse equation. When all four vertices of the rectangle are located in the normalized position in the ellipse equation, it is determined that the rectangular occupant geometry of the corresponding size at the candidate angle can be located within the intraocular geometric constraint region. Obtain geometric adaptation indicators for the rectangular occupant geometry that can be located within the intraocular geometric constraint region, and form multiple candidate schemes.

6. The method for predicting intraocular lens selection and rotational implantation according to claim 1, characterized in that, The method predicts the postoperative arch height, risk probability of abnormal arch height, and postoperative refractive results of each candidate scheme based on the ocular biological measurement parameters and the geometric fit index of the candidate schemes, performs post-processing calibration, and outputs the prediction results of each candidate scheme with confidence intervals or risk probabilities, including: Based on historical data of intraocular lens implantation, a prediction model is integrated by stacking multi-base learners, and the prediction model is trained by adding missing indicator variables; Based on the ocular biological measurement parameters and the geometric fit index of the candidate schemes, the prediction model is used to predict the postoperative arch height, the probability of abnormal arch height, and the postoperative refractive results of each candidate scheme. The risk of the arch height deviating from the safe value is determined based on the predicted postoperative arch height of each candidate treatment, and then converted into the probability of abnormal arch height. The postoperative refractive results are vectorized and modeled to obtain postoperative refractive values, wherein the postoperative refractive values ​​include spherical equivalent, astigmatism vector J0, and astigmatism vector J45; Based on the postoperative refractive value and the expected postoperative refractive value, a residual target is constructed, and the refractive residual is predicted by the residual target to determine the postoperative refractive result; The predicted postoperative arch height, abnormal arch height risk probability, and postoperative refractive results of each candidate scheme are post-processed and calibrated, and the prediction results of each candidate scheme with confidence interval or risk probability are output.

7. The method for predicting intraocular lens selection and rotational implantation according to claim 1, characterized in that, The step of outputting at least one recommended solution based on the sorting order and the geometric fit index of each candidate solution includes: Based on the sorting order, the recommended scheme with horizontal or vertical angle implantation is selected as the optimal recommended scheme. Based on the sorting order, the first candidate solution is the one with the same rectangular placeholder geometry size as the optimal recommended solution and whose candidate angle is the smallest rotation angle from the horizontal or vertical direction. Based on the sorting order, the recommended solutions with candidate angles of horizontal or vertical implantation and adjacent to the rectangular occupant geometry of the optimal recommended solution are selected as the second alternative solutions. Based on the sorting order, the recommended solution with the smallest rotation angle from the horizontal or vertical direction is selected as the third alternative.

8. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the steps of the intraocular lens selection and rotational implantation prediction method as described in any one of claims 1 to 7.

Images