A corneal fine residual lens detection system and method based on multi-modal imaging

Abstract

This invention relates to the fields of ophthalmic medical devices and image processing technology, specifically disclosing a system and method for detecting corneal microlenses based on multimodal imaging. The system includes: an image acquisition module for acquiring corneal density maps, anterior surface elevation maps, and anterior segment OCT images of the examined eye after SMILE surgery; a data processing module for identifying focal areas with low corneal optical density values ​​within the expected lenticule removal area from the corneal density map as suspected microlenses (SRL) areas, and locating microbulges spatially related to the suspected SRL area in the anterior surface elevation map; a confirmation module for determining whether lamellar detachment and compensatory thinning of the superior tissue exist at the suspected SRL area based on the anterior segment OCT image, thus confirming the SRL; and a quantification output module for calculating and outputting the area of ​​the SRL, the number of corneal quadrants it occupies, and its morphological classification. This invention solves the problem of the lack of standardized methods for detecting microlenses after SMILE surgery in the prior art.

Description

Technical Field

[0001] This invention belongs to the field of medical image processing and ophthalmic diagnostic equipment technology, specifically relating to a system and method for detecting, confirming and quantifying minute residual lenses at the corneal interface after corneal refractive surgery, especially after small incision lenticule extraction. Background Technology

[0002] Mini-lenticule extraction (MLE), an advanced refractive surgery, has been widely used to correct refractive errors such as myopia. However, during the procedure, incomplete lenticule separation or extraction may occur, resulting in tiny lenticule fragments remaining between the corneal layers. These fragments, especially the minute residual lenses located in the peripheral area, are difficult to detect with conventional slit-lamp examination due to their small size and concealed location.

[0003] Currently, there is a lack of standardized and objective methods in clinical practice for effectively screening and confirming such surgically transmitted lesions (SRLs). Although anterior segmental OCT can provide high-resolution corneal tomographic images, its efficiency for large-area screening alone is low, and it may be insensitive to extremely thin SRLs. Therefore, how to accurately and efficiently identify SRLs and further assess their potential impact on postoperative visual quality (such as higher-order aberrations) has become a pressing technical problem to be solved in this field. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a highly sensitive, specific, and standardized SRL detection system and method to achieve early detection, precise localization, and quantitative assessment of SRL after SMILE surgery.

[0005] Firstly, to achieve the above objectives, the present invention adopts the following technical solution:

[0006] The core of this invention lies in creatively using corneal density measurement as a primary screening tool, and performing multimodal image fusion and collaborative analysis with anterior surface height mapping and anterior segment OCT.

[0007] 1. Screening Phase: Utilizing the high COD background typically presenting in the lenticule extraction area after SMILE surgery, suspicious areas exhibiting low COD "islands" are identified within this background. Simultaneously, the presence of corresponding microbulges is verified in the anterior surface elevation map. This dual screening significantly improves the sensitivity of detecting SRL (Small Inflammatory Retention).

[0008] 2. Confirmation Phase: For suspicious areas identified during screening, confirmation using AS-OCT is mandatory, serving as the gold standard. The confirmation criteria are the presence of clear interstitial separation and compensatory thinning of the overlying tissue at the corresponding interface.

[0009] 3. Quantification phase: For confirmed SRLs, image analysis techniques are used for quantification, including area measurement, quadrant counting, and morphological description, to provide objective data for clinical decision-making.

[0010] 4. Application Correlation: Correlate the results of SRL detection with the risk assessment of postoperative higher-order aberrations (especially horizontal coma in the 8.0 mm analysis area) to provide objective evidence for interpreting possible night vision symptoms in patients.

[0011] The beneficial effects of this invention are as follows: Compared with the prior art, this invention combines the sensitivity of corneal densitometric measurement with the specificity of AS-OCT, constructing a powerful detection process that can detect subtle SRLs that are easily missed by traditional methods, achieving high sensitivity and specificity. Secondly, this invention provides objective measurement indicators for the area, location, and morphology of SRLs, going beyond subjective qualitative descriptions, which is beneficial for clinical research and surgical quality monitoring, achieving objective quantification. This invention is the first to directly link the objective existence of SRLs with specific higher-order aberrations (negative horizontal coma), providing a new technical path for assessing postoperative visual quality and interpreting patient symptoms, and providing a standardized operating procedure that can be promoted across different medical institutions, unifying the diagnostic criteria for SRLs. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 The chart shown is a standard chart for reporting refractive surgery results 3 months post-surgery.

[0014] The comparison shows the eyes with subtle residual lenses (SRL group, A1-F1) and the contralateral eyes without SRL (non-SRL group, A2-F2). (A1, A2) Postoperative cumulative uncorrected visual acuity and preoperative corrected visual acuity; (B1, B2) Difference between postoperative UDVA and preoperative CDVA (effectiveness); (C1, C2) Change in the number of Snellen lines in postoperative CDVA (safety); (D1, D2) Expected and actual spherical equivalent refractive power (predictability); (E1, E2) Accuracy of spherical equivalent refractive power relative to the target; (F1, F2) Stability of spherical equivalent refractive power at 1 month and 3 months postoperatively.

[0015] Abbreviations: CDVA = Corrected visual acuity; D = Diopter; SRL = Slight residual lens; UDVA = Uncorrected visual acuity.

[0016] Figure 2 The figure shows a multiple linear regression analysis: factors influencing changes in higher-order corneal aberrations from pre-operative to 3 months post-operatively.

[0017] (A) Factors influencing Δ6.0mm SA; (B) Factors influencing Δ8.0mm SA; (C) Factors influencing Δ8.0mm HC. Dots represent estimated regression coefficients (β), and horizontal lines represent 95% confidence intervals. Variables assessed included preoperative parameters (corneal diameter, spherical power, cylindrical power, flat corneal curvature, steep corneal curvature), surgically relevant parameters (central corneal thickness, lenticule thickness), and postoperative parameters (presence of minute residual lenticules, effective optical zone area, postoperative spherical power, postoperative cylindrical power).

[0018] Abbreviations: CI = Confidence Interval; D = Refractive Power; HC = Horizontal Coma; SA = Spherical Aberration; Δ = Change from preoperative to 3 months postoperatively.

[0019] Figure 3 The diagram illustrates the proposed mechanism by which a subtle residual lens in the right eye induces changes in higher-order corneal aberrations of 8.0 mm.

[0020] The green plane represents the ideal aberration-free wavefront, and the red plane depicts the actual wavefront; the difference between the two illustrates the aberrations. (A) Positive spherical aberration induced after standard SMILE surgery. Parallel central rays (solid yellow line) and representative peripheral rays (dashed blue line) focus, resulting in a typical positive Δ8.0 mm SA on the cornea after myopia correction, shown in the associated wavefront map (indicated by black arrows). (B) Negative horizontal coma induced by SRL. SRL is present on the temporal (T) side of the cornea. Rays passing through the SRL region (solid purple line) focus differently relative to central and nasal rays (solid yellow line), with representative peripheral rays represented by dashed blue lines. This differential focusing induces a negative Δ8.0 mm HC and positive SA, shown in the combined wavefront map (indicated by black arrows) and the separated negative HC component.

[0021] Abbreviations: HC = Horizontal coma; SMILE = Micro-incision lenticule extraction; N = Nasal side; SA = Spherical aberration; SRL = Fine-scale image.

[0022] Figure 4 The image shows an example of multimodal imaging of a normal cornea with three different types of fine residual lenses.

[0023] Each row represents a single case at the three-month follow-up: (A) normal contralateral eye for control, (B) worm-shaped SRL, (C) crescent-shaped SRL, and (D) box-shaped SRL. Each case is presented with three images, from left to right: a corneal densitogram (120 µm slice) with a yellow line outlining the approximate lenticule extraction margin and marking the SRL area; an anterior surface elevation map; and an AS-OCT B scan. In cases B and D, key diagnostic features are highlighted: the low COD area of ​​the SRL in the densitogram (red arrow), the corresponding subtle bulge in the elevation map (red arrow), and the clear interlaminar separation in the AS-OCT scan (red arrow). The yellow line outlines the approximate lenticule extraction margin and marks the SRL area.

[0024] Abbreviations: AS-OCT = Anterior Optical Coherence Tomography; COD = Corneal Optical Density; SRL = Fine Residual Lens.

[0025] Figure 5 The image shows the measurement of the area of ​​a fine residual lens using ImageJ software.

[0026] The steps shown are: (A) Capture a corneal densitometry map (120 µm layer) from Pentacam HR ("Corneal Densito" mode); (B) Define the lens boundary in ImageJ; (C) Set the spatial scale calibration (130:3 pixels / mm); (D) Outline the SRL fragment using polygon selection for automatic area measurement; (E) Automatically calculate the area of ​​the SRL region using ImageJ.

[0027] Abbreviation: SRL = Fine residual lenticule; T = Temporal side; Δ = Change from preoperative to 3 months postoperatively. Detailed Implementation

[0028] 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, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0029] Example 1

[0030] The research design and patient selection of a corneal microlentidenal remnant detection system and method based on multimodal imaging specifically include the following steps:

[0031] (1) Patient selection and image acquisition

[0032] The study was conducted in accordance with ethical guidelines. It included nearsighted patients who underwent SMILE surgery in both eyes, completed follow-up for at least 3 months post-surgery, and underwent thorough ophthalmological examinations to exclude other eye diseases and systemic diseases that could affect corneal imaging.

[0033] (2) SRL screening and confirmation process

[0034] Postoperative imaging data were processed using the following standardized procedure: First, slit-lamp examination was performed to rule out other corneal lesions. Subsequently, two observers independently analyzed Pentacam HR images. If a focal low-density area was found within the expected removal area on the corneal densitogram (120µm layer) and a slight elevation was present at the corresponding location on the anterior surface height map, it was marked as a suspected SRL. All suspected cases underwent further AS-OCT examination. SRL was only definitively confirmed when the images showed clear interlaminar separation at the surgical interface and compensatory thinning of the superior tissue. A confirmed SRL was required to occupy fewer than two quadrants on the densitogram.

[0035] (3) Image processing and parameter measurement

[0036] All images underwent standardized analysis. Corneal densitograms were imported into image processing software for spatial calibration. The outline of the confirmed SRL was manually delineated, and its area, the number of corneal quadrants it occupied, and its morphology (e.g., worm-like, crescent-shaped, box-shaped) were automatically calculated and recorded. Using the "Zernike Analysis" mode of a corneal topography system, with the corneal apex as the center, various higher-order corneal aberrations within the 6.0 mm and 8.0 mm analysis areas were measured and recorded. The "Compare 2 Exams" function was used to generate pre- and post-operative corneal curvature difference maps, calculating the effective optical zone area, diameter, and its eccentricity relative to the corneal apex.

[0037] Example 2

[0038] The above research was analyzed in the following specific process:

[0039] From a retrospective cohort of 577 consecutive patients (1085 eyes), 66 eyes from 33 patients (mean age 22.21 ± 0.59 years; 18 (54.5%) male) were ultimately selected for analysis as the final study group. These patients met all eligibility criteria for the study and were confirmed to have unilateral spontaneous retardation (SRL) using our multimodal screening protocol. Of the 33 eyes with confirmed SRL, 17 (51.5%) were right eyes, while 16 (48.5%) were right eyes in the contralateral non-SRL eyes.

[0040] Table 1 compares the preoperative parameters of the SRL eye and the contralateral non-SRL eye. The two groups were well-matched in terms of preoperative mean corneal diameter, spherical power, K1, and K2 (all P > 0.05), while the preoperative cylinder power was slightly lower in the SRL group (P = 0.046). The mean central corneal thickness was also slightly lower (553.12 ± 28.91 μm vs. 555.91 ± 29.20 μm, P = 0.002), while the planned lens thickness and calculated residual stromal thickness were comparable (P > 0.05).

[0041] Table 1 Comparison of ocular clinicopathological parameters

[0042]

[0043] Data are expressed as mean ± standard deviation or as numbers. Paired t-tests and Wilcoxon signed-rank tests were used for SRL and non-SRL variables. Abbreviations: CCT = Central corneal thickness; D = Diopter; EOZ = Effective optical zone; K = Corneal curvature value; RST = Residual stromal thickness; SD = Standard deviation; SRL = Fine residual lenticule.

[0044] *p < 0.05, indicating statistical significance.

[0045] (1) Characteristics of fine residual lenses

[0046] Detailed information on the 33 SRLs is shown in Table 2.

[0047] Table 2 Characteristics of minute residual lenses

[0048] Data are expressed as mean ± standard deviation or as a number. The "Quadrant Occupation" value indicates the average number of corneal quadrants (out of four) extended into by the fine residual lenticule.

[0049] The average area of ​​stromal retinal tracts (SRLs) was 0.875 ± 0.406 mm², and they occupied an average of 0.73 ± 0.26 quadrants, confirming their minute nature. The most common location of SRLs was the 6-12 o'clock sector corresponding to the temporal cornea of ​​the right eye and the nasal cornea of ​​the left eye; n=22, 66.7%, followed by the 6-9 o'clock sector (inferior temporal side of the right eye / inferior nasal side of the left eye; n=6, 18.2%) and the 9-12 o'clock sector (superior temporal side of the right eye / superior nasal side of the left eye; n=5, 15.2%). The main morphologies were worm-shaped (39.4%), crescent-shaped (36.4%), or box-shaped (24.2%).

[0050] Visual and refractive results

[0051] Three months post-surgery, standard visual and refractive outcomes were comparable between the SRL eye and the contralateral control eye. Figure 1 There were no statistically significant differences in postoperative uncorrected or corrected visual acuity, refractive predictability, accuracy, or stability (all P > 0.05). However, compared with the contralateral control group, the SRL eye showed significantly smaller mean effective optical zone area (P < 0.001) and effective optical zone diameter (P = 0.024) (Table 1).

[0052] (2) Higher-order aberrations of the cornea

[0053] Corneal HOA analysis revealed statistically significant differences associated with SRL (Table 3).

[0054] Table 3 Comparison of higher-order aberrations

[0055]

[0056] Data are expressed as mean ± standard deviation or as numbers. Paired t-tests and Wilcoxon signed-rank tests were used for the variables of presence and absence of subtle residual lenses. Abbreviations: HOA = Higher-order aberrations; Pre = Preoperative; Post = Postoperative; RMS = Root Mean Square; SD = Standard Deviation; SRL = Subtle Residual Lens.

[0057] *p < 0.05 indicates statistical significance.

[0058] Compared with the control group, the SRL eyes showed a significantly greater increase in total HOARMS from baseline in the 6.0 mm (P = 0.006) and 8.0 mm regions (P < 0.001). Specifically, ΔSA showed a greater positive shift in the SRL group in both the 6.0 mm (P = 0.026) and 8.0 mm (P = 0.001) regions. ΔHC in the SRL eyes showed a significant negative shift only in the 8.0 mm region compared with the control group (P = 0.001).

[0059] (3) Risk factor analysis of key HOA changes

[0060] Multiple linear regression analysis explored factors that were independently associated with changes in key HOA (Hospital Occupational Abilities). Figure 2No significant independent factors were found to influence Δ6.0 mm SA (all P > 0.05). For Δ8.0 mm SA, higher preoperative spherical error (P < 0.001) and smaller postoperative effective optical area (P < 0.001) were significant independent factors, but the presence of SRL was not independently associated (P = 0.990). Conversely, for Δ8.0 mm HC, the presence of SRL was the only significant independent factor associated with negative offset (P = 0.004).

[0061] This retrospective contralateral study showed that SRL (surface retraction) after SMILE surgery significantly affected postoperative corneal HOA at 3 months, particularly within the 8.0 mm analysis area. The findings suggest a potential mechanism by which SRL influences horizontal coma and spherical aberration.

[0062] A key strength of this study is the implementation of a systematic, multimodal protocol for identifying and confirming SRLs. We define SRLs as fragments spanning fewer than two quadrants, an operational threshold used to distinguish these subtle findings from large residual lenses that can induce complex aberrations and significant astigmatic effects, while our definition focuses on the more regular effects of smaller fragments. While AS-OCT provides deterministic structural confirmation, this study highlights the potential of corneal densitography as a sensitive primary screening tool. Our protocol is based on identifying focal areas with relatively low COD on scans taken 3 months post-surgery, which stand out against the backdrop of generally high COD in lenticule extraction areas at a depth of 120 μm. We hypothesize that these low-density “islands” represent undisturbed tissue from which incomplete lenticule separation has occurred, thus retaining near-normal COD compared to the surrounding stroma. This optical signal may be more sensitive than structural changes alone, potentially identifying SRLs that are too thin to produce a discernible interface elevation on AS-OCT.

[0063] Our key finding is an independent association between SRL and the induction of negative horizontal coma within the 8.0 mm region, with a predisposition in the 6–12 o'clock sector. This positional preference may be explained by surgical technique, as a small peripheral attachment area is often intentionally left to provide counteracting traction during dissection. For right-handed surgeons, the ergonomically challenging inferior temporal (right eye) or inferior nasal (left eye) quadrants pose a higher risk of incomplete dissection or tearing of this attachment area, resulting in debris residue. Optically, such debris located in the 6–12 o'clock sector can cause local corneal steepening, leading to light rays passing through this area (e.g., the negative x-axis corresponds to the temporal side of the right eye or the nasal side of the left eye). Figure 3B) Propagation is faster than that of the corresponding side, causing wavefront differences to advance along the horizontal meridian, directly manifesting as negative horizontal coma. This effect is limited to the 8.0 mm region, consistent with the peripheral nature of SRL, which has a greater effect on larger analytical areas but a diluted effect on the central 6.0 mm region, making it statistically insignificant. Given that coma significantly affects visual acuity under larger pupil conditions, this provides an objective relevance to potential night vision symptoms such as halos, although definitive clinical significance requires correlation with patient-reported outcomes.

[0064] Conversely, the association between SRL and increased positive ΔSA is indirect. Instead, higher preoperative myopia and a smaller effective optical zone are the primary drivers of increased Δ8.0 mm SA. Crucially, we found a significantly smaller effective optical zone in SRL eyes, which itself may be a result of localized epithelial or stromal remodeling induced by the interface irregularities of SRL. A smaller effective optical zone is known to increase positive SA after refractive surgery due to the steeper peripheral cornea relative to the flattened central cornea, especially in higher myopia corrections, resulting in peripheral light passing through a more abrupt transition zone and increasing Δ8.0 mm SA. Figure 3 The lack of significant influencing factors in Δ6.0 mm SA indicates that the effect in this small region is multifactorial or below the detection threshold of this analysis.

Claims

1. A method for detecting minute residual lenticules in the cornea based on multimodal imaging, characterized in that, Includes the following steps: S1: The system automatically controls the Pentacam HR and AS-OCT devices via a standardized interface (DICOM / SDK) to acquire postoperative corneal density maps (120µm layer), anterior surface height maps, and OCT volume data. After acquisition, an affine transformation model based on anatomical landmarks is used to spatially register the multi-source images and establish a unified coordinate system. S2: In the density map, high-density background areas (expected removal areas) are defined by threshold segmentation. A sliding window is used to calculate the local density mean, which is then compared with the global background. If the local density is significantly lower than the global background, it is merged into candidate SRL regions through connected component analysis. S3: Map the coordinates of the candidate SRL region to the front surface height map, and use SIFT / SURF feature matching to achieve sub-pixel level fine registration. Calculate the height difference between the corresponding local region and the surrounding reference area. If there is a significant positive bulge, it is determined as topographic association evidence, supporting the candidate region as a real anomaly. S4: In the registered OCT B-scan, a pre-trained U-Net segmentation model is used to automatically identify the surgical interface structure. The model detects whether there is interlayer separation (interruption of high-reflectivity bright bands) at the interface and whether the overlying tissue exhibits compensatory thinning. The system ultimately confirms SRL only when both are present. S5: For confirmed SRLs, automatically extract the contour from the spatially calibrated image, calculate the actual area and the number of corneal quadrants occupied. Further extract shape features, input them into a classifier built on support vector machines, automatically output its morphological category, and generate a quantification report.

2. The method of claim 1, wherein, In step S2, the intended lens removal area is defined by the region in the corneal density measurement map that generally shows a high corneal optical density value.

3. The method of claim 1, wherein, In step S5, the quantitative analysis is performed by image processing software, including spatially calibrating the corneal density map and then outlining the SRL to automatically calculate its area.

4. The method according to claim 1, characterized in that, In step S5, the morphological classification includes worm-like, crescent-shaped, and box-shaped.

5. The method according to claim 1, characterized in that, The method further includes step S6: based on the confirmation and quantification results of the SRL, a report is generated to assess the risk of postoperative higher-order aberrations in the examined eye, the report specifically highlighting the correlation between the SRL and negative horizontal coma.

6. A corneal micro-residual lenticule detection system for implementing the method of any one of claims 1-5, characterized in that, include: An image acquisition module, configured to acquire image data from a corneal topography instrument and a front-end OCT device; The data processing module is configured to perform collaborative analysis of corneal density maps and anterior surface height maps to screen for suspicious SRL regions; the SRL confirmation module is configured to make final confirmation of suspicious SRL regions based on the interpretation of anterior segment OCT images; and the quantification and output module is configured to measure parameters and output results for confirmed SRLs.

7. The system according to claim 6, characterized in that, The quantization and output module integrates an image processing unit for automatically performing area calculations and quadrant analysis.

8. A computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the method as described in any one of claims 1 to 5.

Images