Construction method and application of histopathologic quantitative evaluation model of choroidal neovascularization

Abstract

The invention discloses a construction method and application of a histopathologic quantitative evaluation model of choroidal neovascularization, and the evaluation model constructed by the method can accurately identify a choroidal neovascularization structure around a laser spot, and can carry out rapid batch analysis on a large number of laid sheets. The reliability and the consistency of analysis results are improved, and differences caused by laser surgery, interference of manual operation factors and the labor cost of analysis are reduced.

Description

Technical Field

[0001] The invention relates to a construction method and application of a histopathological quantitative evaluation model of choroidal neovascularization, belonging to the technical field of image processing and analysis. Background Art

[0002] When conducting nonclinical drug efficacy evaluation, especially in the field of ophthalmic drug efficacy evaluation, the evaluation of the fundus of experimental animals is crucial. Laser photocoagulation is usually used to model experimental animals, causing choroidal neovascularization, which is similar to the lesions of clinical wet age-related macular degeneration (wAMD). One of the biggest advantages of preclinical drug efficacy experiments is that pathological analysis can be performed at the end of the experiment as an endpoint assessment of effectiveness. In preclinical ophthalmic drug efficacy studies, a variety of ophthalmic examinations are usually used for evaluation, such as fluorescein fundus angiography and optical coherence tomography, which can provide real-time, dynamic images of retinal and choroidal blood vessels. However, these methods are currently mostly qualitative or semi-quantitative. There are limitations in the precise evaluation of innovative drug molecules or innovative routes of administration.

[0003] So far, the two most widely reported quantitative techniques for the area of ​​choroidal vascular fibropigmentation reaction area are traditional image analysis software analysis and manual marking. In the traditional method, ImageJ is used to analyze fluorescent staining images based on the user's manual threshold setting or the use of automatic threshold method to segment the image, and the area, intensity and other parameters of the segmented area are calculated and statistically analyzed. When using ImageJ, the grayscale threshold adjustment of digital images of different slides cannot be completely consistent, resulting in standard drift in the analysis of different eye specimens. When performing batch analysis, ImageJ can automate the processing process by writing macros or using plug-ins to reduce manual intervention and improve the repeatability of analysis. The correct use of ImageJ for complex image analysis requires not only considerable professional knowledge and experience, but also the results of batch repeated analysis depend largely on the selected algorithm and parameter settings. Improper settings may lead to inaccurate analysis results. The second method is the traditional method of manually marking different choroidal neovascularization areas, which is prone to drift in the identification of the laser spot boundary. In addition, there may be differences in the annotations between different researchers or the same researcher at different time points, affecting the consistency and repeatability of the data. More importantly, the above two methods are only used to analyze the area containing the laser spot, and the influence of uneven laser spot size caused by surgery on the analysis results cannot be ruled out. In particular, the tissue area at the edge of the laser spot cannot be accurately analyzed.

[0004] In summary, in order to conduct adequate pathological quantitative analysis of choroidal neovascularization and accurately evaluate the effectiveness of innovative drugs and their dose-effect relationship, it is necessary to develop a more precise quantitative method for quantifying the laser area of ​​the choroid, with higher accuracy and sensitivity of analysis and a cost-effective method to evaluate choroidal neovascular tissue proliferation after laser photocoagulation. Summary of the invention

[0005] Purpose of the invention: The purpose of the present invention is to provide a method for constructing a histopathological quantitative evaluation model of choroidal neovascularization and the application of the model.

[0006] Technical solution: The present invention provides a method for constructing a histopathological quantitative evaluation model of choroidal neovascularization, comprising the following steps:

[0007] (1) Pre-treating, staining, and preparing the eyeballs of the choroidal neovascularization animal model to obtain a full choroidal-scleral spread;

[0008] (2) Use a multi-channel fluorescence microscope to collect fluorescence images and bright field images of the entire choroidal scleral flat mount. The sample boundaries in the two images need to completely overlap and be fused into one image;

[0009] (3) annotating the neovascular lesion area in the composite image, and each annotated area is saved as a separate layer; the neovascular lesion area includes the vascular fibrous pigment reaction area and the laser spot;

[0010] (4) Create a new tissue classifier, add the background area, choroid-sclera area, vascular fiber pigment reaction area and laser spot area in sequence, circle the corresponding areas in the layer, and start training. When the Cross-Entropy is reduced to the minimum and stable, confirm its recognition accuracy and stop training and save the tissue classifier to obtain the evaluation model.

[0011] Furthermore, the choroidal neovascularization animal model in step (1) is constructed by laser induction.

[0012] Furthermore, the pretreatment step in step (1) includes: making a hole in the center of the cornea, removing the cornea and lens, peeling off the retina and choroid-sclera complex, and obtaining a sample to be stained.

[0013] Furthermore, the staining step in step (1) includes: blocking the sample to be stained, immersing it in a diluted GS-IB4 solution for incubation in the dark, and washing it after the incubation.

[0014] Furthermore, the light-proof incubation condition is 2°C to 8°C.

[0015] Furthermore, the preparation step in step (1) includes: transferring the cleaned sample to a glass slide, cutting off the optic nerve at its root, cutting the remaining tissue into four petals and laying them flat in a four-leaf clover shape, and sealing the slide to obtain a full choroid-scleral slide.

[0016] Furthermore, the software used to fuse the fluorescence image and bright field image of the whole choroidal sclera paving into one image in step (2) is FCSnap.

[0017] Furthermore, the analysis software used in step (3) and step (4) is HALO.

[0018] The present invention also provides an application of the model constructed based on the above method in analyzing the effect of ocular drugs on the size of the choroidal vascular fiber pigment reaction area of ​​experimental animals.

[0019] Furthermore, the calculation method for analyzing the area of ​​the choroidal vascular fibrous pigment reaction region is to deduct the area of ​​the laser spot region from the area of ​​the neovascular lesion region.

[0020] Beneficial effects: Compared with the prior art, the present invention has the following outstanding significant advantages: the histopathological quantitative evaluation model of choroidal neovascularization constructed by the method of the present application can accurately identify the choroidal neovascularization structure around the laser spot, and can perform rapid batch analysis on a large number of paving slides, thereby improving the reliability and consistency of the analysis results, reducing the variability caused by laser surgery, interference from human operating factors, and the labor cost of analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 The figures are a fluorescence image of the whole choroidal sclera flat mount (A), a composite image of the whole choroidal sclera flat mount fluorescence and bright field (B), and a schematic diagram of the circle annotation when constructing the model of the present invention (C).

[0022] Figure 2 The analysis images of the four analysis methods in Example 2 (A: ImageJ analysis image, B: manual labeling method, C: AI analysis including laser spot, D: AI analysis excluding laser spot) and the comparison results of choroidal neovascularization areas (E). DETAILED DESCRIPTION

[0023] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.

[0024] Example 1

[0025] Step 1. C57BL / 6J mice (male, 13-15 weeks old) were subjected to fundus laser-induced choroidal neovascularization model. The number of laser photocoagulations was 4 per eye. The laser parameters were: wavelength 532 nm, power 200-400 mW, spot diameter 50 μm, exposure time 100 ms, which can be adjusted according to actual conditions.

[0026] Step 2: Remove the mouse eyeball, pierce the center of the cornea with a syringe, and fix it in 4% PFA at room temperature for 30 minutes. Remove the cornea and lens under a stereo microscope, fix it in 4% PFA at room temperature for 10 minutes, wash it with Phosphate Buffered Saline (1×) at room temperature, and then peel off the retina and choroid-sclera complex under a stereo microscope.

[0027] Step 3: Block the treated samples with 3% Triton blocking solution at room temperature for 2 h. After blocking at room temperature, immerse them in 1:50 GS-IB4 solution and incubate them at 2°C to 8°C in the dark for 48 h. Then, wash them on a shaker with Phosphate Buffered Saline (1×) in the dark.

[0028] Step 4: Transfer the cleaned sample to a glass slide, cut off the optic nerve at the root under a stereo microscope, cut the tissue into four petals and lay them flat in a four-leaf clover shape, seal the slide with a coverslip, and then seal the four corners with nail polish.

[0029] Step 5: Use the multi-channel fluorescence microscope Olympus ckx53 to collect the whole choroidal scleral flat-mount images, use the LED fluorescence illumination device BP470 / 40 filter holder model STL-CKXRFA4 to excite the FITC green fluorescence channel of GS-IB4 staining, collect images under the 4x objective, use FCSnap software to capture the images and store them in tiff file format. Turn off the fluorescence channel, keep the slide position unchanged, keep other settings of the fluorescence microscope unchanged, and collect images again in bright field mode to ensure that the sample boundaries in the two images completely overlap.

[0030] Step 6: Open the fluorescence images of the same slide in FCSnap software at the same time ( Figure 1 A) and bright field images, select "Process", "Color Composite", "All Images" in the menu bar, click "Add" under the corresponding image, select the corresponding light channel, and click "OK" to generate the composite image ( Figure 1 B).

[0031] Step 7. Import the digital image of the whole choroidal sclera fold with fluorescence and bright field fusion into the HALO analysis software, right-click the image name in the "Image" box, select "Image Details", and select "Set Resolution" to set the image resolution to 2.5 microns / pixel.

[0032] Step 8. Use the annotation tool in the "Annotations" tab to circle the neovascular lesion area including the laser spot and the vascular fibrochrome reaction area. Each annotation is saved as a separate layer.

[0033] Step 9. In the "Classifiers" tab, select "ClassifierActions", and select "New..." to create a new tissue classifier "DenseNet V2 (HALO AI)". Add the slide background area, choroid sclera area, vascular fiber pigment reaction area, and laser spot area in order, circle the corresponding areas, open "Real-TimeTuning" in "Classify", set Resolution to 3.21μm / px, and adjust Minimum Object Size to 200μm 2 , click "Start Training" to start training. When the cross-entropy is minimized and stable, use "Real-time Tuning" to confirm its recognition accuracy, stop training and save the tissue classifier.

[0034] Circle diagram as shown Figure 1 As shown in C, after the bright field and dark field are superimposed, the scanning can show the laser spot (the bright yellow-white area in the composite image, Figure 1 The area encircled by blue lines in C) and the immunofluorescent staining area of ​​choroidal vascular fibroblast pigment reaction (the green fluorescent area around the laser spot in the composite image, Figure 1 The circled area between the orange and blue lines in C represents the choroidal vascular fibrous pigment reaction area). The area containing the laser spot and the choroidal vascular fibrous pigment reaction area is the neovascular lesion area annotated in the "Annotations" label.

[0035] Step 10. Select all images to be analyzed in the "Image" box, right-click and select "Analyze...", select "Classifiers" in Settings, select the stored Classifier, select "Annotation" in Region of Analysis, check all layers, and start batch analysis.

[0036] Step 11. After the analysis is completed, select all analyzed images, right-click and select Export, Advanced to output the summary data table and the report, legend and other results of the analysis results of each neovascular lesion area.

[0037] Example 2

[0038] ImageJ analysis method (ImageJ), manual labeling method (Manual), AI analysis method including laser spot (HALO LaserSpot+ ) and AI analysis method to eliminate laser spots (HALO LaserSpot- , Example 1 Method) The eyeball samples of mice treated with a marketed positive drug (Eylea, batch number: KTOFCJF, manufacturer: Bayer, dose: 40 μg / eye, single intravitreal injection on Day 7 after laser modeling) were analyzed.

[0039] ImageJ analysis steps:

[0040] 1. Import the fluorescent full choroidal scleral flat mount digital image into ImageJ analysis software, select "Adjust" in the "Image" menu bar, "Type", select "8-bit" to convert the image into an 8-bit grayscale image, then select "Adjust" in the "Image" menu bar, select "Threshold", adjust the threshold range until all neovascular lesion areas are automatically marked, and click "Apply" to apply the threshold setting.

[0041] 2. Select "Wand (tracing) tool" in the toolbar, select the neovascular lesion areas one by one, right-click, select "Add to ROI Manager", select all added neovascular lesion areas in ROI Manager, click "Measure", and the area of ​​each area will be displayed in the results.

[0042] Manual labeling steps:

[0043] Use Olymps OlyVIA to open the fluorescent full choroidal scleral flat mount digital image, select the "point-by-point area drawing" tool, mark along the edge of the neovascular lesion area, and click the "Enter" key to confirm after marking. The area of ​​the marked area will automatically be displayed next to the marking box.

[0044] AI analysis steps including laser spot:

[0045] The steps are basically the same as those in Example 1, except that in step 9, no laser spot area is added, but the entire vascular fibrochrome reaction area is used as the neovascularization area.

[0046] The images of the four analysis methods are shown in Figure 2. Figure 2 As shown in A to D, AI analysis method (HALO LaserSpot- ) and AI analysis including laser spot (HALO LaserSpot+ ), ImageJ analysis showed significant differences in the area of ​​new blood vessels in the analyzed samples ( Figure 2 E). In terms of analysis time, the AI ​​analysis method that excludes laser spots can reduce the analysis time by about 30% compared with the skilled ImageJ analysis method; although there is no significant difference in the analysis compared with the manual marking method, the analysis time can be reduced by about 45%. And when the model of the present application is used to analyze the area of ​​new blood vessels, by building the model once, batch rapid analysis can be performed with higher sensitivity and consistency. In conventional ophthalmic efficacy tests, the number of choroidal paving is large. The method of the present application greatly saves the time cost of project analysis and avoids the human error caused by fatigue of pathologists due to long-term repeated operations. In addition, the variability caused by laser surgery can also be reduced. Since laser surgery is manually operated and the retina choroid in the eyeball is a spherical surface, the level and angle of each point focus will not be completely consistent when the laser is applied, which will cause slight differences in the size of the laser spot. After staining, the entire laser spot and the surrounding vascular fiber pigment reaction area will have green fluorescence coloring, which cannot be distinguished by traditional methods. For green fluorescent coloring areas of similar size, there are two situations: the laser spot is large but the surrounding vascular fiber pigment reaction area is small, and the laser spot is small but the surrounding reaction area is large. When evaluating the effect of a new drug, only the impact on the surrounding reaction area needs to be evaluated. The drug will not change the size of the middle laser spot. Therefore, the evaluation after excluding the laser spot in the present application is more accurate and sensitive.

Claims

1. A method for constructing a histopathological quantitative evaluation model of choroidal neovascularization, characterized in that: The following steps are involved: (1) Pre-treating, staining, and preparing the eyeballs of the choroidal neovascularization animal model to obtain a full choroidal-scleral spread; (2) Use a multi-channel fluorescence microscope to collect fluorescence images and bright field images of the entire choroidal scleral flat mount. The sample boundaries in the two images need to completely overlap and be fused into one image; (3) annotating the neovascular lesion area in the composite image, and each annotated area is saved as a separate layer; the neovascular lesion area includes the vascular fibrous pigment reaction area and the laser spot area; (4) Create a new tissue classifier, add the background area, choroid-sclera area, vascular fiber pigment reaction area and laser spot area in sequence, circle the corresponding areas in the layer, and start training. When the Cross-Entropy is reduced to the minimum and stable, confirm its recognition accuracy and stop training and save the tissue classifier to obtain the evaluation model.

2. The method for constructing a histopathological quantitative evaluation model of choroidal neovascularization according to claim 1, characterized in that: The choroidal neovascularization animal model described in step (1) is constructed by laser induction.

3. The method for constructing a histopathological quantitative evaluation model of choroidal neovascularization according to claim 1, characterized in that: The pretreatment steps in step (1) include: making a hole in the center of the cornea, removing the cornea and lens, peeling off the retina and choroid-sclera complex, and obtaining the sample to be stained.

4. The method for constructing a histopathological quantitative evaluation model of choroidal neovascularization according to claim 1, characterized in that: The staining step in step (1) includes: blocking the sample to be stained, immersing it in a diluted GS-IB4 solution for incubation in the dark, and washing it after the incubation is completed.

5. The method for constructing a histopathological quantitative evaluation model of choroidal neovascularization according to claim 4, characterized in that: The light-proof incubation condition is 2°C to 8°C.

6. The method for constructing a histopathological quantitative evaluation model of choroidal neovascularization according to claim 1, characterized in that: The preparation steps in step (1) include: transferring the cleaned sample to a glass slide, cutting off the optic nerve at its root, cutting the remaining tissue into four petals and laying them flat in a four-leaf clover shape, and sealing the slide to obtain a full choroid-scleral spread.

7. The method for constructing a histopathological quantitative evaluation model of choroidal neovascularization according to claim 1, characterized in that: The software used to fuse the fluorescence image and bright field image of the whole choroidal scleral spread into one image in step (2) is FCSnap.

8. The method for constructing a histopathological quantitative evaluation model of choroidal neovascularization according to claim 1, characterized in that: The analysis software used in step (3) and step (4) is HALO.

9. Use of the model constructed based on the method described in any one of claims 1 to 8 in analyzing the effect of ocular drugs on the size of the choroidal vascular fibrochrome reaction area of ​​experimental animals.

10. The use according to claim 9, characterized in that: The calculation method for analyzing the area of ​​choroidal vascular fibrous pigment reaction is to deduct the area of ​​laser spot from the area of ​​neovascular lesion.

Images