Method for determining duration and size of macular tears, taking into account their forms

Abstract

FIELD: ophthalmology.SUBSTANCE: used to accurately determine the volume and duration of pathological changes in the macula, including the diameter of the hole and its depth taking into account its shape, characterized by an uneven contour, preoperative diagnostic and intraoperative navigation based on the exact dimensions of anatomical structures and the prognosis of the result of surgical treatment, including drug treatment, of retinal diseases using 3D models of the posterior segment of the eye, stereophotogrammetry, vector analysis, artificial intelligence in real time and dynamic observation, in the choice of treatment tactics.EFFECT: improved diagnostics and treatment, including conservative treatment, of macular holes using stereophotogrammetry. The main point of application of the method we have developed is the result of the breakdown of all physiological mechanisms of metabolism of the macular region of the retina - macular hole.1 cl, 20 dwg

Description

[0001] The invention relates to the field of medicine, namely to ophthalmology, and can be used to accurately determine the volume of pathological changes in the macula, including the diameter of the retinal rupture, taking into account its depth, uneven shape and time of occurrence, preoperative diagnostic and intraoperative navigation based on the exact dimensions of anatomical structures and the prognosis of the result of surgical treatment, including drug treatment, of retinal diseases using 3-D models of the posterior segment of the eye, stereophotogrammetry, vector analysis and artificial intelligence in real time and dynamic observation, the choice of treatment tactics.

[0002] Our goal is to improve the diagnosis and treatment, including conservative treatment, of macular holes using stereophotogrammetry. The primary application of our method is the disruption of all physiological metabolic mechanisms in the macular region of the retina—a macular hole.

[0003] The technical result achieved by the invention is the accurate diagnosis of macular holes, assessing the shape, age, and diameter of the retinal hole, taking into account its depth, without the use of optical coherence tomography (OCT). This result consists of analyzing data from a range of diagnostic tests using proposed algorithms, stereophotometry, and 3D imaging, followed by vector analysis and the use of artificial intelligence in real time to predict disease progression, monitor dynamics, and select treatment strategies.

[0004] Currently, there is clinical experience with the use of 3D visualization in ophthalmic surgery. Koichiro and his co-authors proposed a method for creating a three-dimensional fundus image based on multiple partial images by shifting the fundus camera. The prototype is based on the assumption of a spherical fundus, and the image of the crystalline lens is transformed onto a square surface. The disadvantages of this model (a combination of a crystalline lens and a magnifying contact lens) are the approximate optical characteristics of the simulated lens, which affect the creation of the actual image displayed on the square surface. The fundus image is backprojected from multiple images, which affects the quality of the reconstructed sphere.

[0005] Another proposed method is an eye model (with volumetric retinal morphology) developed using optical coherence tomography (OCT) datasets. This model allowed us to create an anatomically correct representation of three-dimensional (3D) retinal layers, while the spherical dimensions of the anatomical structures and their exact sizes were not considered. [Zawadzki R. J. et al. Toward building an anatomically correct solid eye model with volumetric representation of retinal morphology / / Ophthalmic Technologies XX. - SPIE, 2010. - Vol. 7550. - P. 412-418].

[0006] To assess the parameters of macular holes, taking into account their shape, diameter, and depth, specialists typically use specialized devices (prototypes), such as a fundus camera, optical coherence tomography (OCT), and angio-OCT. However, these conventional diagnostic methods have the following drawbacks: limited spatial resolution (the distance between the lips of the hole is visualized, but not its shape), insufficient image contrast and brightness, and the high cost and unavailability of these devices.

[0007] The introduction of stereophotogrammetry into clinical practice has allowed us to improve the determination of the true sizes of the anatomical structures of the eyeball and their relationships, and knowledge of the shape of the rupture makes it possible to assess its duration, size, depth and choose a method for its treatment.

[0008] The stereophotogrammetry method and traditional imaging methods are interchangeable. The key is the ability to use our method without expensive equipment on any retinal imaging device (fundus camera, retinal camera, webcam, digital microscope). Using artificial intelligence, a 3D model can be constructed from a single image.

[0009] The following describes an algorithm for accurately determining the volume of pathological changes in the macula, including the diameter, depth, and age of the retinal tear, taking into account its uneven shape, preoperative diagnostic and intraoperative navigation based on the exact dimensions of anatomical structures and the prognosis of the outcome of surgical treatment, including drug treatment, of retinal diseases using 3-D models of the posterior segment of the eye, stereophotogrammetry, vector analysis and artificial intelligence in real time and dynamic observation, and the choice of treatment tactics.

[0010] 1. Sequential fundus imaging is performed to accurately calibrate the instruments, which in turn ensures the clearest and most contrasting image of the macular region of the retina.

[0011] Technical specifications of the non-medial fundus camera and its calibration:

[0012] Field of view angle: 45° (30° for O3.7mm)

[0013] Pupil diameter: minimum 3.7 mm

[0014] Focal length: 40mm for patient; 60mm for calibration paper.

[0015] Calibration of the non-medial camera (Fig. 3)

[0016] Retinal camera specifications and calibration:

[0017] Digital camera features: resolution 640*480; 24 bit;

[0018] Width of the working part of the module in contact with the eye: 9 mm;

[0019] Lens mount type to camera: Quick Connect;

[0020] The camera is in direct contact with the eye;

[0021] Field of view angle 130 degrees

[0022] Focal length: 12mm

[0023] Calibration of the non-medial camera (Fig. 4)

[0024] Technical specifications of the digital microscope and its calibration:

[0025] Camera Head (CHU)

[0026] HD CMOS: 3 pcs, 1 / 2.9 type

[0027] Resolution: Approx. 2.07 million pixels (1920 × 1080)

[0028] Images: 1080 effective lines in progressive format

[0029] Focal length: 170mm

[0030] Digital microscope calibration (Fig. 5)

[0031] To build a 3D model for MR diagnostics, 30 images of different quadrants of the fundus (posterior pole, superior nasal, superior, superior temporal, temporal, nasal, superior nasal, inferior nasal quadrants) of one eye of one patient are required, obtained using a non-medical fundus camera, a digital camera during the examination of the patient, as well as anamnesis data (rupture diameter according to OCT data, OCT-angio examination protocol).

[0032] 2. A 3D model is created using a point cloud generated from overlapping images. The density of the points that describe the shape of the object being modeled is a determining factor in the model's reliability. Two-dimensional images are converted into a discrete array, in which each point is located in space according to its prototype.

[0033] The results of vector interpretation of three-dimensional models are used to construct a 3D map (approximation model), which allows one to evaluate the shape of the macular hole, determine its diameter, depth, perimeter, area and duration of the hole.

[0034] Three-dimensional models are obtained by photogrammetric processing of images from a digital video camera with a focal length of 170 mm.

[0035] The accuracy of the resulting photogrammetric model is estimated based on physiological measurements and reaches approximately 0.025 mm, which is due to the resolution of the sensor, that is, the pixel size on the retina.

[0036] The following describes algorithms for assessing macular hole parameters based on their shapes. All data listed should be for one eye and performed during a single diagnostic examination.

[0037] Technology for obtaining an approximation model

[0038] 1. The photogrammetric model is used to interpret the edges of the hole (the ridges that form along the edges of the hole are deciphered) and create a vector contour. When a macular hole occurs, its boundaries may initially be distorted, and the hole itself may have an irregular shape. This may be due to damage mechanisms caused, for example, by vitreoretinal traction.

[0039] Over time, if the MR does not heal on its own, its edge fibrosizes, becomes more rounded and defined, and the tear itself increases in size. Thus, the tear's shape can be used to judge its age (Fig. 6), although the morphology of the tear may change.

[0040] a) the larger the area of ​​the rupture, the longer it has been there;

[0041] b) the more serrations there are at the edges of the rupture, the longer it has been there and the worse the prognosis for surgical reconstruction.

[0042] A vector contour is a closed graphic contour (graph) consisting of nodes and lines (Fig. 7). Each node contains coordinates in the local coordinate system of the eye model. These coordinates are used to calculate the perimeter and area of ​​the gap. The perimeter of the gap is calculated as the sum of the lengths of the segments that make up the closed contour. The lengths of the segments are calculated from the coordinates of their ends using the well-known Gauss formula.

[0043] A vector constructed from photogrammetric data and a profile constructed by the OCT system are imported into the program. A geometric figure is constructed based on the photogrammetric contour and profile.

[0044] The height of the gap is determined by the length of the normal dropped from the center of gravity of the figure, which describes the shape of the gap on the base plane. The base is formed by the faces of the lateral surface.

[0045] Figure 7 shows a 3D model of macular hole on the left, and the results with OCT calculations on the right.

[0046] The model volume is determined by an algorithm that automatically partitions the fracture model into tetrahedra using Delaunay triangulation. These elementary tetrahedra allow the geometry of the original model to be fully replicated.

[0047] The body of the program's computational part consists solely of calculation commands for the given coordinates of the vertices of the initial Delaunay tessellation array figure, used to tile the three-dimensional structure with rectangles. Next, in a loop, the program's main function is applied to the vertices of all created rectangles, combining possible combinations of the vertices of the Delaunay tetrahedra and checking which four of them fall within the same sphere.

[0048] Then the volumes of the tetrahedrons are calculated and added together to form the total volume of the figure.

[0049] Before volume calculation, the resulting image files from discontinuity vectorization are converted into a triangle system. This preprocessing, called triangulation, is widely used in mesh encoding, mesh signal processing, and mesh editing. First, the normal direction of the triangle is determined using the vertex order and the right-hand rule.

[0050] After obtaining the normals, we construct a set of triangles by connecting all the vertices of the polygon to the origin.

[0051] Each edge and origin form an elementary triangle, which is the smallest unit for calculations. We determine the area of ​​each elementary triangle as follows: the magnitude of this value is the area of ​​the triangle, and the sign of the value is determined by checking the position of the origin relative to the edge and the direction of the normal.

[0052] In three dimensions, the basic unit of calculation is the tetrahedron. For each triangle, each of its vertices is connected to the origin, forming a tetrahedron (Fig. 8).

[0053] The construction technology is implemented using Python libraries and Blender API (BPY).

[0054] [Jackson TL et al. Ocriplasmin for treatment of vitreomacular traction and macular hole: A systematic literature review and individual participant data meta-analysis of randomized, controlled, double-masked trials / / survey of ophthalmology. - 2022. - T. 67. - No. 3. - pp. 697-711.]

[0055] 3. To obtain the most accurate dimensions, multiple MR measurements are conducted, including grading measurements to create a formula for recalculating the 3D model. The shoulders and transverse diameter of the fracture are measured at different depth levels. To measure the fracture area, the trapezoidal method used for the generalized method of measuring faults is used (Fig. 1, Fig. 2).

[0056] We have developed for the first time a classification that takes into account the rupture diameter:

[0057] - small 90-400 µm;

[0058] -average 400-600 microns;

[0059] -large, over 600 microns.

[0060] This classification justifies further calculations of the rupture depth using the formula for each rupture area.

[0061] The patient's macular hole diameter is calculated according to the classification, the depth of the macular hole is calculated and controlled by OCT, with the error of the measurement method calculated.

[0062] For small ruptures, the rupture depth is between 100 and 620 μm (average depth is 337 μm).

[0063] With average sizes, the rupture depth ranged from 170 to 750 µm (the average depth was 416 µm).

[0064] For large ruptures, the rupture depth is within 100 to 440 μm (the average depth was 338 μm).

[0065] Rules for small, medium and large sizes:

[0066] 1. For small sizes with a diameter from 90 to 400 μm and a height from 100 to 300 μm, the diameter is equal to the depth, which is confirmed by OCT data (the error of the method is 1.4), where a rupture up to 400 μm has a depth of 90 to 450 μm. The diameter-to-height ratio of the rupture is from 0.5 to 1.5, which corresponds to the indicators of medium ruptures. When evaluating 3D models of small ruptures using our calculations, it is possible to more accurately determine the rupture depth from 100 to 620 μm, where the average depth was 337 μm). The area for small ruptures is 0.125 μm. The width error according to stereophotogrammetry data with OCT is 20 μm. Thus, we have proven the accuracy of our proposed method.

[0067] 2. For medium-sized tears with a diameter of 400 to 600 μm and a height of 300 to 450 μm, the diameter is always equal to the depth, which is confirmed by OCT data (the error of the method is 1.8), where a tear greater than 400 μm has a depth of 330 to 750 μm. The diameter-to-height ratio of the tear is from 0.5 to 1.5. When evaluating 3D models of medium-sized tears using our calculations, it is possible to more accurately determine the depth of the tear from 170 to 750 μm, where the average depth was 416 μm).

[0068] 3. For large ruptures with a diameter of 600 to 800 µm or more and a height of 200 to 500 µm or more, the diameter is always greater than the depth, as confirmed by OCT data (the error of the method is 1.6), where a rupture greater than 600 µm has a depth of 100 to 440 µm. The diameter-to-height ratio of the rupture ranges from 1.5 to 6.0. When evaluating 3D models of large ruptures using our calculations, it is possible to more accurately determine the rupture depth within the range of 100 to 440 µm, where the average depth was 338 µm.

[0069] 4. Macular holes with a diameter of more than 400 μm have large errors in 3D models due to uneven contours and misalignment of fixed alignment points along the contour; they require individual 3D image modeling using artificial intelligence.

[0070] Summary table of vector analysis results

[0071] N patient OCT Diameter Height Model Rupture, µm Height µm Perimeter mm Area mm2 Volume mm3 microns microns and along b across h P S V 1 Patient 1 Fig.7.1 0,272 0,408 Fig. 7.2 292 263 328 0,915 0,0676 0,0364 2 Patient 2 Fig.8.1 913 502 Fig. 8.2 720 742 495 2,231 0,403 0,1479 3 Patient 3 Fig.9.1 550 376 Fig.9.2 558 457 389 1,658 0,2088 0,1365 4 Patient 4 Fig.10.1 762 351 Fig.10.2 520 434 322 1,500 0,170 0,0235 5 Patient 5 Fig.11.1 588 379 Fig.11.2 432 392 419 1,351 0,132 0,1554 6 Patient 6 Fig.12.1 430 396 Fig.12.2 463 485 443 2,427 0,457 0,0808

[0072] Clinical example 1.

[0073] Patient M., 56 years old. Case history No. 25-312 of the State Budgetary Health Institution of the Vologda Oblast "VOKOB"

[0074] Complained of deteriorating vision in the right eye for 1.5 weeks.

[0075] Subjective examination:

[0076] Visus OD 0.04 n / c according to ETDRS table Log MAR = 1.4

[0077] Computer static perimetry data - concentric narrowing of the visual fields.

[0078] Amsler test - grid lines are curved in the central area in the form of a spot.

[0079] Objective examination data:

[0080] Biomicroscopy with 78D high-resolution ophthalmoscopy lens.

[0081] Right eye - The optic disc is pale, with clear borders. Arteries are constricted, veins are congested. There is a macular hole in the central region. Peripherally, there are no abnormalities.

[0082] Left eye - The optic disc is pale pink, with clear borders. Arteries are constricted, veins are full. No abnormalities in the central region, no abnormalities in the periphery.

[0083] OST of the right eye - complete macular hole up to 250 μm wide, 259 μm high.

[0084] Calculating depth without using additional equipment:

[0085] 1. Digital webcams (Leica, Topcon) were used to obtain digital images for creating a 3D model, selection and quality control of images;

[0086] 2. Photogrammetric image processing - loading images into software based on MVS-SfM algorithms; performing phototriangulation of the images to determine their spatial position. This will allow us to subsequently obtain measurements corresponding to the actual object based on the image measurements. As a result of this process, we obtain a simplified three-dimensional model of arbitrary scale; scaling the resulting model based on the average sizes of fundus structures; and constructing a dense polygonal model based on depth maps. The model density is determined by the pixel size of the original image; texturing and editing (cropping) of the final model.

[0087] 3. For small fractures with a diameter of 250 µm and a height of 259 µm, the diameter is equal to the depth, as confirmed by OCT data (the error of the method is -1.4), where a fracture up to 250 µm has a depth of 262 µm. The diameter-to-height ratio of the fracture is 0.96, which is consistent with small fractures.

[0088] According to our classification - small.

[0089] The edges of the rupture have a regular rounded shape, which may indicate the recent development of pathology (less than 2 months).

[0090] The method for assessing macular hole parameters based on stereophotogrammetry has minimal error in comparison with known research methods.

[0091] Diagnosis according to ICD10: Right eye - Retinal tears without retinal detachment.

[0092] Recommended: Microinvasive vitrectomy in combination with lensectomy, IOL implantation, and membrane peeling of the right eye.

[0093] After treatment: Visus OS 0.1 n / c.

[0094] Clinical example 2.

[0095] Patient O., 70 years old. Case history No. 25-297, State Budgetary Health Institution of the Vologda Oblast, "VOKOB"

[0096] Complained of deteriorating vision and image distortion in the left eye.

[0097] Subjective review:

[0098] Visus OS 0.08 n / c according to ETDRS table Log MAR = 1.1

[0099] Data from computer static perimetry - visual fields within the age norm.

[0100] Amsler test - a spot is observed in the center of the grid, lines are not visible in front of the left eye.

[0101] Objective examination data:

[0102] Biomicroscopy with 78D high-resolution ophthalmoscopy lens.

[0103] Right eye - The optic disc is pale pink, with clear borders. Arteries are narrowed, veins are full. No pathology in the central region, no focal pathology in the periphery.

[0104] Left eye - The optic disc is pale pink, with clear borders. Arteries are constricted, veins are congested. There is a macular hole in the central region, and laser coagulation is satisfactory in the periphery.

[0105] OST of the left eye - complete macular hole up to 462 µm wide, 381 µm high.

[0106] Calculating depth without using additional equipment:

[0107] 1. Digital webcams (Leica, Topcon) were used to obtain digital images for creating a 3D model, selection and quality control of images;

[0108] 2. Photogrammetric image processing - loading images into software based on MVS-SfM algorithms; performing phototriangulation of the images to determine their spatial position. This will allow us to subsequently obtain measurements corresponding to the actual object based on the image measurements. As a result of this process, we obtain a simplified three-dimensional model of arbitrary scale; scaling the resulting model based on the average sizes of fundus structures; constructing a dense polygonal model based on depth maps. The model density is determined by the pixel size of the original image; texturing and editing (cropping) the final model.

[0109] 3. For average-sized lesions with a diameter of 462 µm and a height of 381 µm, the diameter is always equal to the depth, as confirmed by OCT data (the error of the method is 1.8), where a rupture greater than 462 µm has a depth of 392 µm. The diameter-to-height ratio of the rupture is 1.21.

[0110] According to the classification we developed, the gap is considered average.

[0111] The edges of the tear are irregular in shape, which may indicate that the tear is 4 months old.

[0112] The method for assessing macular hole parameters based on stereophotogrammetry has minimal error in comparison with known research methods.

[0113] Diagnosis according to ICD10: Left eye - Retinal tears without retinal detachment.

[0114] Indications for surgery: progressive deterioration of vision.

[0115] Recommended: Microinvasive vitrectomy in combination with lensectomy, IOL implantation, and membrane peeling of the left eye.

[0116] After treatment: Visus OS 0.2 n / c.

[0117] Clinical example 3.

[0118] Patient R., 74 years old. Case history No. 25-527, VOKOB Healthcare Institution of the Vologda Oblast

[0119] Complained of deterioration of vision in the right eye for a month.

[0120] Subjective review:

[0121] Visus OD 0.1 n / c according to ETDRS table ETDRS = 1.0

[0122] Data from computer static perimetry (campimetry) of the visual field are within the age norm.

[0123] Amsler test - grid lines are curved in the central area.

[0124] Objective examination data:

[0125] Biomicroscopy with 78D high-resolution ophthalmoscopy lens.

[0126] Right eye - The optic disc is pale pink, with clear borders. Arteries are constricted, veins are congested. There is a macular hole in the central region. Peripherally, there are no abnormalities.

[0127] Left eye - The optic disc is pale pink, with clear borders. Arteries are constricted, veins are full. No abnormalities in the central region, no abnormalities in the periphery.

[0128] OST - complete macular hole up to 762 µm wide, 351 µm high.

[0129] Calculating depth without using additional equipment:

[0130] 1. Digital webcams (Leica, Topcon) were used to obtain digital images for creating a 3D model, selection and quality control of images;

[0131] 2. Photogrammetric image processing - loading images into software based on MVS-SfM algorithms; performing phototriangulation of the images to determine their spatial position. This will allow us to subsequently obtain measurements corresponding to the actual object based on the images' measurements. As a result of this process, we obtain a simplified three-dimensional model of arbitrary scale; scaling the resulting model based on the average sizes of fundus structures; and constructing a dense polygonal model based on depth maps. The model density is determined by the pixel size of the original image; texturing and editing (cropping) the final model.

[0132] 3. For larger fractures with a diameter of 620 µm and a height of 375 µm, the diameter is always greater than the depth, as confirmed by OCT data (the error of the method is 1.6), where a fracture greater than 650 µm has a depth of 360 µm. The diameter-to-height ratio of the fracture is 1.65.

[0133] According to the classification we developed, the gap is considered large.

[0134] The edges of the tear are irregular in shape, which may indicate that the tear has been present for more than 4 months.

[0135] The method for assessing macular hole parameters based on stereophotogrammetry has minimal error in comparison with known research methods.

[0136] Diagnosis according to ICD10: Right eye - Retinal tears without retinal detachment.

[0137] Indications for surgery: progressive deterioration of vision.

[0138] Recommended: Microinvasive vitrectomy in combination with lensectomy, IOL implantation, and membrane peeling.

[0139] After treatment: Visus OS 0.1 n / c.

Images