JP2001522679A - Automatic light reflection screening - Google Patents

Abstract

Summary A system and method for virtually positioning and modeling an eye for automatic photorefractive screening. The present invention includes systems and methods for placing the eyes of a patient (16) in a digital image including each eye illuminated by a paraxial flash (12). Automatic detection of light reflection involves the analysis of such light reflections to determine possible pupil-sclera boundaries. The present invention further comprises automatically applying an associated model to such possible pupil-sclera boundaries and analyzing each eye model to determine possible abnormalities in each eye. Based on such analysis, a possible diagnosis is output for each eye. Another aspect of the invention is to measure retinal and corneal reflexes from a model of the eye displayed as an indicator of a patient's (16 eye abnormalities) and to place a flash (12) positioned near the centerline of the camera lens A digital image of the eye of each patient (16) is generated with a camera (10) equipped with a camera, and an image with bright and clear light reflection is generated.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

TECHNICAL FIELD OF THE INVENTION

FIELD OF THE INVENTION The present invention relates to medical machines for measuring eye characteristics, and in particular, to graphically position and model eyes for automatic light refraction screening and to find abnormalities in the visual tissue of a patient. Systems and methods.

[0002]

[Prior art]

Conventional eye examinations are performed by illuminating the inside of the eye and visually examining the uniformity of the normally visible red colored reflected light that fills the pupil. In one eye,
Or any deviation from homogeneity between the eyes indicates a potential problem that has occurred in the patient's visual tissue. Such light reflection screening is therefore a useful tool in patient eye care. Eye disorders such as strabismus, various forms of reflex errors (myopia, hyperopia, astigmatism) and opacification of the ocular medium are major causes of amblyopia or loss of vision. Together, they cause low vision in 5% of the population. However, some forms of visual problems that do not necessarily cause amblyopia appear in 20% of children.

[0003] Amblyopia does not simply reduce visual acuity. In addition to the loss of cognitive sensitivity, the grid sensitivity, vernier sensitivity, contraction sensitivity, shape distortion, spatial placement,
Movement, worse worse crowding stimulus. In addition, amblyopia is a major cause of monocular vision loss in people under the age of 30. While monocular vision loss does not impact perceptual ability, amblyopia and other aesthetic loss disorders are tremendous emotional impacts. Obviously, the unmet need for eyeglasses makes school performance in severe early grades more difficult. Early school success complicates itself and leads to unsatisfactory school potential. In addition, some occupations require excellent vision. Delaying recognition of various visual impairments limits children from choosing these future occupations. Visual development of strings is thought to consist of three steps. (1) a period of development of visual acuity about 3-5 years old, (2) months to 7-8 years, when effective loss causing amblyopia is effective, (3) partial or complete recovery from amblyopia Period of time (at least when teens have low vision). For the foregoing reasons, clinical intervention for amblyopia is most effective when performed as soon as possible. Over the past two decades, investigators have found that many strabismus and amblyopia conditions cause abnormal visual experience in childhood, which can be prevented or reversed by early detection and intervention. Therefore, the earliest defect recognition is very important.
Some studies have indicated that this must be done during the first two years of life for best results. Nevertheless, previous reports have indicated that primary care physicians do not always utilize the latest screening techniques. In fact, one large study estimated that pediatricians screen less than 40% of children aged 3 or younger. This is due to unrealism from a clinical or practical point of view.

In fact, the National Institutes of Health prioritizes the finding of refractive error, strabismus, and amblyopia in infants and children. This priority requires, among other things, the search for better and more cost-effective public health methods for testing visual function. Accordingly, the inventors have realized the need for automated photorefractive screening. Performing an automated light refraction screening requires an accurate model of the eye. The inventors have determined that it would be beneficial to produce such models in an automated manner, or to present these models as diagnostics, either to a physician or other caregiver. Accordingly, the present invention is directed to systems and methods for virtually positioning and modeling an eye for automated light refraction screening and to enable determination of the presence of an anomaly in a patient's visual tissue. ing. The problem solved by the present invention is different from other eye and face detection problems. Students' ability classes are organized by head-mounted devices (eg, H. Kawai and S. Tamura, "Ey
e movement analysis system using fundus images ", in Patter Recognition,
19 (1), 1986, pp. 77-84), which is done via a fixation of the eye position relative to the image. Eye tracking is based on constrained updates of the eye model (eg, X. X
ie, R. Sukhakar and H. Zhuang, "On improving eye feature extraction usin
g deformable templates ", in Pattern Recognition Letters, 27 (6), 1994, p
p. 791-799: A. Yuille and P. Hallinan, "Deformable Templates", Chapter 2
in Active Vision, MIT Press, 1992, pp. 21-38). Face recognition (for example,
International Conference on Automatic Face and Gesture Recognition, proc
eedings of, edited by M. Bichsel: 1995; Second International Conference
on Automatic Face and Gesture Recognition, Proceedings of, 1996. in Kil
Lington. Vermont) is performed using various image transformations or feature extraction. General image processing references, e.g., K. Castleman, Digital Image Processing, Pent
ice-Hall, 1996; R. Haralick, L. Shapiro, Comuter and Robot Vision, 1,2,
Addison, 1992; and R. Jain, R. Kasturi and B. Schunck, Machine Vsion, Mc
Graw-Hill, 1995) may find some additional information.

[0005]

[Means for Solving the Problems]

A system and method for virtually positioning and modeling an eye for automated light screening. A preferred embodiment of the present invention is a flash with a lens attached to obtain a digital image of the individual, placing the individual's eye in the digital image, modeling the structure inside the eye, and digitizing the individual for eye disease. A digital camera with a suitably programmed processor, such as a versatile digital computer, for analyzing the affected eye and providing treatment recommendations. In particular, according to one aspect, the invention includes systems and methods for positioning a patient's eye in a digital image that includes each eye illuminated by a paraxial flash. Includes auto-sensing light reflections indicating the placement of each eye in the digital image. Auto-sensing light reflection involves the analysis of such light reflections to determine possible pupil-sclera boundaries. The present invention further provides for the automatic fitting of the relevant model to such possible pupil-sclera boundaries, determining possible abnormalities, and making possible diagnoses for each eye based on such analysis. An analysis of each eye model for output is provided. According to another aspect, the present invention comprises displaying the patient's eye, including each eye illuminated by a paraxial flash, with bright dots in the digital image as possible corneal reflections, and for each eye For each eye, find and display a red-black, black-white slope, each with a set of tilt points around such a bright point, as the borderline between each possible pupil and sclera. Fitting multiple circles as possible models to such a subset of tilt points, each eye model has an associated strength, classify the eye model for each eye, and for each eye the strongest association Systems and methods for placement and modeling in a digital image for display as a model of an isolated eye.
Another aspect of the invention involves the generation of a digital image of each patient's eye with a camera equipped with a flash near the central axis of the camera lens to generate an image with bright and clear light reflections.

The target population of the present invention is an unachieved high population, such as school pupils. No special equipment is used to keep the image processing simple and inexpensive. By using the properties of paraxial flash and novel processing techniques, the eye can appear anywhere on the patient's face image (simplifying the use of the present invention by field eye care professionals) and the patient's eye Need not fill the frame of the camera image,
The camera can be placed away from the patient, out of the way of the patient, and there is no need to secure the patient's head. One or more details of the invention are set forth in the accompanying drawings and the description below. Other features, objects, and advantages of the invention will be apparent from the description and drawings, and from the claims.

[0007]

BEST MODE FOR CARRYING OUT THE INVENTION

Image Generation System FIG. 1a is a block diagram illustrating a preferred physical structure of the present invention. Digital camera
10 is an improvement to provide a “paraxial” flash 12 that is positioned somewhat off the optical axis of the lens 14 of the camera 10. The flash 12 is, for example, a camera 10
A ring flash mounted near the optical axis of the lens 14 or one or more small flash devices may be used. A Kodak Model DC120 Digital Camera is suitable. It comprises a 6X telephoto lens 14 and a small flash unit 12 mounted somewhat off-center at the front of the telephoto lens. Telephoto lenses increase the field of view of the camera when photographing the patient's eyes, thereby increasing the number of pixels available for analysis. The close proximity of the flash 12 to the center of the lens 14 provides a bright, sharp light reflection that implements the image processing features of the present invention that accurately and reliably model the patient's eye. The use of a flash rather than continuous illumination means that the pupil can be completely dilated without significant discomfort to the patient. FIG. 1b is a perspective view of a lens 14 and a flash 12 according to the present invention. The flash 12 in the embodiment uses visible light. However, instead of a suitable camera, a flash that emits mainly non-visible light, such as an infrared flash, may be used. The camera 10 is used to capture digital images of the face of a patient 16 and download the images directly to a conventional computer 18 for processing. In a preferred embodiment of the present invention, the camera 10 is a high quality digital image of the patient's face and eyes, preferably full color (eg, RGB), high resolution (eg, 8 bits / pixel for each color). At least about 640 × 480 pixels, preferably at least about 1024 × 768 pixels. Although color images are preferred, the techniques of the present invention can also be used for grayscale images. However, here, a color digital image is used for this explanation.

In another embodiment, camera 10 is a conventional film camera, such as an instant photo camera, the images of which are optically captured by computer 18. In yet another embodiment, the image processing according to the present invention is performed in a digital camera 10 that is specially programmed to allow for an integrated photo screening system.

Photo Screening Photo screening is by taking two separate light reflections from the eye, the corneal reflection and the retinal reflection. Light from the flash 12 is reflected between the air in the cornea of the eye and the tear film. Since the corneal surface is essentially spherical, when reflected from the cornea, the point of the cornea closest to the camera 10 reflects light to the camera. The eye may be squint and the associated light reflection points may be off the center of the pupil of the eye. 2a-2b show some corneal reflection states. FIG. 2a shows the light path from the flash to a pair of normal eyes, and from the eye to the flash, with illustrations of corneal reflections from the camera's field of view. As shown, the light reflection 20 of the cornea is basically concentrated within the outer circumference of the iris 22. FIG. 2b is a diagram showing the corneal reflection from the camera field of view, as well as the optical path from the flash to two abnormally one eyes and from the eye to the flash. As shown, in the left eye, the corneal light reflection 20 is basically concentrated within the outer periphery of the iris, while in the right eye, the corneal light reflection 24 is off-center. Note that the tolerance of "concentration" depends in part on the displacement of flash 12 from the optical axis of camera lens 14. Light from the flash 12 also reflects off the back of the eye through the iris. From the light reflection of the iris, information on the abnormality of the eye and the refraction state of the eye can be obtained. If there is any abnormality in one eye (eg, tumor, bleeding, cataract, etc.), less light will be reflected by the retina than in normal cases. There is a discernable difference in reflectance between the abnormal and normal retina. In an eye without abnormal reflexes, the light reflection of the retina appears as a uniform and clear circle. However, if the eye has abnormal reflexes, the light reflex of the retina appears as a crescent. The crescent shaped reflex size to pupil diameter is generally associated with reflex abnormalities. Therefore, the difference in the observed reflectance is global (difference between the eyes) or local (difference in the visual field of one eye). 2c-2f show some retinal reflection states. FIG. 2c is a diagram showing retinal reflexes 26 of a pair of normal eyes, showing identical and uniform reflectivity in both eyes and no abnormal reflexes. FIG. 2d is a diagram showing retinal reflexes of a pair of eyes, and a decrease in reflectance 28 is seen due to a retinal abnormality in one of the eyes. FIG. 2e is a diagram showing retinal reflexes of a pair of eyes. Here, each shows a crescent-shaped reflection 30, and it can be seen that there is a similar reflection abnormality in both eyes. FIG. 2f shows the retinal reflex of a pair of eyes. Here, the shapes 32, 34 of the binocular crescents are different, and it can be seen that there are different amounts of reflex abnormalities between the eyes.

Eye Model In order to automate photo screening, it is necessary to create an eye model of the patient's face with images. In a preferred embodiment of the invention, the front projection of the eye is modeled by a pair of concentric circles. Formula (1) Formula (2) Where (A, B) are the coordinates of the center of the eye, C ₁ indicates the boundary between the pupil and the iris,
C ² indicates the boundary between the iris and the sclera. This model establishes means for arranging eyes in a two-dimensional coordinate system and measuring reflection from eyes. Once the model parameters have been generated, the model can be used to automatically generate diagnostic or screening decisions for ocular abnormalities. As an example of how such a model "fits" the eye, FIG. 3a is an enlarged photograph of the patient's eye, and FIG. 3b is a patient superimposed image of the eye model used in the present invention. It is an enlarged photograph of the eye. C ₁ indicates the boundary between the iris and the pupil, and C ₂ indicates the boundary between the pupil and the sclera. As another example, FIG. 4a is an enlarged photograph of the eye of a patient having an abnormal pupil region. FIG. 4b shows an enlarged photograph of a patient's eye having an abnormal pupil region superimposed with an image of an eye model used in the present invention. In a preferred embodiment of the present invention, a set of reflections is measured for one eye. Corneal reflection is the reflection of light from the front of the eye (cornea). This generally appears as a bright spot. The corneal reflex CR is modeled as four consecutive regions (described below). In FIG. 3b, the corneal reflex is indicated by an overlaid region CR.
(The corneal reflection CR in FIG. 3b need not necessarily be located in the center of the model.) The corneal reflection always appears in the image of the eye taken by the camera 10 and the flash 12. This fact is
It is used to automatically position the eyes within the image of the patient's face. This can be used even when the eye is not in the field of view of the camera. Retinal (or red) reflex RR is a type of pupil reflex. Pupillary reflex is the light reflection from the inner surface (retina) of the eye returning through the pupil. In a color image, retinal reflexes generally cause the pupil to become reddish and pale. In the preferred embodiment, the retinal reflex RR is modeled as a single light-dark value related to the average amount of red color seen in the pupil. (In a grayscale image, the retinal reflex appears as a set of clearly distinguishable gray values and is modeled similarly).

A crescent-shaped reflection MR is also a kind of pupil reflex. Depending on the curvature of the eye and the position of the flash, a stronger concentration of reflected light (usually a crescent) appears around the border of the pupil. When this appears, the crescent shaped reflection MR is preferably modeled as four consecutive regions (described below). In FIG. 3b, a crescent-shaped reflection is indicated by an overlapping region MR. Another anatomical problem can also cause non-uniform pupil reflex. For example, in cataracts, some of the pupil reflexes appear brighter than normal. In this case, each abnormal pupil region AR is modeled as four continuous regions (described below). In FIG. 4b, one abnormal pupil region is displayed by the overlapping region AR. The four consecutive regions are defined as follows. Formally, it is assumed that an image is composed of a two-dimensional raster of pixels, and is composed of an integer grid of R columns and a column C. Also, let the pixel position in the image be (r, c). Two pixel _{(r i, c i) (} r j, c j) consecutive to the four paths P between is defined as follows. Formula (3) The four continuous areas S are defined as follows. Formula (4)

For simplicity, Equation 4 shows any set of consecutive pixels in I, where any pixel in the set of pixels is any other set of pixels in the set of pixels. From a pixel, it is reached via the path of at least one pixel in the set of pixels. In the preferred embodiment, the possible actual corneal reflex, crescent reflex, and other pupil abnormal regions described herein are all modeled as four consecutive regions. Eye Placement, Modeling, Analysis, and DiagnosisOne aspect of the present invention is the need for digitizing images of a patient's face and eyes to position each eye and build a model for each eye as described above. Is to obtain the appropriate parameters. To this end, the present invention takes advantage of the properties of certain illuminated flashes. This flash produces light reflections that cannot be seen with normal intensity images. The preferred embodiment of the present invention utilizes such light reflection by image processing to position and model the eye to enable automatic light reflection screening. In particular, the camera image of the assumed content generated as described above is effectively used to reduce the complexity of searching for possible eye models. Here, each eye is saturated (
Suppose we have a corneal reflex CR that appears as a (bright) point. Assume that each pupil has some red shading. This red shading greatly tilts the red band at the pupil-iris boundary. (This will be referred to below for convenience as the "red-black" slope. However, similarly the boundaries can be seen in the grayscale image as differences in individual grayscale values. Is assumed to appear as a white shadow, producing a full-colored pronounced slope at the iris-sclera interface). (This is conveniently referred to below as the "black-white" slope. Similar slopes can be identified as distinct grayscale differences in the graysrail image). These three features perform the search process.

FIG. 5 is a flowchart showing the entire steps of the image processing according to the preferred embodiment of the present invention. In a broad sense, the preferred processing steps for positioning, modeling, analyzing, and diagnosing a patient's eye are as follows, each of which is described in further detail below. (1) Generate a digitized image of the patient's eye illuminated with a paraxial flash (step 50). (2) Search for bright spots in the image as possible corneal reflections (step 51). (3) Look for red-black and black-white slopes around bright points as possible pupil and scleral boundaries. (4) As a possible eye model, match the circle with the lower arc of the inclined point (step
53). (5) Classify the eye model by intensity (ie, similar to the exact model of the eye) and choose the two best assumptions (step 54) (this assumes that the patient has two eyes) Assumption). (6) The retinal reflex and the corneal reflex are measured from the selected eye model, and an abnormality, that is, an abnormal reflex is searched (step 55). (7) Recommend appropriate diagnosis (step 56).

(1) Generate digitized images of the patient's face and eyes as illuminated by the paraxial flash As described above, the camera 10 and the paraxial are used to capture images of the patient's faces and eyes. Use flash 12. If the camera 10 is digital, the image data is downloaded directly to the computer 18 for processing as a digital image. camera
If 10 is a conventional film camera, such as an instant photo camera, the images are optically scanned in computer 18 and stored as digital images. View or print the raw image on a computer monitor (black and white or color)
And notify using appropriate conventional graphics software. (2) Search for bright spots in the image as possible corneal reflections (step 51) Each of the three band (ie, RGB) image inputs to the computer 18 provides a separate search for a potential corneal reflection area. Preferably, the threshold value is within the band. To generate a binary image of a bright area, the images are ANDed. Next, using the queue-based paint-fill algorithm, the bright pixels are spatially classified into four consecutive regions. Bright areas having portions within the selected magnitude range are identified as possible or hypothetical as corneal reflexes. Formally, the input image I is a threshold value for generating a binary image I _bright indicating which pixels are bright. Formula (5) _{Here, I red, I green, I} blue is a separate band of the input image, I _bright = 1 indicates that the pixel is bright, T _B is the selected threshold. Next, paint using a queue to spatially classify into four consecutive regions using bright pixels-
Apply the fill algorithm. The pseudo code for this algorithm is shown below (in this case, the image I called this pseudo code is a bright image I _bright ): Algorithm (1) Then, the size of each bright area is counted as the number of pixels with the same label: Equation (6) Here, (r, c) indicates the positions of all pixels in the image. Formula (7) The range that satisfies the above may be corneal reflex, where T _S and T _L are selectable thresholds.

(3) Possible pupil boundaries, scleral boundaries, red-black and black-
Look for white slope For each of the possible corneal reflex (area) LABELs, a set of pixels exhibiting a strong slope at a distance within the expected range of the pupil and iris radius is found in the annular circumference of the corneal reflex Was. This can be achieved by considering the center of the region, along with a set of pixels along the circumference from an evenly distributed center, forming a set of rays.
Pixels were fitted with linear gradient filters along each ray segment. For each segment, the pixel with the largest black-white slope and the largest red-black slope are reported. Any direction where the largest slope point is not within the expected range of distance from the center, or where the black-white slope is closer to the center than the red-black slope, is a deviated pixel outside the region ("out-of-range pixel" ) Is discarded. Formally, the centers (x _c , y _c ) of four consecutive regions with the value LABEL are found as follows: Formula (8) The centroid forms a set of rays with a circumferential conformal set. Where Tθ (Degree) is a selectable algorithm parameter that controls the angular resolution of the set. Pixels (integer coordinates) along each ray θ can be calculated by the following pseudo code by using the following equation (10). And T ₁ and T ₄ are algorithm parameters indicating the minimum and maximum expected radii of the pupil and iris, respectively. Algorithm 2 After such an enumeration, a list of points along each ray θ is shown below. Formula (11) Where Nθ is the number of points in the list returned by the above pseudo code. table 1

The black-white gradient BWθ _{: I} (Equation 12) and the red-black gradient RBθ _{: I} (
To calculate equation 13), the 1 × 7 linear gradient filter shown in Table 1 above is convolved with points along the ray. Formula (12) Formula (13) Here, T2 and T3 are algorithm parameters indicating the highest and lowest expected radii (in pixels) of the pupil and iris, respectively. The pixel with the largest black-white slope RAYθ _{: W} is found with the following equation _: Formula (14) A pixel with a red-black slope RAYθ _{: r} can be found with the following equation _: Formula (15)

These pixels form two lists of points located radially around possible corneal reflections. For each list, there is one point per ray. Formula (16) The list of red-black slope points is given by: Formula (17) These lists form a closed loop, so a circular wraparound
Point RAYθ _{: I} = RAYθ-360 _{: I} = RAYθ _{+ 360: I}

(4) As a possibility of an eye model, a circle equation is applied to the tilt point found in the previous step, using the least squares method that matches the circle to the lower arc of the tilt point. For each possible corneal reflection, one circle is applied to the black-white tilt pixel and one circle is applied to the red-black tilt point. The preferred embodiment of the present invention uses a novel method for removing out-of-range pixels within the applied circle based on the sub-arc.
Next, the elimination of out-of-range pixels as a subset of the arc is naturally known from the context of the problem. If the eye is taken in a direction other than forward, or if the eyelashes cover an area of the iris, or if retinal reflex occurs only in one part of the pupil, the strongest large lower arc By fitting a circle to, one can generally search for the most reliable model. In particular, the least squares method is used to fit a circular equation to eight subsets of various slope points. Each subset starts at a 45 degree increment and covers 270 ° of the arc. For each fitted circle, three intensity measurements, residuals, mean slope, and arc range are calculated. These values are normalized and summed as intensity values. The circle with the global maximum intensity value (in the example, out of eight possible) is chosen as the model for the hypothesized corneal reflex. Formally, one circle models a set of pixels, and Equation (18) Equation (19) Here, a, b, and r are model parameters, and N is the number of points to be modeled. Red
- a pair of inclined point of the black tilt point _{(RAYθ: r,} Equation 17) the pupil from - iris boundary (model (A, B, C _1), Equation 1) for obtaining, also black - white gradient point The following processing is executed to obtain the iris-sclera boundary (model (A, B, C ₂ ), equation 2) from the set (RAYθ _{: w} , equation 16).
In the first case, the notation permutation is (a, b, r) = (A, B, C ₁ ) and (x _i , y _i ) = (xθ _{: r} , yθ _{: r} )
Here, if θ 1 increases, i increases by Tθ, and N = 360 / Tθ. In the second case,
The notation permutation is (a, b, r) = (A, B, C ₂ ) and (x _i , y _i ) = (xθ _{: w} , yθ _{: w} ), where if θ increases by 1, i increases by Tθ, and N = 360 / Tθ. To find the least-squares method that fits the problem, the error term that sums the distance between the model (a, b, r) and each point (x _i , y _i ) is denoted by x. Here, the answer is a model for the X ² to a minimum. The following equation is obtained by expanding Equation 20. Equation (21) By substituting Equation 21 below, Equation (22) Equation (23) It is recognized as linear within the unknowns a, b, α. Given these conditions, the answer to Equation 23 is: (See, for example, W. Press et al, Numerical Recipes in C: The Art of Scie).
See ntific Computing, Chambridge Press, the (2 ^nd edition) 1992. Here, A is N × 3, x is 3 × 1, b is N × 1, and the matrix is configured as follows. Formula (25) The answer to equation 24 is: From this equation, a and b are directly obtained, and are obtained by replacing Equation 22. In a preferred embodiment, a Gauss-Jordan elimination method (eg, W. Press et al., Numerical Recipes in C: The Art of Scientif) is used to solve the inversion matrix of Equation 26.
ic Computing, Cabridge Press, are using the (2 ^nd edition) 1992).

If all points (x _i , y _i ) are “pixels in range” and the noise of the measurement is evenly distributed, Equation 26 gives the optimal answer (best model). Points located through large slopes rarely meet these constraints. For example, FIGS. 6a-6d are enlarged pictures of a patient's eye showing various steps in removing out-of-range pixels. Figure 6a
Is an enlarged photograph of the patient's eye without the model superimposed. FIG. 6b shows a black-white tilt point 60 arranged for the image of FIG. 6a. These pixels illustrate the eyelash spurious response, as well as some unconstrained out-of-range pixels. A common method for fitting the presence of out-of-range pixels is the least median square method (eg, PJ Rousseeuw, "Least median of squares regression", in Jo
urnal of American Statistics Association, Vol. 79, 1984, pp. 871-880, P.
Meer, D. Mintz and A. Rosenfeld, "Robust Regression Methods for Computer
Vision: A Review "in International Journal of Computer Vision, 6: 1, 199
1, pp. 59-70). This algorithm is effective when there are up to 50% out-of-range pixels, but is O (n ² ) in computational complexity. To prevent computational delays, the preferred embodiment uses a combination of the new method and some constraints to eliminate out-of-range pixels in O (n) time. (A) Any pixel RAYθ _{: w} is discarded, and equation (27) (B) Any pixel RAYθ _{: r} is discarded, and equation (28) In both cases (a) and (b), slopes less than 100 include almost invisible boundaries, probably a response to noise. (C) Any pixel that is a copy of another pixel is discarded. Formula (29) Such copied pixels appear when Tθ is too small, resulting in oversampling. (D) In the following cases, any pair of pixels RAYθ _{: w} and RAYθ _{: r} are discarded
Can be Formula (30) This fits the natural constraint that the iris boundary should be outside the pupil boundary. (E) If a pixel is found in isolation, it is discarded. Pixel (x _i , y _i ) is not considered isolated if: Formula (31)

Since all discarded pixels are treated as out-of-range pixels, they are not used in the process to which the least squares method is applied. FIG. 6c is the result of fitting equations 27-31 to the black-white tilt points shown in FIG. 6b. Note that only indeterminate out-of-range pixels have been removed from the image. Certain out-of-range pixels continue to appear (due to the reaction from the eyelashes). For the remaining in-range pixels, Equation 26 is applied to the eight-point subset RAYθ _{: w} and the eight-point subset RAYθ _{: r} . Each of the subsets is formed by a set of points within the aforementioned range of θ. The effect is that each of the subsets includes points on the lower arc of the original sampling circle. Formally, eight subsets of RAYθ _{: w} and eight subsets of RAYθ _{: r} are defined as follows. Formula (32) Formula (33) Each of the partial concentrations represents a ramp point located in a continuous 270 ° arc around the possible corneal reflex, beginning with a 45 ° increase.

To calculate the value, the subset SUBARCσ_:θ_{: w}Or SUBARCσ_:θ_{: r}And the results
(A, b, r), residual error, average gradient strength, and arc range. All values depend on the number of in-range pixels used to fit the circle. N_inThis number, indicated by, is 360
/ Tθ (original set RAYθ_{: w} Or RAYθ_{: r}From the number of pixels in
Images discarded or omitted in any appropriate steps given in 33
Calculate by subtracting prime numbers. The residual is calculated as follows. Formula (34) SUBARCσ_:θ_{: w}Is calculated as follows. Formula (35) SUBARCσ_:θ_{: r}Is calculated as follows. Formula (36) The average range is calculated as follows. Formula (37)Where r is the radius of the fitted circle. Subset is trusted if range <0.3
Is considered impossible (the fitted point is 30% or less than the circle border)
Only spreads). 8 subsets SUBARCσ_:θ_{: r}And the eight subsets SUBARCσ_:θ _{: w} If all are considered unreliable, a possible corneal reflex LABEL is disproved as a candidate eye model. Inclination_BWAnd inclined_RBIs limited to 500, and larger values are set to 500. Value SUBARCσ_:θ_{: r}And SUBARCσ_:θ_{: w}Is calculated as follows. Formula (38)Formula (39)The maximum value in each case is 3. The weight factor was obtained experimentally. At least a subset SUBARCσ of red-black slope points_:θ_{: r}Is the largest part if you can trust
Set SUBARCΔ_:θ_{: r}Is found and is as follows. Formula (40)

The parameters (a, b, r) of the circle fitted to the point SUBARCΔ _: θ _{: r} yield the parameters (Equation 1) for a possible pupil-iris boundary as follows: A = a, B = b, C ₁ = r. The possible iris-sclera boundary radius C ₂ (Equation 2) is calculated as the model radius of BWθ _{: w} , where the model is determined as the maximum number at an integer radius, where The number is calculated as follows: Formula (41) If the black-white slope point SUBARCσ _: θ _{: r of} at least one subset is reliable, the maximum subset SUBARCΔ _: θ _{: r} was determined as follows. Formula (42) FIG. 6d shows the lower arc of the maximum of the black-white slope points from the set shown in FIG. 6c. If there is no reliable subset of the red-black slope points SUBARCσ _: θ _{:: W} , the point SUBARCΔ _: θ _{: W} must be set to obtain the parameters (Equation 2) for the iris-sclera boundary. The parameters of the fitted circle (a, b, r) were used as follows. A = a, B = b, C 2 = r. In this case, the pupil - since the boundary of the iris is known, C ₁ value (Equation 1) is undefined. The total number of possible eye models associated with the possible corneal reflex LABEL is calculated as follows. Formula (43) Note that if no plausible subarc model is found for the pupil-iris circle, the first half of the total number is zero. Although it is possible to model the eye without pupil boundaries, the reduced number reflects reduced certainty in the model.

(5) Classify the eye model by intensity and choose the two best hypotheses Each hypothesized corneal reflex has the overall model intensity (ie, similar to the eye). In the next step, the intensity sets are classified. The two strongest models are selected as accurate eye models. The exact eye models are shown in the output from the computer and include a superimposed graphical representation of each model on the associated eye image. In particular, each of the possible corneal reflex LABELs (Equations 1-2) resulting from the possible eye modeling is an associated number (Equation 43). The two possible eye models with the highest total number are taken as real eye models, within the constraints shown below, and are given by: First, the center of the eye model must be at least two mutually separated eye diameters. Formula (44) Second, the orientation of the patient's face in the image is assumed to be either vertical or horizontal, but in each case a level having one of a parallel set of image boundaries. This is proved by: Formula (45) The orientation of the patient's face in the image provided by the user (upper right, left side, or right side) to determine which model in the strongest pair is associated with the left eye and which model is associated with the right eye Use If no possible eye model pair passes these constraints, or if fewer than two possible eye models are found, the image is reported as atypical and the Processing of the preferred embodiment of the invention stops.

(6) Measure retinal and corneal reflexes from the selected eye model and look for abnormalities or metamorphosis in these reflexes. In a preferred implementation, the best pair of possible eye models is Used to perform some tests to find out if any anomalies are found in the image. Test A: In a normal eye, the corneal reflex is only slightly deviated from the center (some deviation is caused by the distance between the camera flash and the center of the camera lens). In this test, for each eye, the deviation of the corneal reflex is measured as the ratio of the distance between the corneal reflex centroid and the center of the eye to the iris. Formula (46) The reflection is off-center in both corneas, and the patient is not looking directly into the camera, and the computer system indicates so. If only one corneal reflex is off-center, the patient is suffering from strabismus and the computer system indicates that the patient should consult an ophthalmologist. Formally, these conditions are tested as follows: Formula (48) Here, _Ter is a selectable threshold value.

Test B: In normal eyes, the corneal reflex appears equally and evenly in both eyes. To test for arc-like uniformity, for each eye model (left and right), (A, B, C ₁ ,
C ₂ ) there is an associated pair of eccentric circles modeled by (Eq. 1-2), the corneal reflex modeled by CR (Eq. 4), and its association modeled by (x _c , y _c ) There is a centroid (Equation 8). The corneal reflex RR for each eye model excludes pixels labeled as belonging to the corneal reflex and its two deep surrounding boundaries (preferably excluding this boundary to minimize errors). , Is preferably measured as the average red value of pixels within the iris-pupil boundary. Formally, it is used to calculate the set of pixels in the pupil, denoted by PR, and is defined as: Formula (49) The corneal reflection RR is calculated as follows. Formula (50) Here, (r, c) indicates all pixel positions in the image. If the corneal reflections have unequal brightness, the computer system indicates that the patient should be referred to an ophthalmologist. Formally, this is tested as follows: Here, T _rr is a selectable threshold value.

Test C: Pixels (excluding corneal reflex) within the pupil boundary that are brighter than the average intensity inside the pupil are partitioned into ranges. Any region of sufficient size with a suitable perimeter in proximity to the pupil boundary is identified as a crescent reflex. No crescent-shaped reflex is seen in normal eyes. If either eye exhibits a crescent-shaped reflex, the patient has a refractive error and the computer system indicates that the patient should be referred to an ophthalmologist. Any area that is large enough but not adjacent to the pupil boundary is identified as a pupil abnormal region (eg, the area may be crescent shaped) and the computer system informs the ophthalmologist that the patient has Indicates that this should be the case. The pupil-iris boundary line is not located in one eye, and testing for crescent reflex and abnormal pupil range is undefined in the preferred embodiment of the present invention. Formally, the average green intensity in the pupil is denoted by GR and is calculated as follows: Formula (52) The average blue intensity in the pupil is denoted by BR and is calculated as follows: Formula (53) An image of a bright pixel in the _pupil is denoted by I _pupil and is created as follows. Formula (54) Here, I _pupil = 1 indicates that the pixel is bright. The bright pixels are spatially classified into four consecutive regions using the paint-fill pseudocode algorithm using the queues listed above (the pseudocode image I referenced in the pseudocode is In that case I _pupil ). The size of each bright area is counted as the number of pixels with the same label. Formula (55) The perimeter of each bright region is calculated as the number of bright pixels adjacent to the non-bright pixels. Formula (56)

The content of the bright range with multiple boundaries is calculated as the number of bright pixels within a short pupil-iris circle distance. Formula (57) Bright areas are classified according to the following criteria. Formula (58) Formula (59) Formula (60)

(7) Recommend Appropriate Diagnosis FIG. 7 is a flowchart of a preferred test sequence using the present invention after the images have been obtained and the eye has been positioned, modeled, and analyzed. First, a determination is made whether corneal or retinal light reflections are visible in the image taken by camera 10 (step 100). Otherwise, the test needs to be repeated (ie another image is obtained) (step 102). If so, then a determination is made as to whether corneal light reflection is concentrated (step 104). Otherwise, an indication appears that the patient is strabismus and further testing is needed (step 106). Otherwise, a determination is made whether the brightness of the retinal (red) light reflection in both eyes is equal (step 108). Otherwise, the patient's retina is indicated as abnormal and should be submitted for further testing (step 110). Otherwise, it is determined whether a retinal crescent light reflection is present (step 112). If so, then an indication is given that the patient has a refractive error and that further testing should be performed. If not, a determination is made whether another abnormal region exists in the retinal light reflection (step 116). If so, it indicates that the patient may be suffering from media opacity and needs further testing (step 118). Otherwise, the test sequence ends (step 120). Note that other test sequences can be devised, and the tests described below can be performed in a different order and / or can be terminated after any temporary diagnostic steps.

Preferred Imaging Environment Preferred test protocols for human patients include: (1) measuring the distance from camera 10 to the patient, (2) placing the camera over the patient's horizontal center landmark, 3) Take an image. In addition, the inventor has determined the following practical constraints on the preferred image environment for use of the camera 10 of FIG. -One human patient is assumed, so one pair of eyes is assumed per image. This image does not include the eyes of the patient or guardian, since the parent or guardian has a small child on his lap. The arc patient 16 is expected to look directly into the camera lens 14 with both eyes open,
The appropriate reflection from the eye is imaged. Generally, in order for the test to be valid, it is necessary to be able to see the entire pupil of each eye. -The path from the camera 10 to the patient 16 should be free of obstacles. For example, patient 16 should not wear glasses. -Patient 16 is required to look directly into the camera and is imaged using conventional lens 14 at approximately conventional distance from the camera. Images in the range where this stable shape was expected,
Produces visual anatomical features of obvious size. The constraint that the patient 16 must not be too far from the camera 10 does not need to affect the function of the centralized system for positioning the patient's eyes. Rather, this constraint is enforced to ensure that the eye area is imaged by a sufficient number of pixels. • The patient's eyes must be dilated. Generally, normal dilation after 3-5 minutes in a darkened room is sufficient. The ambient lighting level in the room where the picture is taken should be as dark as possible to maintain dilation. The embodiments of the invention described herein are designed to be active in the presence of complex backgrounds, and the images are expected to be reasonably uncluttered. For example,
The background should not include a life-size photograph of the person (partially only the face). Normally, jewelry is OK, but all eye-like shapes should be removed. -The patient's eye is expected to be approximately parallel to the horizontal or vertical image boundaries. Two orientations are allowed in tests for defined visual problems.

Calibration of Algorithm Parameters In the description of our algorithm for locating, modeling, and screening eyes, several thresholds and parameters have been identified. The values for these parameters depend on several calibration factors: the size of the digital image (rows by columns in pixels), the quantization of the image (bits / pixel), the distance from the camera to the object, the camera The optical power of the lens (s) attached to the lens, the displacement of the flash from the center of the lens, the intensity of the flash (lumens). Calibration of the position of the paraxial flash is achieved on an empirical basis. Development is
Includes correction of non-threshold distance from patient. This requires adjusting the position of the flash with respect to the optical axis of the camera lens in a manner that maximizes image screening reflections. If the flash is placed too close to the optical heart, the reaction will be overwhelmed. Conversely, if it is too far away, the reaction will be extinguished. Adjusting the camera light sensor is also essential to ensure a standard number of lumens per flash. We have successfully configured and calibrated two working systems based on different cameras. However, conceptually changing these values does not change the operation of the system of the present invention. From the two configurations tested, we chose the one with greater image resolution. This is because, in general, a larger number of samples (in this case, pixels) gives a more stable result by statistics. In the currently preferred form, Ko captures a 1280x960x24 bit (RGB) image.
I am using a dak DC120 digital camera. The object is approximately 3.5 feet (ie,
(Slightly longer than a meter). This distance helps to avoid altering the refractive properties of the patient's eye by accommodation (ie, muscle-induced alterations in the curvature of the eye by focusing on nearby material) and is not annoying to the patient.

The camera is a built-in 3 × lens with an additional 2 × lens attached to it to obtain an effective 6-magnification magnification. The flash is a standard model from Kodak Corporation consisting of a flash tube housed in a reflector. The flash tube is about 27.25 (length) x about 2.5 mm (diameter). The reflector is about 7.8mm (width) x about 2
It is 5.1 mm (length) x about 5.5 mm (depth). The flash was measured from the center of the reflector, which was also located approximately 17.25 mm away from the optical axis (center of flash tube-center of optical axis). For this camera and flash configuration, the values of the above algorithm parameters were adjusted as follows. T ₁ = 6 T _L = 80 T ₂ = 39 T _B = 200 T ₃ = 17 Tθ = 2.0 T ₄ = 83 Tπ = 20 T _S = 4 T _cr = 0.2

Computer Implementation The invention can be implemented in hardware or software, or in a combination of both (eg, a programmable logic array).
Unless stated otherwise, the algorithms included as part of the present invention are not related to any particular computer or other particular device. In particular, various general-purpose devices can be used with programs written by the techniques described herein. Alternatively, it is more convenient to configure more specialized equipment to perform the necessary method steps. However, the invention relates to a program comprising at least one processor, at least one data storage system (volatile and non-volatile memory and / or storage elements), at least one input device and at least one output device. Preferably, it is implemented in one or more computer programs running on a possible system. Program code has been applied to input data to perform the functions described herein and to generate output information. The output information has been applied to one or more conventional output devices. Each such program may be implemented in any desired computer language (including machines, assemblies, high-level procedures, or object-oriented programming languages) for communicating computer systems. In any case, the language may be a compiled or translated language. Each such computer program preferably executes the procedures described herein, and configures and operates a computer when the storage medium or device is read by the computer, preferably a general or special purpose computer. It is preferably stored on a storage medium or device (eg, ROM, CD-ROM, or magnetic or optical storage medium) readable by a programmable computer for various purposes.

A number of embodiments of the invention have been described. Nevertheless, it will be understood that various modifications may be made without departing from the spirit and scope of the invention. For example, certain steps for certain purposes described above may be absolute or relatively
It can be performed in any order, and thus in an order other than the order described above. As another example, other diagnostic tests may be performed or devised using the eye model generated by the present invention. As yet another example, the systems and methods of the present invention can be used with non-human patients. Therefore, other embodiments are also within the claims.

[0034]

[Brief description of the drawings]

FIG. 1a is a block diagram illustrating a preferred physical structure of the present invention used to image a patient's eye.

FIG. 1b is a perspective view of a lens 14 and a flash 12 according to the present invention.

FIG. 2a is a diagram showing the light path from the flash to a pair of normal eyes and back to the camera, as well as the appearance of corneal reflections from the camera's field of view.

FIG. 2b is a diagram showing the light path from a flash to a pair of eyes with abnormalities, and back to the camera, as well as the appearance of corneal reflections from the camera's field of view.

FIG. 2c is a diagram of a retinal reflection 26 for a pair of normal eyes, showing uniform reflections equal in both eyes and without reflection errors.

FIG. 2d is a retinal reflex diagram for a pair of eyes with retinal pathologies, one of which produces a low reflex.

FIG. 2e is a diagram of retinal reflection for a pair of eyes, each exhibiting a crescent-shaped reflectance, showing the same type of reflection error in both eyes.

FIG. 2f is a diagram of retinal reflexes for two eyes exhibiting different crescent reflectivities, representing differences in the amount of reflection error in the eyes.

FIG. 3a is a close-up photograph of a patient's eye.

FIG. 3b is a close-up photograph of a patient's eye with a superimposed graphical representation of the model used in the present invention.

FIG. 4a is a close-up photograph of a patient's eye including an abnormal pupil area.

FIG. 4b is a close-up photograph of a patient's eye including an abnormal pupil area, with a graphical representation of the eye model used in the present invention.

FIG. 5 is a flowchart showing the entire image processing steps according to the preferred embodiment of the present invention.

FIG. 6a is a close-up photograph of the patient's eye without superimposing the model.

FIG. 6b is a close-up photograph of the patient's eye showing the black-white tilt points arranged in the image in FIG. 6a.

FIG. 6c is a close-up photograph of a patient's eye illustrating the result of adding a constraint to the black-white tilt point shown in FIG. 6b.

FIG. 6d is a close-up photograph of the patient's eye showing the highest scoring subset of black-white slope points from the set shown in FIG. 6c.

FIG. 7 is a flowchart of a preferred test sequence using the present invention. Once the image is obtained, the eyes are positioned and modeled. Like reference numerals are used for like elements in the various figures.

──────────────────────────────────────────────────続 き Continuation of front page (81) Designated country EP (AT, BE, CH, CY, DE, DK, ES, FI, FR, GB, GR, IE, IT, LU, MC, NL, PT, SE ), OA (BF, BJ, CF, CG, CI, CM, GA, GN, GW, ML, MR, NE, SN, TD, TG), AP (GH, GM, KE, LS, MW, SD, SZ, UG, ZW), EA (AM, AZ, BY, KG, KZ, MD, RU, TJ, TM), AL, AM, AT, AU, AZ, BA, BB, BG, BR, BY, CA, CH, CN, CU, CZ, DE, DK, EE, ES, FI, GB, GD, GE, GH, GM, HR, HU, ID, IL, IS, JP, KE, KG , KP, KR, KZ, LC, LK, LR, LS, LT, LU, LV, MD, MG, MK, MN, MW, MX, NO, NZ, PL, PT, RO, RU, SD, SE, SG, SI, SK, SL, TJ, TM, TR, TT, UA, UG, UZ, VN, YU, ZW (72) Inventor Brady, Barbara La Jolla, California, United States of America USA (72) Inventor Birch, Dark -Uwe, San Diego, California, United States of America (72) Inventor Granette, David San Diego, California, United States [continued abstract]

Claims (32)

[Claims] 1. A computer-implemented method for placing a patient's eyes in a digital image that includes each eye illuminated by a paraxial flash, the method comprising: a light reflection in the digital image indicating a position of each eye. Automatically searching for. 2. The method of claim 1, wherein automatically retrieving light reflections comprises analyzing such light reflections to determine possible pupil-sclera boundaries. The method according to 1. 3. The method of claim 2, further comprising the step of automatically fitting an associated model to such possible pupil-sclera boundaries. 4. The method of claim 3, further comprising the step of analyzing a model of each eye to determine the likelihood of an abnormality in each eye. 5. The method of claim 4, further comprising outputting a possible diagnosis to each eye based on such analysis. 6. A computer-implemented method for modeling a patient's eye in a digital image that includes each eye illuminated by a paraxial flash, the digital image as possible light reflections from each eye. Automatically fitting a model to each eye of the digital image based on the location of the light reflections in the eye. 7. The method of automatically fitting a model to a digital image, comprising the step of analyzing such possible light reflections to determine possible pupil-sclera boundaries. Item 7. The method according to Item 6. 8. The method of claim 6, further comprising analyzing a model of each eye to determine possible abnormalities in each eye. 9. The method of claim 6, further comprising outputting a possible diagnosis for each eye based on such analysis. 10. A computer-implemented method for positioning and modeling a patient's eye in a digital image including each eye illuminated by a paraxial flash, comprising: (a) as a possible corneal reflection of each eye; (B) searching for and indicating bright spots in the digital image; and (b) as possible pupil-sclera boundaries around such bright spots, red-black, black-black, each having a set of tilt points. (C) fitting a plurality of circles to a subset of such tilt points of a possible model for each eye, wherein the model of each eye is (D) classifying the eye model by intensity for each eye, and indicating the strongest associated eye model as the best representation of each eye; Features and how to. 11. The method of claim 10, further comprising measuring retinal and corneal reflexes from each of the eye models indicated as indicators of possible abnormalities in each eye. . 12. The method of claim 11, further comprising outputting a possible diagnosis for each eye based on such measurements. 13. A computer-implemented method for positioning and modeling a patient's eye, comprising: (a) illuminating the patient's eye with a paraxial flash; and (b) digitizing having each eye. Generating an image; (c) searching for and indicating a bright spot in the digital image as a possible corneal reflection of each eye; and (d) for each eye, a possible pupil-sclera boundary. Around such a bright point as a line, look for red-black, black-white slopes, each having a set of slope points, and indicate: (e) this as a possible model for each eye For a subset of such slope points,
Fitting a plurality of circles, wherein each eye model has an associated intensity; and (f) for each eye, classifying the eye model by intensity and identifying a strongest associated eye model. Indicating the best of each eye. 14. The method of claim 13, further comprising measuring retinal and corneal reflexes from each of the indicated eye models as indicators of possible abnormalities in each eye. . 15. The method of claim 14, further comprising outputting a possible diagnosis for each eye based on such measurements. 16. A computer-implemented method for positioning and modeling a patient's eye, comprising: (a) having a flash positioned near a centerline of a camera lens that produces an image with bright, clear light reflections. Generating a digital image of each of the patient's eyes with a camera; (b) searching for and indicating a bright spot as a possible corneal reflection of each eye; and (c) for each eye a possible pupil and Around such a bright point as the boundary of the sclera, looking for and indicating red-black, black-white slopes each having a set of slope points, and (d) as a possible model for each eye For a subset of these slope points,
Fitting a plurality of circles, wherein each eye model has an associated intensity; and (e) for each eye, classifying said eye model by intensity and identifying the strongest associated eye model. Indicating the best of each eye. 17. The method of claim 16, further comprising measuring retinal and corneal reflexes from each of the indicated eye models as indicators of possible abnormalities in each eye. the method of. 18. The method of claim 17, further comprising outputting a possible diagnosis for each eye based on such measurements. 19. The step of finding and displaying bright spots in the digital image as possible corneal reflections, comprising: (a) spatially classifying bright pixels of each digital image into four consecutive ranges. And (b) identifying such four consecutive regions as regions within a size selected as possible corneal reflections. The described method. 20. The method of claiming and displaying red-black, black-white slopes comprises: (a) determining a center for each possible corneal reflection; and (b) a set of compasses from centroids. Along the direction, together with the pixels of the digital image, define a set of rays as each center; (c) applying a gradient filter to each ray segment; (d) for each such segment: Determining the pixel with the highest red-black slope and the pixel with the highest black-white slope, wherein the set of pixels with the highest red-black slope from all segments is 20. The set of pixels having the highest black-white slope from all segments having a possible pupil boundary and having a possible scleral boundary. the method of. 21. Fitting a plurality of circles to such a subset of tilt points as possible eye models, comprising: (a) for each possible corneal reflection, at least one of the sub-arc forms Applying a least squares method to fit the circular equation to a slope point having a possible pupil boundary and a possible scleral boundary; and (b) a circle intensity value for fitting each of at least one circular equation. 21. The method of claim 20, comprising: calculating. 22. The step of classifying the eye model by intensity, comprising: (a) normalizing and storing circle intensity for each eye; and (b) best circle for the associated eye. 22. The method of claim 21, comprising: selecting a circular equation having an intensity score. 23. The step of measuring retinal and corneal reflexes from an eye model, each indicated as an indicator of a possible abnormality in each eye, wherein the retinal reflexes appear equally and equally in both eyes. 15. The method according to claim 11 or 14, comprising the step of determining and displaying. 24. Retinal and corneal reflexes from each of the displayed eye models,
15. The method of claim 11 or claim 14, wherein measuring as a possible anomaly indicator in each eye comprises determining and displaying if only one corneal reflex is off-center. Method. 25. The retinal and corneal reflexes from each displayed eye model,
15. The method of claim 11 or 14, wherein measuring as an indicator of a possible abnormality in each eye comprises determining and displaying if a crescent reflex is present in the eye. 26. The retinal and corneal reflexes from each of the displayed eye models,
Measuring as an indicator of a possible abnormality in each eye, the determining and displaying steps if the pupil abnormal region in the eye is located away from a plurality of boundaries displayed by the associated eye model. 12. The method according to claim 11, wherein
Or the method of 14. 27. A system for placing a patient's eye, modeling the patient's eye, and determining the presence of an anomaly in the patient's visual tissue, comprising: (a) a bright, clear light reflection; A camera having a flash positioned near the optical axis of the camera lens to generate a digital image of each patient's eye; and (b) automatically generating light reflections indicating the position of each eye in the digital image. A computer programmed to search. 28. The system of claim 27, wherein the computer is further programmed to analyze such light reflections to determine possible pupil-sclera boundaries. 29. The system of claim 28, wherein the computer is further programmed to fit an associated model to such possible pupil-sclera boundaries. 30. The system of claim 29, wherein the computer is further programmed to analyze each eye model to determine possible abnormalities in the eye. 31. The system of claim 30, wherein the computer is further programmed to output a possible diagnosis in the eye based on such analysis. 32. A camera for generating a digital image of each patient's eye, comprising: (a) an eccentric digital camera with an optical lens; and (b) an optical axis of the lens during use of the camera. The digital image generated by the camera has a bright, clear light reflection from the eye, with a flash positioned nearby, and such a bright, clear light reflection is used to position and model the eye. Camera capable of being analyzed by the computer system of claim 1.