U.S. patent application number 11/064382 was filed with the patent office on 2005-09-15 for system and method for analyzing wavefront aberrations.
Invention is credited to Dreher, Andreas W., Lai, Shui T..
Application Number | 20050200809 11/064382 |
Document ID | / |
Family ID | 34886263 |
Filed Date | 2005-09-15 |
United States Patent
Application |
20050200809 |
Kind Code |
A1 |
Dreher, Andreas W. ; et
al. |
September 15, 2005 |
System and method for analyzing wavefront aberrations
Abstract
A system and method for specifying a vision correction
prescription for a patient's eye, wherein in one embodiment, the
method includes obtaining a wavefront aberration measurement of the
patient's eye, applying at least one value from the wavefront
aberration measurement to a statistical model trained using a
plurality of objectively measured aberration values and a plurality
subjectively measured visual acuity values as training data; and
predicting a vision correction prescription for the patient's eye
based on the at least one value and the statistical model.
Inventors: |
Dreher, Andreas W.;
(Escondido, CA) ; Lai, Shui T.; (Encinitas,
CA) |
Correspondence
Address: |
MORRISON & FOERSTER LLP
3811 VALLEY CENTRE DRIVE
SUITE 500
SAN DIEGO
CA
92130-2332
US
|
Family ID: |
34886263 |
Appl. No.: |
11/064382 |
Filed: |
February 22, 2005 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
60546378 |
Feb 20, 2004 |
|
|
|
Current U.S.
Class: |
351/246 ;
351/205 |
Current CPC
Class: |
B29D 11/00528 20130101;
G02C 13/003 20130101; B29D 11/00442 20130101; G02C 7/02 20130101;
A61B 3/0025 20130101; G02C 2202/22 20130101; G02C 7/027 20130101;
B29D 11/00413 20130101 |
Class at
Publication: |
351/246 ;
351/205 |
International
Class: |
A61B 003/10; A61B
003/00 |
Claims
What is claimed is:
1. A method for specifying a vision correction prescription for a
patient's eye, comprising: obtaining a wavefront aberration
measurement of a patient's eye; applying at least one value derived
from the wavefront aberration measurement to a statistical model
trained to analyze the at least one value; and predicting a vision
correction prescription for the patient's eye based at least in
part on the analysis of the at least one value by the statistical
model.
2. The method of claim 1 wherein the statistical model is trained
using a plurality of objectively measured aberration values and a
plurality subjectively measured visual acuity values as training
data.
3. The method of claim 2 wherein the statistical model is further
trained using patient personal profile data.
4. The method of claim 3 wherein the patient personal profile data
comprises at least one data parameter selected from a group of data
parameters consisting of: patient age, patient preferences, patient
prescription history and side effects reported by a patient.
5. The method of claim 2 wherein the plurality of subjectively
measured visual acuity values comprises best corrected visual
acuity measurements from a plurality of patients whose eyes have
been measured to also provide the plurality of objectively measured
aberration values.
6. The method of claim 1 wherein the measured wavefront aberration
may be expressed with Zernike polynomials of the third order or
higher.
7. The method of claim 1 wherein the statistical model comprises a
neural network.
8. The method of claim 1 wherein the predicted visual correction
prescription comprises spherical, cylindrical and axis parameters
associated with a standard prescription.
9. The method of claim 8 further comprising using the predicted
vision correction prescription to corroborate a subjectively
measured prescription.
10. The method of claim 8 wherein the predicted vision correction
prescription further prescribes adjusting at least one high order
aberration associated with the patient's eye.
11. The method of claim 10 wherein the step of adjusting comprises
reducing or increasing the at least one high order aberration.
12. The method of claim 10 wherein the step of adjusting comprises
introducing the at least one high order aberration.
13. The method of claim 10 further comprising providing a graphic
illustration of the at least one high order aberration.
14. The method of claim 13 wherein the graphic illustration
comprises a color-coded table depicting a relative importance of
the at least one high order aberration.
15. The method of claim 14 wherein the at least one high order
aberration comprises one or more high order aberrations selected
from a group consisting of: spherical aberration, coma, trefoil,
tetrafoil and astigmatic tetrafoil.
16. The method of claim 1 wherein the statistical model is
customized for a particular category of eye types.
17. The method of claim 1 wherein the wavefront aberration
measurement provides one or more measurements of high order
aberrations selected from a group consisting of: spherical
aberration, coma, trefoil, tetrafoil and astigmatic tetrafoil.
18. A method of predicting a difference in visual acuity resulting
from high order correction versus standard correction of a
patient's eyes, the method comprising: obtaining a wavefront
aberration measurement of a patient's eye; inputting at least one
value derived from the wavefront aberration measurement into a
statistical model trained to analyze the at least one value; and
predicting a difference in visual acuity resulting from high order
correction versus standard correction of the patient's eye, based
on the at least one value and the statistical model.
19. The method of claim 18 wherein the statistical model is trained
using a plurality of objectively measured aberration values and a
plurality of difference values indicative of differences between
visual acuity of patients having high order aberration corrected
vision and visual acuity of the same respective patients having
standard corrected vision.
20. The method of claim 19 wherein the statistical model is further
trained using patient personal profile data.
21. The method of claim 20 wherein the patient personal profile
data comprises at least one data parameter selected from a group of
data parameters consisting of: patient age, patient preferences,
patient prescription history and side effects reported by a
patient.
22. The method of claim 18 further comprising illustrating the
difference visually.
23. The method of claim 22 wherein the step of illustrating
comprises providing a first image of a Snellen chart indicative of
standard corrected vision for the patient and a second image of a
Snellen chart indicative of high order aberration corrected vision
for the patient.
24. The method of claim 18 wherein the statistical model is capable
of predicting which measured high order aberrations of the
patient's eye are beneficial, detrimental or neutral.
25. The method of claim 24 further comprising providing a vision
correction prescription, wherein the vision correction prescription
prescribes adjusting at least one high order aberration measured
for the patient's eye based on whether the at least one aberration
is predicted to be beneficial, detrimental or neutral.
26. The method of claim 25 wherein the step of adjusting comprises
reducing or increasing at least one high order aberration predicted
to be detrimental or beneficial, respectively.
27. The method of claim 25 wherein the step of adjusting comprises
introducing at least one high order aberration predicted to be
beneficial
28. The method of claim 25 further comprising providing a graphic
illustration of the at least one high order aberration.
29. The method of claim 28 wherein the graphic illustration
comprises a color-coded table depicting whether the at least one
high order aberration is beneficial, detrimental or neutral.
30. The method of claim 25 wherein the at least one high order
aberration comprises one or more high order aberrations selected
from a group consisting of: spherical aberration, coma, trefoil,
tetrafoil and astigmatic tetrafoil.
31. A method of creating a statistical model for use in predicting
vision correction prescriptions for patients, the method
comprising: obtaining a plurality of wavefront aberration
measurements from a plurality of patient's eyes; obtaining a
plurality of visual acuity measurements from the plurality of
patients; applying values associated with the plurality of
wavefront measurements to an input layer of a statistical model;
applying values associated with the plurality of visual acuity
measurements to an output layer of the statistical model; and
generating a plurality of weight values associated with respective
nodes of the statistical model based on the applied values
associated with the plurality of wavefront measurements and
corresponding values associated with the plurality of visual acuity
measurements.
32. The method of claim 31 wherein the plurality of wavefront
aberration measurements comprises measurements of one or more high
order aberrations selected from a group consisting of: spherical
aberration, coma, trefoil, tetrafoil and astigmatic tetrafoil.
33. The method of claim 31 wherein the statistical model is created
from training data received from a plurality of patients having a
common eye type.
34. The method of claim 31 wherein the statistical model comprises
a neural network.
35. The method of claim 31 wherein the values associated with the
plurality of visual acuity measurements comprises a plurality of
difference values indicative of differences between visual acuity
of patients having high order aberration corrected vision and
visual acuity of the same respective patients having standard
corrected vision.
36. The method of claim 35 wherein the statistical model is capable
of predicting which measured high order aberrations of the
patient's eye are beneficial, detrimental or neutral.
37. A data structure created using the method of claim 31.
38. A computer readable medium that stores computer executable
code, which when executed performs a method for specifying a vision
correction prescription for a patient's eye, the method comprising:
obtaining a wavefront aberration measurement of a patient's eye;
applying at least one value from the wavefront aberration
measurement to a statistical model trained to analyze the at least
one value; and predicting a vision correction prescription for the
patient's eye based on the at least one value and the statistical
model.
39. The computer readable medium of claim 38 wherein the
statistical model is trained using a plurality of objectively
measured aberration values and a plurality subjectively measured
visual acuity values as training data.
40. The computer readable medium of claim 39 wherein the
statistical model is trained using a plurality of objectively
measured aberration values and a plurality of difference values
indicative of differences between visual acuity of patients having
high order aberration corrected vision and visual acuity of the
same respective patients having standard corrected vision.
41. The computer readable medium of claim 40 wherein the method
further comprises predicting a difference in visual acuity
resulting from high order correction versus standard correction of
a patient's eyes, based on the at least one value and the
statistical model.
42. A system for specifying a vision correction prescription for a
patient's eye, comprising: means for obtaining a wavefront
aberration measurement of a patient's eye; means for analyzing at
least one value derived from the wavefront aberration measurement
to predict a vision correction prescription for the patient's eye
based at least in part on the at least one value.
43. The system of claim 42 wherein the means for analyzing
comprises a statistical model.
44. The system of claim 43 wherein the statistical model comprises
a neural network.
45. The system of claim 42 wherein the measured wavefront
aberration may be expressed with Zernike polynomials of the third
order or higher.
46. The system of claim 42 wherein the predicted visual correction
prescription comprises spherical, cylindrical and axis parameters
associated with a standard prescription.
47. The system of claim 46 wherein the predicted vision correction
prescription further prescribes adjusting at least one high order
aberration associated with the patient's eye.
48. The system of claim 47 wherein adjusting comprises minimizing
the at least one high order aberration.
49. The system of claim 47 wherein adjusting comprises increasing
the at least one high order aberration.
50. The system of claim 47 wherein adjusting comprises introducing
the at least one high order aberration.
51. The system of claim 42 further comprising means for providing a
graphic illustration of the at least one high order aberration.
52. The system of claim 51 wherein the graphic illustration
comprises a color-coded table depicting a relative importance of
the at least one high order aberration.
53. The system of claim 42 wherein the means for analyzing
comprises a statistical model capable of predicting whether a
measured high order aberration of the patient's eye is beneficial,
detrimental or neutral.
54. A spectacle lens constructed in accordance with the predicted
vision correction prescription obtained using the method of claim
1.
55. A contact lens constructed in accordance with the predicted
vision correction prescription obtained using the method of claim
1.
56. An intraocular lens constructed in accordance with the
predicted vision correction prescription obtained using the method
of claim 1.
57. A method of correcting a patient's vision comprising performing
a surgical procedure on the patient's eye in accordance with the
predicted vision correction prescription obtained using the method
of claim 1.
Description
RELATED APPLICATIONS
[0001] The present application claims the benefit of priority under
35 U.S.C. .sctn. 119(e) to U.S. Provisional Patent Application Ser.
No. 60/546,378, entitled "SYSTEM AND METHOD FOR ANALYZING WAVEFRONT
ABERRATIONS," filed on Feb. 20, 2004.
[0002] This application is related to U.S. patent application Ser.
No. 10/218049, entitled "Apparatus and Method of Correcting High
order Aberrations of the Human Eye," filed Aug. 12, 2002; U.S.
patent application Ser. No. 10/076218, entitled "Apparatus and
Method for Determining Objective Refraction using Wavefront
Sensing," filed Feb. 13, 2002, issued as U.S. Pat. No. 6,761,454;
and U.S. patent application Ser. No. 10/014037, entitled "System
and Method for Wavefront Measurement," filed Dec. 10, 2001, issued
as U.S. Pat. No. 6,781,681. Each of the foregoing patents and
applications is hereby incorporated by reference in its
entirety.
BACKGROUND OF THE INVENTION
[0003] 1. Field of the Invention
[0004] The present invention relates to a system and method for
correcting aberrations of the human eye. More particularly, the
invention relates to method and system for statistically
correlating objective measurements of aberrations associated with
the human eye with subjective measurements of visual acuity to
create a statistical model that can be used to predict vision
correction prescriptions based at least in part on objectively
measured aberrations of a patient's eye.
[0005] 2. Description of the Related Art
[0006] The human eye, namely the cornea and lens, can exhibit a
variety of optical aberrations that diminish the optical
performance of the eye, resulting in blurred vision. The correction
of blurred vision by lenses has typically been limited to
correction of low order aberrations only, such as defocus and
astigmatism. Traditionally, high order aberrations, e.g. those
describable with Zernike polynomials of the third order or higher,
could not be corrected using lenses. In addition, due to lens
manufacturing limitations and expenses, defocus and astigmatism are
typically only corrected in discrete steps, with any correction
being made to the nearest one quarter (1/4) diopter. Unfortunately,
the resolution of one quarter (1/4) diopter results in incomplete
vision corrections and limits the performance of the patient's
eye.
[0007] Recently, systems employing wavefront aberrometry, such as
described in the co-pending patent applications incorporated
herein, have enabled the high order aberrations in a patient's eyes
to be measured objectively, accurately, and rapidly. When coupled
with a custom aberrator that is adapted to correct some or all of
the high order aberrations in a patient's eyes, such as also
described in the co-pending patent applications incorporated
herein, a patient's visual acuity may be improved to as well as
20/10.
BRIEF SUMMARY OF THE INVENTION
[0008] The invention relates to correlating or establishing a
relationship between objectively measured values or parameters of a
patient's eyes with subjectively measured or reported values or
parameters of visual performance.
[0009] In one embodiment, this correlation is performed via machine
learning techniques that create a statistical model (e.g., a neural
network).
[0010] In another embodiment, a method for specifying a vision
correction prescription for a patient's eye, comprises: obtaining a
wavefront aberration measurement of a patient's eye; applying at
least one value from the wavefront aberration measurement to a
statistical model trained to analyze the at least one value; and
predicting a vision correction prescription for the patient's eye
based at least in part on the analysis of the at least one value by
the statistical model.
[0011] In a further embodiment, a method of predicting a difference
in visual acuity resulting from high order correction versus
standard correction of a patient's eyes, comprises: obtaining a
wavefront aberration measurement of a patient's eye; inputting at
least one value derived from the wavefront aberration measurement
into a statistical model trained to analyze the at least one value;
and predicting a difference in visual acuity resulting from high
order correction versus standard correction of a patient's eyes,
based on the at least one value and the statistical model.
[0012] In yet another embodiment, the invention provides a method
of creating a statistical model for use in predicting vision
correction prescriptions for patients, the method comprising:
obtaining a plurality of wavefront aberration measurements from a
plurality of patient's eyes; obtaining a plurality of visual acuity
measurements from the plurality of patients; applying values
associated with the plurality of wavefront measurements to an input
layer of a statistical model; applying values associated with the
plurality of visual acuity measurements to an output layer of the
statistical model; and generating a plurality of weight values
associated with respective input nodes of the input layer based on
the applied values associated with the plurality of wavefront
measurements and corresponding values associated with the plurality
of visual acuity measurements.
[0013] In a further embodiment, the invention provides a learned
data structure created using the above method of creating a
statistical model.
[0014] In a further embodiment, the invention provides a computer
readable medium that stores computer executable code, which when
executed performs a method for specifying a vision correction
prescription for a patient's eye, the method comprising: obtaining
a wavefront aberration measurement of a patient's eye; applying at
least one value derived from the wavefront aberration measurement
to a statistical model trained to analyze the at least one value;
and predicting a vision correction prescription for the patient's
eye based on the at least one value and the statistical model.
[0015] In another embodiment, a system for specifying a vision
correction prescription for a patient's eye, comprises: means for
obtaining a wavefront aberration measurement of a patient's eye;
and means for analyzing at least one value derived from the
wavefront aberration measurement to predict a vision correction
prescription for the patient's eye based at least in part on the at
least one value.
BRIEF DESCRIPTION OF THE DRAWINGS
[0016] FIG. 1 graphically depicts one embodiment of a neural
network for predicting visual acuity based on measured aberrations
in a patient's eye.
[0017] FIG. 2 is a perspective view of one embodiment of an
apparatus for measuring wavefront aberrations in a human eye.
[0018] FIG. 3 is a detailed perspective view of a portion of the
apparatus of FIG. 2 being operated in conjunction with a
patient.
[0019] FIG. 4 is a block diagram depicting the components of the
apparatus of FIG. 2.
[0020] FIG. 5 is a flow chart depicting one embodiment of a process
for predicting a standard prescription in a patient by using an
embodiment of a neural network similar to that of FIG. 1.
[0021] FIG. 6 is a flow chart depicting an embodiment of a process
for using an embodiment of the neural network similar to that in
FIG. 1 to help determine the relative importance of correcting
aberrations found in a patient's eyes.
[0022] FIG. 7 illustrates a color-coded table and graphic images of
high order aberrations in accordance with one embodiment of the
invention.
[0023] FIG. 8 depicts an embodiment of a process using an
embodiment of a neural network similar to that of FIG. 1 for
predicting what level of improvement in visual acuity may be
obtained by a patient by choosing eyeglasses or contact lenses
containing custom wavefront aberrators.
[0024] FIG. 9 illustrates a graphic depiction of side-by-side
comparative images of blurred vision resulting from standard
correction compared to corrected vision resulting from high order
aberration correction of a patient's eye, in accordance with one
embodiment of the invention.
DETAILED DESCRIPTION OF THE INVENTION
[0025] The following detailed description is directed to certain
specific embodiments of the invention. However, the invention can
be embodied in a multitude of different ways as defined and covered
by the claims. In this description, reference is made to the
drawings wherein like parts are designated with like numerals
throughout.
[0026] One practical disadvantage of the objective wavefront
aberration measurements, expressed in terms of Zernike polynomials,
is that these polynomials may be difficult to correlate with a
subjective refraction measurement made by a eye care professional,
such as from using a phoropter, a Snellen chart, and a patient's
feedback. This tends to make it difficult for the eye care
professional to make an informed recommendation to the patient of a
course of action for correcting the patient's vision.
[0027] Theoretically, wavefront aberrometry measurements of
low-order aberrations in a patient's eyes should correspond to the
traditional prescription of sphere, cylinder, and axis. However,
interaction of high order aberrations may contribute to the
patient's perceived blur on the retina, and therefore should be
accounted for in predicting "subjective" measures of visual
performance. Subjective measures of visual performance include, for
example, Snellen visual acuity, based on a patient's ability to
read letters of decreasing size in rows of a Snellen chart, and
contrast sensitivity.
[0028] Ideally, if all aberrations of a patient's eye are
corrected, a diffraction-limited image is formed on the retina of
the patient's eye. The visual acuity of the eye is then determined
by the entrance size of the pupil. Thus, methods of predicting
subjective measures of visual performance from objective high order
aberration data have focused on algorithms based on maximizing the
optical quality (or minimizing the blur) of light being focused on
a patient's eye.
[0029] However, performance of a human's visual system is dependent
on both the optical properties of the eye and on the neurological
response by the patient's retina and brain. Moreover, it has been
found that the presence of some degree of certain high order
aberrations, which technically increase the blur at a patient's
retina, actually improve subjective measures of a patient's visual
performance. For example, studies of the eyes of people with 20/15
or better natural visual acuity have found the presence of high
order aberrations to be relatively common despite these people's
high visual acuity. This result contradicts the theory behind using
optical quality based algorithms to predict visual acuity. Rather,
subjective visual acuity is therefore determined by many factors
such as the observable wavefront aberrations of a patient's eyes,
interactions between each aberration in a patient's eyes, and
interactions between the aberrations and the patient's neurological
visual processing path.
[0030] Machine learning techniques have been applied in a variety
of fields, including financial forecasting, business decision
making, medical diagnosis, and pattern recognition, to make
predictions where direct observation of all relevant factors is not
possible. In particular, inductive learning techniques take as
input a training set of observed data points to "learn" an
equation, a set of rules, or some other data structure. This
learned structure or statistical model may then be used to make
generalizations about the training set or predictions about new
data. As used herein, "statistical model" refers to any learned
and/or statistical data structure that establishes or predicts a
relationship between two or more data parameters (e.g., inputs and
outputs). Although the invention is described below with reference
to neural networks, other types of statistical models may be
employed in accordance with the present invention.
[0031] For example, each data point of the training data set may
include a set of values that correlate with, or predict, another
value in the data point. Thus, in a medical diagnosis application,
a set of biometrics, such as blood pressure and age, of heart
attack patients may predict a five-year likelihood of another heart
attack. By using, as a training set, data points consisting of
measurements of the biometric values along with the actual outcome
for each patient, an inductive learning technique may construct a
data structure that, when given a new patient's biometrics,
predicts the new patient's five-year likelihood of another heart
attack.
[0032] Inductive learning methods that are well known to those of
skill in the art include techniques such as, for example, Bayesian
reasoning, memory based density estimation, parametric density
estimation, and neural networks. In neural networks, a training set
is used to learn weights attached to a data structure comprising a
network. See, e.g., Neural Networks for Pattern Recognition,
Christopher M. Bishop (Oxford University Press, 1995). A neural
network generally includes a set of nodes for inputting data
values, the input layer, and an output layer for producing a
predicted data value. Back-propagation neural networks also
comprise one or more hidden layers, between the input and output
layers. The input layer and each hidden layer are represented as
having outputs to each node of the next layer, either another
hidden layer or an output layer. After training is complete, the
neural network comprises a set of weights for each node in the
network. These weight values express a relationship between the
training data that is applied to input nodes and training data
applied to output nodes, as well as any hidden layer node values.
Using this trained relationship, the neural network can predict one
or more output values when new input values or parameters are fed
into corresponding nodes of the input layer. For each node, a
weight for the node is applied as a coefficient to a fixed equation
that expresses the trained relationship (which may vary in
different embodiments of the neural network) relating each of the
node's inputs to at least one output value. The output value of one
or more output layer nodes constitutes the output of the
network.
[0033] It is to be appreciated that while certain embodiments are
discussed with respect to a particular neural network topology,
other network topologies, such as, for example, having different
numbers of hidden layers, or differing numbers of input or output
nodes, can also be used. Further, while certain embodiments are
discussed with respect to using neural networks, other inductive
machine learning techniques or statistical analysis techniques,
such as, for example, those mentioned above, can be used.
[0034] FIG. 1 depicts one embodiment of a neural network 100
adapted for receiving wavefront aberration data having five input
nodes 110, a hidden layer with three nodes 120, and one output
layer with a single node 130 for producing a predicted visual
acuity value. Embodiments of the neural network 100 adapted to
receive a different number, or all, of the measured wavefront
aberrations generally include one input node per wavefront
aberration. Embodiments of the neural network 100 may be
implemented to execute on any general purpose or specialized
computer processor. It is also to be appreciated that the neural
network 100 may be implemented in software, which may itself be
produced using any computer language or environment, including
general-purpose languages such as C or FORTRAN. Preferably, the
neural network software is generated using a neural network
development tool such as, for example, Brainmaker Pro, produced by
California Scientific Software, Nevada City, Calif.
[0035] FIGS. 2 and 3 depict one embodiment of an apparatus for
measuring wavefront aberrations in a patient's eyes both to collect
the training data for the neural network 100 and for subsequent use
in conjunction with the neural network 100. The apparatus 10, as
described in more detail in the co-pending patent applications
incorporated herein, comprises a housing 12 that may be mounted on
a movable stand 14, for positioning the housing 12 in front of a
patient 15 who may sit in an examination chair 16. In one
embodiment, the apparatus 10 may also comprise a phoropter (not
pictured). In one embodiment depicted in FIG. 4, the housing 12
comprises a set of optics that is configured to measure wavefront
aberrations in the eye of patient 15. It is appreciated by those of
skill in the art that there exist different devices, systems and
techniques, such as a Schemer disk wavefront sensor, Shack-Hartmann
wavefront sensor, Fizeau interferometer, Twymann-Green
interferometer, holographic element wavefront sensor, binocular
wavefront sensor, etc., for measuring aberrations of the human eye,
or other optical systems. Any of these wavefront measuring devices,
systems and/or techniques may be utilized in accordance with the
present invention.
[0036] Using the collected training data, one embodiment of the
neural network 100 has been trained using a set of data relating
the measured wavefront aberrations of a patient to one measure of
visual acuity, e.g., the best corrected spectacle visual acuity
(BCSVA), of the patient. A learning set comprising patient
aberrations corresponding to the Zernike polynomials that are
generally known in the art as spherical aberration, coma, trefoil,
tetrafoil, and astigmatic tetrafoil as input values and BCSVA
values as the output values was constructed for 85 myopic eyes and
167 astigmatic eyes. The resulting trained neural network predicted
visual acuity in a random subset of the data set accurately in 85%
of the test cases. In other embodiments, improved accuracy may be
obtained in some circumstances by including a greater number of, or
all, measured high order aberrations during training and subsequent
testing. Furthermore, in other embodiments additional output
values, such as contrast sensitivity are used to train an
embodiment of the network 100 to obtain greater accuracy. It is
appreciated that other measures of visual acuity such as best
corrected visual acuity of a patient's eye using contacts,
intraocular lenses or Laser In Situ Keratomileuis (LASIK), for
example, may also be used as training data for the neural network
100.
[0037] FIG. 5 is a flow chart depicting one embodiment of a process
200 for predicting a standard prescription in a patient based on
objectively measured aberrations rather than a subjective phoropter
test. A standard prescription, the patient's subjective refraction
measure, comprises measurements of sphere, cylinder, and axis,
which correspond to low order aberrations. However, as discussed
above, subjective measures of visual acuity may also be affected in
a particular eye by the interaction of higher order aberrations and
the patient's neurological visual path. A training set of patients
is first measured to objectively determine the aberrations in their
eyes. The subjective refraction of each training patient is also
measured by using, for example, a phoropter and a Snellen Chart. In
one embodiment, the neural network 100 is trained using the
objectively measured aberrations as input and the subjective
measurements of sphere, cylinder, and axis as output. It is
appreciated that any number of additional data parameters that may
have some bearing on a patient's visual acuity can be used to train
and create a statistical model in accordance with the present
invention. For example, in one embodiment, factors or parameters
such as a patient's age, a patient's preferences for improving
vision (e.g., improved long distance vision, improved short
distance vision, improved contrast, improved resolution of fine
detail, improved depth of focus, etc.), and information concerning
the patient's prescription history, for example, may serve as
training data for building a statistical model. Additionally,
parameters or factors such as whether a patient experiences
headaches, blurred vision or "halos" at night, or other side
effects or symptoms (favorable or unfavorable), can be taken into
account during training of the statistical model. These types of
data parameters are generally referred to herein as "patient
personal profile parameters" or "profile parameters."
[0038] After a statistical model has been trained, the model may
then be used to predict a prescription of a new patient. Such a
process 200 begins at a step 210 where the aberrations of the
patient's eyes are objectively measured using, for example, the
measurement apparatus 10. Moving on to a step 220, the patient's
aberration measurements are input into the trained network 100
which is run to produce predicted sphere, cylinder, and axis
measurements based at least in part on objectively measured
aberration parameters. These predicted values of sphere, cylinder
and axis may then be used, for example, to provide a patient with a
standard prescription based at least in part on objective
measurements or parameters, to allow the patient and doctor to
compare different treatment options (e.g., corrective lenses,
contacts or surgery), or in the formulation of a LASIK scheme, for
example, to confirm a subjectively measured prescription before
executing the LASIK procedure.
[0039] It is further appreciated that the statistical model can
express relationships between various high order aberrations and
visual acuity such that the relative impact of certain high order
aberrations and/or combinations thereof can be predicted. For
example, certain high order aberrations or combinations of
aberrations may consistently reveal poor vision or superior vision
among patients having such aberrations. This would indicate that
such aberrations or combinations thereof may contribute to the poor
or superior vision. Given such a statistical model, it is easy to
see how a predicted visual correction prescription may be generated
that not only corrects for standard low order aberrations (sphere,
cylinder and axis) but also indicates which high order aberrations
(e.g., spherical aberration, coma, trefoil, tetrafoil, astigmatic
tetrafoil) and/or combinations thereof may be beneficial,
detrimental or neutral to a patient's subjectively perceived visual
acuity.
[0040] It has been further found that different high order
aberrations have different levels of importance for myopic,
hyperopic, presbyopic, and astigmatic eyes. Thus, in one embodiment
of process 200, a different neural network 100 may be used for eyes
in each of those categories of eyes. For example, in one
embodiment, a neural network 100 may be trained using primarily
patients with astigmatic eyes, resulting in a statistical model
that is customized especially for patients having astigmatic
aberrations.
[0041] As discussed above, it has been found that not all high
order aberrations have a negative impact on visual acuity. Thus,
eye correction techniques or devices that correct high order
aberrations such as eyeglasses or contact lenses, for example,
containing custom wavefront aberrators, as described in the
co-pending patent applications incorporated herein, may be further
improved by correcting only those aberrations that have a negative
affect on visual acuity, and leaving those that have a positive
affect.
[0042] FIG. 6 is a flow chart depicting another embodiment of a
process 300 for using a neural network or statistical model 100 to
help determine a course of corrective action for a patient's eyes.
In one embodiment, a neural network 100 is trained using some or
all objectively measured Zernike aberrations as the training
inputs. One or more measures of subjective visual performance, such
as visual acuity, and/or contrast sensitivity are measured for each
patient using traditional, low-order corrections, as with a
phoropter or eyeglasses. Additionally, visual performance is
subjectively measured using custom high order wavefront aberrators
correcting each of the high order aberrations (e.g., spherical
aberration, coma, trefoil, tetrafoil and astigmatic tetrafoil) in
the input set for each patient. The difference, or the improvement
measure between standard corrected vision and custom high-order
aberration corrected vision, which may be low or even negative in
the eye of a patient whose beneficial high order aberrations were
removed, is used as the output value for each patient in the
training set. Thus after training, the weights W.sub.1n of each of
the input layer nodes 110 indicate, for each aberration, the
relative importance, when the weight is positive, of correcting
that aberration for improving visual acuity. In cases where the
weight of the node 110 is negative, the corresponding aberration is
one that improves vision, i.e. correcting the aberrations does not
improve vision, but rather worsens visual acuity.
[0043] Thus, returning to process 300, beginning at a step 310, the
aberrations of a patient's eyes are measured. Moving on to a step
320, each aberration is analyzed with reference to the weights of
the input layer nodes 110 of the neural network 100 trained with
the aberrations as inputs and visual performance improvement as
outputs and assigned an importance based on the relative magnitude
and sign of the weight. Next at a step 330, a table of measured
aberrations, such as Zernike polynomials in a table according to
the order of the polynomial, from, for example, second order to
sixth order, is presented with each aberration color coded based on
the relative importance of adjusting that aberration as determined
in the analysis of the step 320. FIG. 7 illustrates one example of
a color-coded table and, additionally, a color-coded graphic image
that illustrates high-order aberrations based on their relative
importance or status (e.g., beneficial or detrimental to visual
acuity), in accordance with one embodiment of the invention. Note
that adjusting the aberration may include minimizing the magnitude
of the aberration, increasing the magnitude of the aberration, or
introducing the aberration. In one embodiment, aberrations that
have small amplitudes, or that do not correlate to an improvement
in vision are keyed in green, for example, aberrations having
smaller, but still significant, amplitudes are keyed in yellow, for
example, and aberrations having a large amplitude are keyed in red,
for example, to indicate the importance of adjusting those
aberrations. Note that adjusting an aberration may include
decreasing the magnitude of the aberration or increasing the
magnitude of a beneficial aberration. Moreover, some or all
aberrations, whether present in the patient's eyes or not, may also
be keyed to a color, for example, to indicate whether introducing
the aberration may improve visual acuity. Moving onto a step 340,
the eye care professional and the patient may use this information
to develop a plan for correcting the patient's vision, and allow
the eye care professional to advise the client more accurately on
the value of correcting the higher order aberrations using, for
example, spectacles, contacts, intraocular lenses, or surgery
(e.g., LASIK). As indicated above, this correction plan may include
introducing or increasing aberrations that have a beneficial impact
on visual acuity.
[0044] FIG. 8 depicts an embodiment of a process 400 for predicting
what level of improvement in visual acuity may be obtained by a
patient by choosing eyeglasses or contact lenses, for example,
containing custom wavefront aberrators. An embodiment of a neural
network is constructed as discussed for use in process 300 above.
Starting at a step 410, the aberrations of a patient's eyes are
measured. Moving to a step 420, a predicted improvement in visual
acuity from using custom high order wavefront corrected glasses
versus standard eyeglasses is calculated using the neural network
100. Custom high order wavefront corrected glasses include a custom
wavefront aberrator that corrects the patient's vision by adjusting
wavefront aberrations in the patient's visual path. This adjusting
may include any combination of the following adjustments:
minimizing the magnitude of existing aberrations; increasing the
magnitude of existing, beneficial, aberrations; or introducing
beneficial aberrations. Next at a step 430, this improvement is
presented. In one embodiment, the improvement is depicted
graphically in terms of the additional number of Snellen chart
letters that high order corrected glasses would allow the patient
to read versus standard eyeglasses. In another embodiment, a
graphical depiction may be presented using a computerized display
to simulate the relative blurring of the letters of the chart as
shown in exemplary FIG. 9. The letters in the left chart are more
blurred to indicate poorer vision resulting from standard
correction or no correction. The right chart is clearer and more
focused to illustrate the improved vision possible with high order
aberration correction. It is understood that such graphics may
provide a best approximation of a patient's visual acuity based on
predicted improvement using standard correction and high-order
correction, respectively. For example, if a patient's predicted
improvement using standard correction is 20/25 vision and using
high-order correction is 20/15 vision, for example, respective
graphic charts selected (automatically or manually) to represent
corresponding blurring may be presented to the patient.
[0045] In addition to being used in reference to the selection and
design of custom wave front aberrator eyeglasses, embodiments of
neural network 100 may also be advantageously used to predict the
results of more permanent corrections, including surgical
procedures and techniques such as intraocular lens insertion and
corneal sculpting such as Radial Keratotomy, Astigmatic Keratotomy,
Automated Lamellar Keratoplasty, Photo Refractive Keratectomy, or
LASIK. The predictive ability of the trained neural network 100 is
of even greater value when used in conjunction with the planning of
these procedures than with custom wavefront aberrator eyeglasses
due to the permanent nature of the changes that are made.
[0046] Thus, an embodiment of a process similar to the process 400
includes generating an embodiment of the neural network 100 using
training data that uses improvements resulting from high-order
correction in LASIK patients when compared to standard correction
for those patients, measured prior to performing a LASIK procedure.
Thus, a statistical model can be trained using objectively measured
high order aberrations as training input values for the input layer
nodes and the subjective improvement measurement values as training
output values for the output node of the statistical model. After
the statistical model is trained using training data, it may then
be used to determine or predict the relative importance and status
(e.g., beneficial, detrimental, neutral) of various high order
aberrations found in LASIK patient candidates. After identifying
the relative importance of the high order aberrations in the
patient, the ablation pattern of the LASIK laser can be changed to
optimize the patient's visual performance as a result of LASIK
surgery. Further, in one embodiment of a process paralleling the
process 400, the patient is presented with a visual depiction of
the relative improvement (or worsening) of vision that is predicted
to result from LASIK that is similar to the depictions illustrated
in FIG. 9, for example. This process may be combined with steps of
process 400 to provide the patient with a comparison of the
predicted outcome of regular eyeglasses, LASIK (with standard or
high-order correction), and of using eyeglasses incorporating
custom aberrators. A process similar to process 400 may be
performed for patients using or considering contact lenses or
intraocular lenses. By using patients with corrective contact
lenses or intraocular lenses to obtain training data, statistical
models may be created to predict relative results between the
different types of treatments (e.g., intraocular lenses v. regular
eyeglasses v. contacts) and the different types of corrections
(e.g., standard correction v. high-order aberration
correction).
[0047] It is to be appreciated that in embodiments of processes of
using trained neural networks, such as described herein, the
computational steps comprising using the neural network 100 on a
new patient, and presenting the result on a display, may be
performed directly on the measurement apparatus 10 using, for
example, an attached video display to present the result, or the
measured aberrations may be input into another computerized system
for immediate use as described herein, or for storage and later use
and presentation by a different display system. Furthermore, it
should be realized that, for each process disclosed herein,
additional steps may be added, others removed, steps merged, or the
order of the steps rearranged without departing from the scope of
the invention.
[0048] While the above detailed description has shown, described,
and pointed out novel features of the invention as applied to
various embodiments, it will be understood that various omissions,
substitutions, and changes in the form and details of the device or
process illustrated may be made by those skilled in the art without
departing from the spirit of the invention. As will be recognized,
the present invention may be embodied within a form that does not
provide all of the features and benefits set forth herein, as some
features may be used or practiced separately from others.
* * * * *