LABORATORY SCIENCE

Scheimpflug image–processing method for

accurate measurement of ocular surfaces
Ji C. He, PhD, Yanwen Fang, MD

PURPOSE: To develop a method to process Scheimpflug images to estimate higher-order variations

in ocular surfaces.
SETTING: New England College of Optometry, Boston, Massachusetts, USA.
METHODS: Zernike polynomial functions were used rather than lower-order surface functions to fit
the ocular surfaces to evaluate higher-order variations. A related ray-tracing method was developed
to correct optical distortion of the Scheimpflug image.
RESULTS: The new method to process Scheimpflug images showed higher-order variations in the
posterior corneal surface. In a simulation, ignoring higher-order variations in the anterior corneal
surface, as in previous studies, produced an error as high as 202% for some estimated Zernike
coefficients of the posterior corneal surface.
CONCLUSION: The results indicate that it is necessary to correct refractive distortion in Scheimp-
flug images using higher-order surface functions and that the method developed in this study can
be used to process scheimpflug images for measurement of ocular wavefront aberrations.
Financial Disclosure: Neither author has a financial or proprietary interest in any material or
method mentioned.
J Cataract Refract Surg 2010; 36:838–842 Q 2010 ASCRS and ESCRS

Accurate determination of surface geometry of the Scheimpflug images are not adequate for estimating
cornea and the lens in the human eye is important wavefront aberrations of the eye.
for exploring the sources of ocular wavefront aberra- The Scheimpflug image is a record of the sagittal
tion. Photography based on the Scheimpflug principle section of the anterior segment of the eye that is illumi-
has long been used for in vivo assessment of the ante- nated by a thin-slit parallel light beam. The image is
rior segment of the eye. Recently, Scheimpflug imag- formed by the Scheimpflug camera with the optical
ing systems with high-resolution cameras and new axis oriented 45 degrees to the optical axis of the
image-processing programs have been used to study illuminating slit beam and its image plane at a right
accommodation and presbyopia and in clinical angle to the object plane of the illuminated eye section.
practice to diagnose eye disease.1 Although current In the Scheimpflug image, structures of the cornea and
Scheimpflug imaging systems have been used success- lens are clearly visible; however, the size and shape are
fully in eye research,2–12 the algorithms for processing distorted because of errors inherent in the Scheimpflug
principle.2–4 The first error is the distortion caused by
the specific relationship of geometric projection
between the object and image planes defined in the
Submitted: September 28, 2009.
Scheimpflug technique. The projection distortion,
Final revision submitted: November 24, 2009.
Accepted: November 30, 2009. however, is mathematically predictable if design
parameters of the system are known2; therefore, it is
New England College of Optometry (He), Boston, Massachusetts, adequately corrected in most current Scheimpflug
USA, and Shanghai Eye & ENT Hospital (Fang), Fudan University, systems.
Shanghai, China. Another error is the distortion caused by the eye it-
Corresponding author: Ji C. He, PhD, New England College of self because optically, the outer structure of the eye
Optometry, 424 Beacon Street, Boston, Massachusetts 02115, (eg, the cornea) is an additional optical component in
USA. E-mail: hej@[Link]. the Scheimpflug optical system during image

838 Q 2010 ASCRS and ESCRS 0886-3350/10/$dsee front matter

Published by Elsevier Inc. doi:10.1016/[Link].2009.11.022
 LABORATORY SCIENCE: METHOD FOR SCHEIMPFLUG IMAGE PROCESSING 839

formation for the inner structure (eg, the lens). How- the anterior corneal surface, this study used Zernike polyno-
ever, the section of the corneal surface participating mial functions to fit the anterior corneal surface, as per-
formed in previous studies of wavefront analysis of
in Scheimpflug image formation is not rotationally
corneal aberrations.15,16 The fitting can be achieved by using
symmetrical to the optical axis of the Scheimpflug the least-squares matrix inversion method with the aid of the
camera due to the tilt between the optical axes of Gram-Schmidt orthogonalization procedure.16 After the
the camera and the eye. All surfaces of the cornea fitting, the anterior corneal surface is mathematically
and lens, except the anterior corneal surface, suffer described as
this type of optical or refractive distortion in the X
N

Scheimpflug image. To correct the optical distortion, Zac ðx; yÞ Z Cj Zj ðx; yÞ (1)
jZ1
2 procedures are required: 3-dimensional (3-D) recon-
struction of the outer surface and ray tracing of each where Zj is the j term of Zernike polynomial functions of N
point in the inner surface.2 However, in current terms under a unit circle area, which could be those recom-
Scheimpflug image processing, the surface function mended by the Optical Society of America,17 and Cj is the
corresponding Zernike coefficient. According to previous
used to reconstruct the anterior corneal surface in studies,15 the term number N up to 35th (7th-order Zernike
most systems is a 2nd-order function.2,3,5 This does functions) would be enough to precisely describe normal
not account for the higher-order local variation in the corneal surface.
surface shape, which mainly contributes to higher- Given that the anterior corneal surface is accurately deter-
order wavefront aberrations of the cornea. mined, the first step in ray tracing a point in the inner surface
section of the Scheimpflug image is to find the intercept of
In this study, a method was developed to process the ray from the image point to the nodal point of the
the Scheimpflug image for capturing the higher- Scheimpflug camera. Figure 1 shows a reconstructed 3-D an-
order variation of the ocular surfaces using Zernike terior corneal surface with a Scheimpflug anterior corneal
polynomial functions. A new ray-tracing procedure section and a Scheimpflug posterior corneal section. Both
required for the higher-order functions was also sections are assumed to be corrected for the projection distor-
tion at this stage. At the posterior corneal section, the
developed and is presented here. Scheimpflug camera views point P at its nodal point through
the anterior corneal surface. Because of refraction at the ante-
MATERIALS AND METHODS rior corneal surface, the apparent point P is deviated from its
true point in the eye, which has to be traced back. The ray
In a Scheimpflug imaging system capable of assessing 3-D from nodal point (xn, yn, zn) to image point (xp, yp, zp) is
surfaces of the cornea and lens, a series of Scheimpflug expressed by vector
images, each at a different sagittal meridian, are usually ob-      
tained by repeating image shutting at different meridians S Z xp  xn i þ yp  yn j þ zp  zn k (2)
while rotationally turning an illuminating slit beam and
camera simultaneously. Image processing of the Scheimp- and its unit vector s is
flug images starts from registering the 4 surface boundaries s Z s=jsj Z sx i þ sy j þ sz k (3)
of the cornea and lens on each image and subsequently
deriving quantitative parametric descriptions of the 4 curves. where
This can be done by using the conventional image- h 2  2  2 i1=2
processing method suggested in previous studies.2–4 jSj Z xp  xn þ yp  yn þ zp  zn
Because of the projection distortion of the Scheimpflug tech-
nique, none of the 4 curves reflects the true size of the corre-
sponding surface section. Correction of the projection To trace the ray back into the eye, it is necessary to find the
distortion is usually performed at this stage using the intercept C (xc, yc, zc) of the ray S at the anterior corneal
method in Appendix 2 in a study by Richards et al.2 This pro- surface Zac. However, it is impossible to derive the intercept
cedure requires the use of the design parameters of the analytically by solving equations of the ray and the corneal
Scheimpflug system. After the projection distortion is surface, as suggested by Richards et al.,2 because the corneal
corrected, the true size of the anterior corneal section at the surface is not described by a low-order function now but
meridian under correction is recovered because the only rather by the higher-order Zernike polynomials. In this
anterior corneal surface is subject to projection distortion. study, a method of computational numerical iteration was
The true size of the posterior cornea and lens surfaces, used to find the intercept by approaching the difference
however, is still not reflected because their images are also between the points in the ray and the corneal surface to zero.
optically distorted; this must be corrected by further image Given that the intercept point at the corneal surface was
processing with a ray-tracing procedure. Ray tracing re- derived, the normal at the intercept can be found with
quires 3-D reconstruction of the anterior corneal surface
from the series of corrected anterior corneal images; it is usu- n Z vFðx; y; zÞ=vx i D vFðx; y; zÞ=vy jD vFðx; y; zÞ=vz k (4)
ally derived by fitting the corrected anterior corneal curves to where F(x,y,z) Z 0 is the general description of the anterior
surface function with least-square procedure. Most previous corneal surface Zac, and thus
studies used a 2nd-order elliptic paraboloid as the surface
PN
function2; however, this does not accurately describe the vFðx; y; zÞ=vx Z cj vZj ðx; yÞ=vx
true anterior corneal surface because the real corneal surface PjNZ 1
vFðx; y; zÞ=vy Z j Z 1 cj vZj ðx; yÞ=vx (5)
has higher-order local variation, as seen in recent studies of
corneal wavefront aberrations.13–15 To accurately describe vFðx; y; zÞ=vz Z  1

J CATARACT REFRACT SURG - VOL 36, MAY 2010

840 LABORATORY SCIENCE: METHOD FOR SCHEIMPFLUG IMAGE PROCESSING

Table 1. Zernike coefficients of the anterior and posterior cor-

neal surfaces over a 6.0 mm diameter central area.

Value (mm)

j Term/Coefficient Anterior Cornea Posterior Cornea

j Z 0/Z(0,0) 295.76317 632.69041

j Z 1/Z(1,1) 0.0 0.0
j Z 2/Z(1,1) 0.0 0.0
j Z 3/Z(2,2) 0.0 0.6
j Z 4/Z(2,0) 172.59732 187.92693
j Z 5/Z(2,2) 2.0 1.2
j Z 6/Z(3,3) 0.0 0.3
j Z 7/Z(3,1) 0.0 0.5
j Z 8/Z(3,1) 1.0 0.8
j Z 9/Z(3,3) 0.0 0.2
j Z 10/Z(4,4) 0.0 0.1
j Z 11/Z(4,2) 0.0 0.05
Figure 1. Schematic of Scheimpflug image ray tracing (ACS Z sec- j Z 12/Z(4,0) 1.45174 0.95409
tion of anterior corneal surface; C Z intercept of the ray and the cor- j Z 13/Z(4,2) 0.0 0.2
neal surface; N Z nodal point of Scheimpflug camera; Nc Z normal
j Z 14/Z(4,4) 0.0 0.0
of the surface at C; P Z image point at the posterior corneal section;
j Z 15/Z(5,5) 0.0 0.0
PCS Z Scheimpflug image section of posterior corneal surface;
P0 Z true point at the posterior corneal surface). j Z 16/Z(5,3) 0.0 0.0
j Z 17/Z(5,1) 0.0 0.0
j Z 18/Z(5,1) 0.0 0.0
describes the components along the respective axes. j Z 19/Z(5,3) 0.0 0.0
Substituting the intercept point (xc, yc, zc) to equations 4 j Z 20/Z(5,5) 0.0 0.0
and 5, the normal is found.
With the incident ray vector, the intercept point at the
anterior corneal surface, and the normal at the intercept ignoring the higher-order and asymmetric variations. This
point, the refracted ray vector after it passes the anterior can be achieved by setting the Zernike coefficients of j Z 5
corneal surface can be derived applying Snell’s law.18 and j Z 8 to zero. In fact, the rest of the Zernike terms in
Once the ray vector inside the eye is determined, the true Table 1 describe a conic anterior cornea with a radius of
point P0 on the posterior corneal section corresponding to 7.77 mm and a corneal asphericity of 0.18.
the Scheimpflug image, point P can be derived by finding
the intercept of the refracted ray on the illuminated x–z
plane (y Z 0 in Figure 1). This corrects the refractive distor-
RESULTS
tion of the posterior corneal section in the Scheimpflug
image. The whole true posterior corneal shape can then In the simulation, image-processing method (12 sec-
be accurately recovered by fitting the corrected posterior tions in steps of 15 degrees were simulated) developed
corneal sections with Zernike polynomial functions, as de- in this study was used to correct refractive distortion of
scribed above. For correcting refractive distortion of the
lens, the same procedure used to correct the posterior cor- the posterior corneal surface. Figure 2 shows the Zer-
neal surface distortion must be performed again by ray nike coefficients up to 5th order. The distribution of
tracing back the anterior corneal surfaces and the posterior the Zernike coefficients in Figure 2 therefore describes
corneal surface. the true posterior corneal surface because both projec-
A simulation of the correction of refractive distortion for tion and refractive distortions are corrected.
a posterior corneal surface was performed by implement-
ing the image-processing method described above to an as- Figure 3 shows the simulation-derived posterior
sumed cornea. Table 1 shows the Zernike coefficients of the corneal surface using a conic anterior corneal surface.
assumed anterior and posterior corneal images, which were Figure 4 shows the difference in the Zernike coeffi-
corrected for projection distortion using the image- cients of the posterior corneal surface between the 2
processing method. In the anterior corneal surface, 2 asym- types of anterior corneal surfaces; it can be seen that
metric terms, j Z 5 and j Z 8, were given with the amounts
often observed in a normal eyes for a 6.0 mm area; other the use of the conic function produced significant dif-
terms were set at zero. The Zernike coefficients for the ferences in Zernike coefficients. The difference in the
anterior cornea are assumed to accurately describe the posterior corneal surface was introduced not only at
anterior corneal surface at this point because the only dis- two Zernike terms (j Z 5 and j Z 8), which were ig-
tortion has been corrected. nored from the assumed anterior corneal surface, but
To test whether there is a difference in the corrected
posterior corneal surface between the current method and also at other terms (eg, j Z 6, j Z 9, j Z 14). The value
methods used in previous studies, a second-order conic func- of j Z 5 in the posterior cornea increased in amplitude
tion was used to model the anterior corneal surface by from 2.14 mm to 3.23 mm (52% change), and j Z 8

J CATARACT REFRACT SURG - VOL 36, MAY 2010

 LABORATORY SCIENCE: METHOD FOR SCHEIMPFLUG IMAGE PROCESSING 841

Figure 2. Distribution of derived Zernike coefficients of the posterior Figure 4. Distribution of derived Zernike coefficients of the posterior
corneal surface processed for Scheimpflug image with higher-order corneal surface processed for Scheimpflug image with higher-order
variation in the anterior corneal surface taken into account. variation in the anterior corneal surface ignored.

increased from 1.46 mm to 2.01 mm (34% increase) DISCUSSION

when the two Zernike terms in the anterior corneal Scheimpflug imaging is a useful tool for in vivo assess-
surface were ignored. Meanwhile, the values of j Z 6 ing of the anterior segment of the eye; however, its use-
and j Z 9 changed from 0.48 mm to 0.89 mm (87% fulness in ophthalmology research or in clinical
change) and from 0.35 mm to 0.36 mm (202% change), applications has not been fully explored. Scheimpflug
respectively. imaging has an advantage over the traditional Placido
Therefore, most Zernike terms of the posterior cor- ring technique for evaluating wavefront aberrations in
neal surface derived by ignoring higher-order varia- the eye because it tests both the anterior cornea and
tion in the anterior corneal surface were not true posterior cornea as well as the surfaces of the lens. It
because the real anterior cornea had higher-order var- is therefore capable of measuring wavefront aberra-
iation and the variation was not restricted to the two tions in all ocular surfaces instead of in the anterior
terms simulated here but was also distributed on corneal surface only, as in conventional corneal topog-
many other terms. The large error in the derived shape raphy. This is important for understanding the sources
of the posterior corneal surface means that there of the wavefront aberrations in the whole eye and for
would be significant error in further estimates of planning wavefront-guided treatment in clinical prac-
wavefront aberrations in the posterior corneal surface tice, such as for phakic intraocular lens, refractive, and
if the Scheimpflug-imaging technique were used to cataract surgery.
measure wavefront aberrations in the posterior cornea However, the application of the Scheimpflug-
and the lens. imaging system relies on the accuracy of its measure-
ments. Several factors limit the measurement accuracy
of current Scheimpflug systems. For example, due to
eye movement, the instrument’s axis during measure-
ment is not always maintained in the same position
with respect to the eye. The corneal surface shift result-
ing from the eye movement causes the peak to deviate
from the corneal vertex in some Scheimpflug images;
therefore, the shape of the reconstructed anterior
corneal surface is not the true shape. This misalignment
will introduce measurement error unless an eye-
tracking system is used. In addition, the image resolu-
tion of the surface curves at each section decreases
with the distance from the center because of projection
distortion. Lower resolution will produce lower
accuracy, thus introducing measurement error.
Figure 3. Difference in derived Zernike coefficients of the posterior
corneal surface between the anterior corneal surface with and with- Although techniques for solving these issues must be
out higher-order variations ignored during Scheimpflug image improved, higher-order variation in the ocular surfaces
processing. should not be ignored. Biologic ocular surfaces, like the

J CATARACT REFRACT SURG - VOL 36, MAY 2010

842 LABORATORY SCIENCE: METHOD FOR SCHEIMPFLUG IMAGE PROCESSING

anterior corneal surface, cannot be accurately described 6. Dubbelman M, van der Heijde GL. The shape of the aging human
by low-order geometric surfaces. Recent studies of cor- lens: curvature, equivalent refractive index and the lens paradox.
Vision Res 2001; 41:1867–1877
neal aberrations13–15 show that the shape of corneal sur- 7. Dubbelman M, Weeber HA, van der Heijde RGL, Völker-
faces varies significantly and that higher-order corneal Dieben HJ. Radius and asphericity of the posterior corneal
variations change substantially from person to person, surface determined by corrected Scheimpflug photography.
and even from left eye to right eye of the same person. Acta Ophthalmol Scand 2002; 80:379–383
However, in current methods of Scheimpflug image 8. Dubbelman M, van der Heijde GL, Weeber HA. Change in shape
of the aging human crystalline lens with accommodation. Vision
process, the surface for reconstructing the corneal sur- Res 2005; 45:117–132
face is mathematically described with 2nd-order func- 9. Dubbelman M, Sicam VADP, van der Heijde GL. The shape of
tions.2,3,5 For example, Richards et al.2 used ellipses, the anterior and posterior surface of the aging human cornea.
Kampfer et al.3 used circles, and Dubbelman et al.5 Vision Res 2006; 46:993–1001
used conics. These studies ignored higher-order local 10. Sicam VADP, Dubbelman M, van der Heijde RGL. Spherical ab-
erration of the anterior and posterior surfaces of the human cor-
variation in surface shape, and thus introduced errors nea. J Opt Soc Am A Opt Image Sci Vis 2006; 23:544–549
when estimating the inner surface shapes or the wave- 11. Dubbelman M, Sicam VADP, van der Heijde RGL. The contri-
front aberrations in the inner structures. The error is di- bution of the posterior surface to the coma aberration of the hu-
rectly dependent on the characteristics and amount of man cornea. J Vis 2007; 7(7):. 10,1–8. Available at [Link]
the higher-order variations in the outer surface as [Link]/7/7/10/[Link]. Ac-
cessed January 20, 2010
well as on how inner is the surface located. The more in- 12. Hermans E, Dubbelman M, van der Heijde R, Heethaar R. The
ner the surface (eg, the posterior lens surface), the larger shape of the human lens nucleus with accommodation. J Vis
the estimated error was introduced because of the error 2007; 7(10):. 16,1–10. Available at [Link]
accumulation. org/7/10/16/[Link]. Accessed January
Our results show that the method we developed in 20, 2010
13. Artal P, Guirao A, Berrio E, Williams DR. Compensation of cor-
this study is capable of detecting higher-order varia- neal aberrations by the internal optics of the human eye. J Vis
tions in the surfaces of the cornea and the lens. The 2001; 1(1):. 1–8. Available at [Link]
use of higher-order Zernike polynomial functions 1/1/[Link]. Accessed January 20, 2010
and the new ray-tracing technique will, therefore, 14. Salmon TO, Thibos LN. Videokeratoscope–line-of-sight mis-
improve the accuracy of the Scheimpflug-imaging alignment and its effect on measurements of corneal and inter-
nal ocular aberrations. J Opt Soc Am A Opt Image Sci Vis
system and thus extend its applications in ophthalmol- 2002; 19:657–669; errata 2003; 20:195
ogy research and clinical practice. 15. He JC, Gwiazda J, Thorn F, Held R. Wave-front aberrations in
the anterior corneal surface and the whole eye. J Opt Soc Am
A Opt Image Sci Vis 2003; 20:1155–1163
REFERENCES
16. Wang JY, Silva DE. Wavefront interpretation with Zernike poly-
1. Koretz JF, Strenk SA, Strenk LM, Semmlow JL. Scheimpflug nomials. Appl Opt 1980; 19:1510–1518
and high-resolution magnetic resonance imaging of the anterior 17. Thibos LN, Applegate RA, Schwiegerling JT, Webb R. Stan-
segment: a comparative study. J Opt Soc Am A Opt Image Sci dards for reporting the optical aberrations of eye; VSIA Stan-
Vis 2004; 21:346–354 dards Taskforce Members. J Refract Surg 2002; 18:S652–S660
2. Richards DW, Russell SR, Anderson DR. A method for improved 18. Born M, Wolf E. Principle of Optics; Electromagnetic Theory of
biometry of the anterior chamber with a Scheimpflug technique. Propagation, Interference, and Diffraction of Light 5th ed.. New
Invest Ophthalmol Vis Sci 1988; 29:1826–1835. Available at York, NY, Pergamon, 1975;; 190–196
[Link] Accessed January
20, 2010
3. Kampfer T, Wegener A, Dragomirescu V, Hockwin O. Improved
biometry of the anterior eye segment. Ophthalmic Res 1989; 21:;
First author:
239–24
4. Lapuerta P, Schein SJ. A four-surface schematic eye of ma-
Ji C. He, PhD
caque monkey obtained by an optical method. Vision Res New England College of Optometry,
1995; 35:2245–2254 Boston, Massachusetts, USA
5. Dubbelman M, van der Heijde GL, Weeber HA. The thickness of
the aging human lens obtained from corrected Scheimpflug im-
ages. Optom Vis Sci 2001; 78:411–416

J CATARACT REFRACT SURG - VOL 36, MAY 2010