be generally the case and is certainly not the case for a
keratoconic cornea or irregular astigmatic cornea.
e Keratometer measurements become increasingly difficult
when surface irregularities distort the mires.
e The point-by-point nature of the measurement process
makes it impractical to compile a complete mapping of
corneal topography.
The keratometer fulfils its primary role, that of providing a
radius of curvature or, on the basis of an assumed corneal
refractive index, a value of the corneal refractive power, for
routine clinical assessment of normal human corneas. Its
application beyond this role is very limited.
Photokeratoscopes are also, in many respects, inadequate:
* The central and peripheral regions of the cornea cannot
normally be measured because of an absence of reflected
mires, corneal curvature, and obstruction caused by the
nose and brow (Bores 1991, Gormley et al 1988, Klyce
and Wilson 1989a, Mammone et al 1990, Warnicki et al
1988). Accurate measurement of the central reflected mire
is critical to the mathematical derivation of corneal shape
(Missotten 1994).
e The location of reflected mires is highly sensitive to
corneal distortion. This is because the position of the
reflected mire will be a function of both slope and
displacement. On normal corneas, this sensitivity can be
advantageous. | However when there are very rapid
changes in topography the image soon becomes too
distorted to measure reliably (Klyce and Wilson 1989b,
Friedlander et al 1991).
e Visual interpretation of a keratogram can only provide
approximate data; clinically significant changes are
commonly not detectable (Friedlander et a/ 1991, Wilson
and Klyce 1991).
e Several commercial instruments use dithering techniques
to fill gaps in the acquired data. The assumptions that are
made, typically that the cornea is spherical over any areas
that cannot be imaged, are unwarranted (eg Klyce and
Wilson 1989a, Bores 1991).
e Photokeratoscopic data are highly sensitive to
misalignment with the corneal axis (Heath et al 1991) and
misjudgment of the focussing position and therefore the
distance to the cornea (Saarloos and Constable 1991,
Missotten 1994).
e The instruments rely on corneal reflectivity, in turn
determined by the condition of the unstable tear film that
coats the surface of a healthy cornea and the surface
roughness of the corneal epithelium (Duke-Elder 1970).
For a healthy cornea, reflectivity is only about 4% at the
corneal centre and decreases to near 2% at the periphery
(Clark 1973a). Abnormal corneas with low reflectivity
cannot be measured (Warnicki et al 1988).
Photokeratoscopes cannot be used for assessment of
corneal topography during surgery because of the
inevitable non-reflectivity of the corneal surface.
e Exact topographic data cannot be calculated from a
photokeratoscope image. The curve fitting techniques
applied are of limited value for several reasons but
primarily because of the non-uniqueness of the corneal
surface for a given image (eg Wise et al 1986, Mammone
et al 1990), the asphericity of the corneal surface, and
because they cannot model abrupt changes that may occur,
for example, at the edge of a photorefractive keratectomy
(Missotten 1994).
3. THE KERATOCON
The limited quantitative information provided by the
keratometer and photokeratoscope together with an
increasing requirement in modern ophthalmology for
accurate topographic mapping of the entire anterior surface
of both normal and abnormal corneas made it apparent that a
new instrument was required. It follows from the discussions
above that such an instrument should:
i. be capable of measuring the entire cornea,
ii. not rely on corneal reflectance,
iii. not rely on a precorneal tear film,
iv. measure corneal topography with sufficient density and
accuracy to provide reliable and clinically interpretable
representations of corneal topography and corneal
power, and
v. not require that assumptions be made about the
geometry of the cornea.
A schematic of the instrument appears below. The cameras
used were two 35mm motor-driven Leica R4's fitted with
Leica 200mm focal length lenses, positioned at
approximately 25° convergence. Because the cameras were
non-metric it was necessary to design the prototype so that
on-the-job calibrations could be performed. This required
photocontrol very close to the cornea — within the limited
depth of field of both cameras and sufficiently close to be
photogrammetrically reliable. There are two operational
considerations. Firstly, the clinician must have access to the
eye; secondly, patients cannot be expected to tolerate close
proximity to any part of the instrument. To overcome these
problems, photocontrol was reflected from a beamsplitter
into the optical axis of the system, so that it appeared to
surround the cornea in each photograph. A similar technique
has been used by Scott (1981, 1987) for his reflex measuring
instruments and reported by Mikhail (1974). The control
points were a pattern of 30 marks burnt through a thin
opaque surface deposited onto the outside of an accurately
polished glass sphere. The sphere is flash illuminated from
behind at the moment that the cameras are fired. The control
points were coordinated to better than +5um. The cameras,
photocontrol and beamsplitter were mounted onto the
platform of a Sun PKS1000 photokeratoscope. This
provided a mechanism for controlling the height of the
instrument and the position of the patient's head relative to
the cameras.
beamsplitter
ri
"4A ES |
FI RSS A
0 IRR
POL KXX]
bod 12292 li
7777 gon
pivot
Figure 1. Schematic of the Keratocon.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996
A variety
including
1961, B:
particula)
as the :
Warnicki
method a
contact 1
prototype
1993 per
indicatioi
simpler t
onto it.
redirect
where m
cornea to
bundle fi
control sj
Figur
in the
The calib
using bot
and Kara
Karara 1¢
routine d.
space con
monosco]
prior to p:
It was in
accuracy
effects of
base distz
of object
constraine
of field cc
patient's |
found to
and conve
Formulae
Abdel-Az
precision
geometrie
l obje
the ii
