Graefe’s Arch Clin Exp Ophthalmol (2000)

238:765–773 © Springer-Verlag 2000 C L I N I C A L I N V E S T I G AT I O N

Wolfgang Haigis Comparison of immersion ultrasound biometry

Barbara Lege
Nicole Miller and partial coherence interferometry
Britta Schneider
for intraocular lens calculation according
to Haigis

Received: 8 March 2000 Abstract Background: The preci- ferent conversion formulas were
Revised: 22 May 2000 sion of intraocular lens (IOL) calcu- used, among them the relation
Accepted: 8 June 2000 lation is essentially determined by which is built into the Zeiss IOL-
the accuracy of the measurement of Master. IOL calculation was carried
axial length. In addition to classical out according to Haigis with and
ultrasound biometry, partial coher- without optimization of constants.
ence interferometry serves as a new Results: On the basis of ultrasound
optical method for axial length deter- immersion data from our Grieshaber
mination. A functional prototype Biometric System (GBS), postopera-
from Carl Zeiss Jena implementing tive refraction after implantation of
this principle was compared with im- a Rayner IOL type 755 U was pre-
mersion ultrasound biometry in our dicted correctly within ±1 D in
laboratory. Patients and methods: In 85.7% and within ±2 D in 99% of all
108 patients attending the biometry cases. An analogous result was
Presented in part at the Annual Meeting laboratory for planning of cataract achieved with optical axial length
of the German Ophthalmological Society surgery, axial lengths were addition- data after suitable transformation of
(DOG), Berlin, 1999 ally measured optically. Whereas optical path lengths into geometrical
surgical decisions were based on ul- distances. Conclusions: Partial co-
W. Haigis (✉) · B. Lege · N. Miller
B. Schneider trasound data, we used postoperative herence interferometry is a non-
University Eye Clinic, refraction measurements to calculate contact, user- and patient-friendly
Josef-Schneider-Strasse 11, retrospectively what results would method for axial length determina-
97080 Würzburg, Germany have been obtained if optical axial tion and IOL planning with an accu-
e-mail:
[email protected] length data had been used for IOL racy comparable to that of high-pre-
Tel.: +49-931-2015640 calculation. For the translation of op- cision immersion ultrasound.
Fax: +49-931-2012454 tical to geometrical lengths, five dif-

Introduction The first in vivo application of PCI measuring the ax-

ial length of a human eye was reported by Fercher et al.
Some 200 years ago, the Italian scholar Lazzaro Spallan- [5, 6] and dates back to 1986. Since then, the technique
zani found out that bats were using ultrasound to “see has been well accepted by the ophthalmic community in
with their ears”. Today, ultrasound biometry is a well- its tomographic form (optical coherence tomography,
established and precise method for the measurement of oc- OCT), whereas the distance-measuring capability of PCI
ular distances, especially determination of the axial length is just about to grow up and leave the research laboratory
and its ocular segments for the calculation of the required (e.g. [7, 14, 17, 19, 20, 21]).
power of intraocular lenses (IOL). Recently, with laser in- We had the opportunity to test [14, 21] a pre-production
terference biometry (LIB), a new optical method based on function model (Zeiss ALM) of such an optical measuring
the principles of partial coherence interferometry (PCI) has device manufactured by Carl Zeiss Jena in our laboratory
come on the scene as a possible alternative to ultrasound. and help in the development of this instrument into a clini-
766

cally applicable biometry system, which appeared on the Biometry

market in autumn 1999 under the name of Zeiss IOLMas-
Ultrasound biometry was performed in immersion technique with
ter. the Grieshaber Biometric System (GBS). This high-precision in-
In a first stage, we had set up the correlation [14] strument uses four digital 40-MHz counters (resulting in accuracy
between optical and acoustic axial lengths, using high- of ≈20 µm) triggered by the rising edges of the ocular interface
precision immersion ultrasound measurements. echoes. For each transmitted pulse, all ocular segments are ac-
quired simultaneously. Measurement results are given as means of
Whereas surgical decisions were still based on ultra- 30 of 32 successive single-shot measurements, with the lowest and
sound data, it was the purpose of this study to find out the highest value left out. For each eye, five independent measure-
what refractive results would have been obtained if opti- ments were taken; the most representative set of segmental dis-
cal instead of acoustic axial length data had been used tances (anterior chamber depth, lens thickness, vitreous length, ax-
for IOL calculation. ial length) was selected as the final result. The obtainable repro-
ducibility (average standard deviation for five consecutive axial
length measurements) with this instrument is 22±24 µm. A typical
echogram thus obtained is shown in Fig. 1.
Patients and methods For the optical axial length measurements with LIB, we used –
as has already been mentioned – the functional prototype of
Patients today’s Zeiss IOLMaster, denoted by “ALM”.
If OCT is compared to ultrasound B-scan imaging, then LIB is
Since October 1997, we have been routinely performing optical equivalent to obtaining an ultrasound A-scan. The basic principle
axial length measurements in addition to our standard ultrasound of optical biometry – PCI – is depicted schematically in Fig. 2 (af-
biometry on patients who attend our laboratory for cataract sur- ter [4, 18]): Infrared radiation (λ=780 nm) of short coherence
gery planning. length lC (≈160 µm) is emitted by a laser diode LS in a Michelson
For this study we retrospectively reviewed the data of 136 eyes interferometer setup and split into two separate coaxial beams E’
of 108 patients who had undergone routine cataract surgery with and E’’, with E’’ being delayed by twice the displacement d of the
subsequent implantation of a biconvex PMMA IOL type Rayner measuring mirror M’’. Both partial beams illuminate the eye to be
755 U in the capsular bag between July 1997 and October 1998. measured and are reflected at the cornea (C) and the retina (R).
Optical axial lengths were obtained for 118 eyes. In two eyes, no After passing through a beam splitter, all beam components are
PCI measurements were performed. In 16 (12%)of 134 eyes opti- detected by a photodetector (PD). Interference between different
cal axial length measurements could not be carried out. components (e.g. E’R and E’’C) can take place if their optical path
Postoperative refraction and visual acuity data were available length difference is smaller than the coherence length lC, i.e. if
for 103 of 118 eyes. Of these, four eyes with a best corrected visu- |2L–2d|≤lC. One of the characteristics of this two-beam set-up is
al acuity (BCVA) <0.3 were excluded from the study. Among its insensitivity to longitudinal eye movements.
these was one case of diabetic macular degeneration, one of cho- To obtain an “optical A-scan”, the interferometer mirror M’’ is
roidal atrophy and one of retinal detachment; one case was not moved with a constant speed giving rise to a Doppler modulation of
documented. A fifth case with BCVA=0.3 was also eliminated the interference intensity at the photo detector. The envelope of the
from the study due to a prominent maculopathy. Thus, 98 eyes of modulation signal is recorded as a function of the position of the mea-
88 patients were left for further evaluation. The mean age of the surement mirror M’’, which can be determined very precisely. As a
31 men (35%) and 57 women (65%) in the patient group was 71 result, the optical path length L between cornea and retina is obtained.
years (range 44–91 years) at the time of (the first) surgery.
Three surgeons were involved in the study. One surgeon per-
formed phacoemulsification with a corneoscleral tunnel incision
(6 mm) in 44 of the 98 eyes of the study group, another one extra-
capsular cataract extraction (ECCE) with corneoscleral incision
(7–9 mm) in 22 eyes; the third surgeon made use of phacoemulsi-
fication in 15 and ECCE in another 17 eyes – in all cases with cor-
neoscleral tunnel incisions (6 mm). The size of the capsulorhexis
in the phacoemulsification cases was 6 mm, as was the capsuloto-
my for ECCE. The no-stitch technique was used with phacoemul-
sification, whereas two to five single sutures were applied in
ECCE.
Anterior vitrectomies had to be performed in two cases due to
posterior capsule defects during phacoemulsification. In one
ECCE case, a prophylactic peripheral iridectomy was carried out;
in another one, a partial zonulolysis occurred. None of these intra-
operative complications, however, was found to be connected to
any long-term change in refraction.
Refraction of 91 of these 98 eyes was measured between
the 101st and 400th day postoperatively; four eyes were measured
within 100 days after surgery, three eyes after 400 days or more. The
mean value for the date of postoperative refraction determination
was 8.2±2.8 months (245±85 days; range 2–515 days) postoperative-
ly. One patient incidentally had his second eye operated on during
Fig. 1 Typical echogram obtained with the Grieshaber Biometric
his visit for refraction of his first eye. On this occasion, 2 days post-
System (GBS) in immersion coupling: four 40-MHz counters are
operatively, his second eye’s refraction was also measured.
triggered by the rising edges of the ocular interfaces, denoted by
the thin downward noses. Thus, for each transmitting pulse, all oc-
ular segments are acquired simultaneously
 767

Table 1 Ocular tissue data for optical biometry: group refractive length including retinal thickness. The individual group refractive
indices, valid for λ=780 nm, and (vertex) distances for ocular seg- indices were deduced by Hitzenberger [18], starting from the
ments. CTH Corneal thickness; ACD anterior chamber depth; LTH known phase refractive indices at λ=550 nm and assuming the dis-
lens thickness; VIT vitreous; RET retinal thickness; AXL axial persion of water for the ocular segments

Ocular segment

CTH ACD LTH VIT RET AXL nmean

Group refractive index n at 780 nm [17] 1.3856 1.3459 1.4070 1.3445 1.3445 (?)
Theoretical eye Gullstrand data (mm) [9] 0.50 3.10 3.60 16.80 0.15 24.15 1.3549
Biometric eye ultrasound data (mm) [1, 10] 0.51 2.62 4.48 15.87 0.14 23.62 1.3574

Fig. 4 Optical “A-scan” of an hyperopic eye with an axial length

of 21.48 mm as displayed by the present version of the ZEISS
IOLMaster (SNR signal-to-noise ratio)

length (GPL) by OPL=n GPL. In the ALM instrument, a group re-

fractive index of n=1.3549 [17] was “wired in”. Hence, the axial
length of AL=21.616 mm of Fig. 3 corresponds to an optical path
length of 29.29 mm.
Another optical trace of a short eye is shown in Fig. 4. This
Fig. 2 Optical biometry: principle of a dual-beam partial coher- plot stems from the final Zeiss PCI instrument, the IOLMaster.
ence interferometer (after [6, 19]) Analogous to the ultrasound measurements, we performed five
consecutive single measurements of the axial length with the
ALM instrument. The result with the best signal-to-noise ratio was
then selected as representative.
Optically and acoustically measured axial lengths cannot be
expected to yield the same values (cf. e.g. [14]): First, ultrasound
measures the distance from the anterior corneal vertex to the ILM,
whereas optically the distance up to the RPE is measured. Optical
axial lengths will thus be greater than acoustic ones. Second, seg-
mental measurements with individual sound velocities are possible
with ultrasound; optically, however, a mean group refractive index
equivalent to using a mean velocity in ultrasound has to be applied
to convert OPLs into geometrical distances. Last but not least: ul-
trasound measures along the optical axis of the eye, PCI – as a fix-
ation-bound method – along the eye’s visual axis.
In an earlier study ([14, 21]), we had obtained a relation be-
Fig. 3 Typical optical A-scan (regular and zoomed signal) as ob- tween axial lengths ALGBS from precision segmental immersion
tained with the ALM prototype of the ZEISS IOLMaster (ILM in- measurements with the GBS and OPL from PCI measurements
ternal limiting membrane; RPE retinal pigment epithelium; s side with the Zeiss ALM instrument:
lobes of the main signal, characteristic for the laser diode used
OPL/1.3549=ALGBS·0.9571+1.3033 (1)
This relation is also used (i.e. wired in) in the market version of
A typical PCI scan recorded with the ALM instrument is given the ZEISS IOLMaster. The value of 1.3549 denotes the average
in Fig. 3. In addition to the anatomically relevant peaks of the in- group refractive index for the eye at the instrument’s infrared radi-
ner limiting membrane (ILM) (about 0.2 mm apart) and the retinal ation wavelength λ=780 nm, derived by Hitzenberger [17] on the
pigment epithelium (RPE), two artifact signals denoted by “s” basis of the Gullstrand eye. Rather than theoretical data, we prefer
show up in Fig. 3. These signals are characteristic for the laser di- true biometric results ([1, 10]) (compiled in Table 1) for the calcu-
ode used and appear as side lobes to the respective main signal, lation of the mean refractive index nmean. With
equally spaced ≈0.8 mm apart from each other. The ILM peak, 5 d
however, is not regularly found on every optical scan – an obser- nmean = ∑ ni ⋅ i
vation which has already been made by other investigators, e.g. 1 L
[17]. (L=AXL; di=CTH, ACD, LTH, VIT, RET; and ni=respective indi-
The optical path length (OPL) in a medium is affected by its ces of refraction in Table 1) for the weighted mean, we arrived at a
group refractive index [4] n and related to the geometrical path value of 1.3574 [14].
 768

Table 2 Overview of different conversion algorithms used to In optimized mode, the constants a0, a1 and a2 are obtained by
translate optical path length (OPL) as acquired by the ALM instru- a separate optimization process. For each patient, the actual post-
ment into geometrical axial length (AL). AL0 Displayed by the operative refraction is used to calculate the corresponding optical
ALM instrument. Note: the conversion relation for AL4 is used in ACD. For this purpose, Eq. 2 is algebraically transformed to give
the Zeiss IOLMaster an expression for d. These values of d for all patients are then cor-
related with the preoperative ultrasound measurements of the
ID Biometry principle AL derived from (acoustic) ACD and the axial length. Double linear regression
analysis then yields the constants a0, a1 and a2.
GBS Immersion ultrasound Sum of ocular segments Characteristic for the optical ACD (Eq. 4) is its dependence on
AL0 Laser interference (PCI) OPL/1.3549 AC and L; Eq. 4 is not valid for AC=0. In this case the following
AL1 Laser interference (PCI) OPL/1.3549–0.14 approximation may be used:
AL2 Laser interference (PCI) OPL/1.3574
AL3 Laser interference (PCI) OPL/1.3574–0.14 d=(a0+u a1)+(a2+v a1) L for AC=0 (7)
AL4 Laser interference (PCI) (OPL/1.3549–1.3033)/0.9571 where u=–0.241 and v=0.139. To derive the numerical values of u
and v, use has been made of model calculations [13] yielding
mean values of AC=2.53 mm at axial lengths of L=20.0 mm and
Thus, for comparison with the ultrasound axial length, we de- AC=3.50 mm at L=27.0 mm. Equation (7) produces results com-
duced the respective PCI axial lengths from the primarily mea- parable to the relation d=ACDconst–3.976+0.17 L from [8].
sured OPL under five different conditions: for mean refractive in-
dices of 1.3549 and 1.3574 with and without compensating for the
retinal thickness of 0.14 mm [1] and, additionally, for relation 1. Results
These different conversion algorithms to translate OPL into geo-
metrical axial length – subsequently denoted by AL0 to AL4 –
are compiled in Table 2. Conversion AL0 corresponds to the mea- Table 3 gives statistical results of the immersion ultra-
surement display of the ALM instrument; AL4 – as mentioned sound measurements for the axial length AL and anterior
earlier – represents the IOLMaster calibration. chamber depth AC. Also, mean corneal radii CR, aver-
age implant lens powers IOL and average achieved post-
IOL calculation operative refractions REF are listed in Table 3. Table 4
summarizes means, standard deviations and ranges of the
IOL calculation was carried out according to Haigis [12]. Based ALs obtained from the different conversion algorithms
on the thin lens formula AL0–AL4 (cf. Table 2). While – as pointed out earlier –
RX conversion AL0 represents the measurement display of
DL = n − n with z = DC + and
L−d n −d 1 − RX ⋅ d x the ALM prototype, AL4 is used in the commercially
z
n −1 available version of the ZEISS IOLMaster. A compari-
DC = C (2) son of these two optical axial length conversions with
R
the GBS immersion ultrasound values is plotted in
where DL is power of implant lens, DC is corneal power, RX is de-
sired refraction, n is refractive index of aqueous and vitreous Fig. 5.
(=1.336), nC is fictitious refractive index of cornea (=1.3315), dX Axial length is the most influential parameter in IOL
is vertex distance between cornea and spectacles (=12 mm), R is calculation. Thus, for the ultrasonically measured AL as
corneal radius, L is axial length (as measured by ultrasound), d is well as for each optically derived length value, we used
optical ACD, the resultant refraction RX for an implant lens DL and
an optical ACD d is readily deduced: Eq. 2 to calculate the refraction expected from preopera-
tive biometry and keratometry as well as from the power
q − DC
RX = with of the IOL actually implanted.
1 + d x ⋅ (q − DC )
In standard, non-optimized mode, the optical ACD d
n ⋅ [n − DL ⋅ ( L − d )] was calculated according to Eq. 4 with a1=0.4, a2=0.1
h q= . (3)
n ⋅ ( L − d ) + d ⋅ [n − DL ⋅ ( L − d )] and a0 given by Eq. 5. For the ACD constant of the
The optical ACD d is given by Rayner lens 755 U, a value of 4.96 mm equivalent to an
d=a0+a1 AC+a2 L for AC≠0 (4) A-constant of 118.0 (D) was used. Although the manu-
facturer’s ACD constant in the lens data sheet is 4.7 mm
where AC is preoperative anterior chamber depth and L is preop-
erative axial length, as measured e.g. by ultrasound. and not 4.96 mm, we preferred this A-constant-based
The parameters a0, a1 and a2 are constants describing the im- value since – from our experience with lens constants of
plant IOL. In standard or default mode, a1 and a2 are set to various other IOLs – we have the impression that manu-
a1=0.4, a2=0.1 and facturers’ A-constants are more reliable than their ACD
a0=ACDconst–0.4 mean(AC)–0.1 mean (AL)=ACDconst–3.687 (5) constants. The refraction thus calculated (REF calc) is
with mean(AC)=3.37 mm, mean (AL)=23.39 mm [15] and
compared to the spherical equivalent of the actually
ACDconst=manufacturer’s ACD constant. Often, the A-constant achieved postoperative (subjective) refraction (REF true)
given by the manufacturer seems to be more reliable than the at best corrected vision. Figure 6 shows a box plot of the
ACD constant. The relationship between the two constants is pro- mean difference REF true–calc obtained in standard
vided by [22]: mode (std), i.e. with no optimization of lens constants
ACDconst=0.62467·A-constant–68.747. (6) a0, a1 and a2.
 769

Table 3 Statistical data (mean, standard deviation, range; n=98) for axial length (AL) and anterior chamber depth (AC) as measured with the
GBS in immersion ultrasound, corneal radius (CR), power (IOL) of intraocular lens and actually achieved postoperative refraction (REF)

AL (mm) AC (mm) CR (mm) IOL (D) REF (D)

Mean ± SD 23.37±1.22 3.06±0.50 7.72±0.32 21.1±3.0 –0.01±1.13

Range 20.32–27.45 1.95–4.62 6.76–9.01 9.0–29.0 –3.50–+2.50

Table 4 Statistical data (mean, standard deviation, range) for axial

length (AL) as measured with the GBS in immersion ultrasound
and deduced from PCI measurements with different conversion al-
gorithms

ID AL (mm)

Mean ± SD Range

GBS 23.37±1.22 20.32–27.45

AL0 23.71±1.17 21.01–27.52
AL1 23.57±1.17 20.87–27.38
AL2 23.66±1.17 20.97–27.47
AL3 23.52±1.17 20.83–27.33
AL4 23.41±1.23 20.59–27.39

Fig. 5 Comparison of optical axial length calibrations AL0 and

Table 5 Constants a0, a1, a2 and correlation coefficient R for the AL4 (cf. Table 3) with the acoustic axial length obtained by im-
prediction of the optical ACD d according to Eq. 3, obtained from mersion ultrasound with the GBS
double linear regression analysis and different axial length defini-
tions as given in Table 2

ID a0 a1 a2 R (%)

GBS 1.512 0.239 0.107 36.2

AL0 3.743 0.249 0.038 27.1
AL1 3.709 0.225 0.031 24.7
AL2 3.821 0.242 0.032 25.9
AL3 3.785 0.218 0.025 23.5
AL4 1.602 0.205 0.110 36.6

In optimized mode, we first calculated – for every pa-

tient – the very value of d in Eq. 2 which would have
produced the observed refraction on the basis of the ac-
tually implanted IOL power. Subsequently, in a double
linear regression analysis, these values of d were corre-
lated with the preoperative ACD (as measured by ultra-
sound) and the respective AL. As a result of this analy-
sis, three independent constants a0, a1, and a2 are ob-
tained. Data for the triples (a0, a1, a2) together with val-
ues for the correlation coefficients R for all of our differ- Fig. 6 Box plot of the mean deviation REF true–calc (∆REF) for
ent AL definitions are compiled in Table 5. By defini- the IOL calculation based on different axial lengths in standard
tion, the triple (a0, a1, a2) causes a zero prediction error mode, i.e. with no optimization of lens constants a0, a1 and a2
for the postoperative refraction. The respective box plots
for the mean prediction errors (analogous to Fig. 6) are
depicted in Fig. 7. Table 6 lists the numerical values for Finally, to check for a possible AL dependence of the
∆REF (REF true–calc) with and without optimization, prediction error, the individual deviations REF true–calc
together with the respective percentages of correct pre- (∆REF) are depicted in Fig. 10 versus ultrasound ALs
dictions within ±1 D and ±2 D. These percentages and for the calculations based on GBS data as well as the
their changes due to optimization of constants are also AL4 conversion.
plotted in Fig. 8 for the ±1 D range and in Fig. 9 for the
±2 D range.
770

Fig. 10 Axial length dependence of the deviation REF true–calc

(∆REF) for the IOL calculation based on different axial lengths

Fig. 7 Box plot of the mean deviation REF true–calc (∆REF) for Discussion
the IOL calculation based on different axial lengths in optimized
mode, i.e. with optimization of lens constants a0, a1 and a2 Comparing the mean optical ALs for the different con-
version algorithms in Table 4, all PCI lengths are found
to be longer than the ultrasound (GBS) data: e.g.
AL0–GBS=0.34±0.20 mm, AL2–GBS=0.30±0.20 mm,
and AL4–GBS=0.04±0.20 mm. This was expected and
has already been found in an earlier study [14], where
similar results had been obtained for 265 eyes:
AL0–GBS=0.30±0.17 mm, AL2–GBS=0.25±0.17 mm.
Drexler et al. [2] obtained a difference between optical
and acoustic ALs of 0.460 mm. However, apart from us-
ing a wavelength of λ=855 nm in their original experi-
mental Vienna instrument which also allowed segmental
AL measurements, their ultrasound measurements were
performed in applanation technique. In contrast to the
immersion method we applied, applanation ultrasound is
subject to the zero point error as well as to possible
Fig. 8 Percentages of correct refraction predictions within ±1 D shortenings of the AL due to globe compression during
with (opt) and without (std) optimization for axial lengths derived transducer contact. Thus, these results can not readily be
by different methods
compared. In an earlier study, Hitzenberger et al. [20],
compared PCI measurements at 780 nm to both immer-
sion and applanation ultrasound. They obtained a differ-
ence of 0.47±0.25 mm for 179 applanation measure-
ments and 0.18±0.12 mm for 50 immersion measure-
ments. Their Kretz 7200 MA instrument used for immer-
sion ultrasound cannot, however, match the precision of
the GBS device.
Conversion relation AL4 represents the original re-
gression dependence between optical and high-precision
acoustic ALs [14]. Hence, AL4 and GBS data should be
more or less identical, which is reflected in the observed
mean difference of 0.04 mm between AL4 and GBS.
This can also be seen from Fig. 5. The regression line
through the AL4 data points is given by y=0.9906x+
Fig. 9 Percentages of correct refraction predictions within ±2 D 0.2589, the slope of which being statistically not differ-
with (opt) and without (std) optimization for axial lengths derived ent from 1. On the other hand, the regression line for
by different methods AL0 is given by y=0.9481x+1.5515. Its slope of 0.9481
 771

Table 6 Mean prediction er-

rors ∆REF (true postoperative Without optimization (std) With optimization (opt)
minus calculated refraction ac-
cording to Eq. 2) and percent- ID ∆REF (D) ±1 D (%) ±2 D (%) ∆REF (D) ±1 D (%) ±2 D (%)
ages of correct predictions
within ±1 and ±2 D with and GBS –0.17±0.71 85.7 99.0 –0.03±0.69 86.7 99.0
without optimization for differ- AL0 0.74±0.76 69.4 94.9 –0.02±0.72 84.7 98.0
ent axial length definitions as AL1 0.37±0.75 78.6 96.9 –0.02±0.71 85.7 99.0
given in Table 2 (n=98) AL2 0.63±0.76 72.4 95.9 –0.02±0.72 85.7 99.0
AL3 0.26±0.75 79.6 95.9 –0.02±0.71 85.7 99.0
AL4 –0.06±0.72 85.7 98.0 –0.01±0.71 84.7 99.0

is significantly different from 1, not so from 0.9571, the ages of correct predictions within ±1 D and ±2 D with
slope of Eq. 6. Thus, our earlier findings in [14] are con- and without optimization. Since the manufacturer’s con-
firmed. A slope <1 is caused by using a mean (group) re- stant together with the GBS and AL4 data had already
fractive index for the conversion of optical to geometri- produced prediction errors close to zero (see above), not
cal ALs, which cannot account for the overproportional much was left to be optimized. For AL0 to AL3, howev-
influence of the lens in short eyes. This behavior was er, the number of correct predictions within ±1 D and
found and described for the first time by Haigis and Lege ±2 D could be significantly increased, e.g. for AL0 from
[14]. A detailed publication in English is presently being 69.4% to 84.7% (±1 D) and 94.9% to 98% (±2 D). Thus,
prepared. by optimization, it was possible to achieve correct re-
In standard IOL calculation mode (std), the differ- fraction predictions within ±1 D in 84.7%–85.7%, within
ences in mean AL due to the specific conversion formu- ±2 D in 98.0%–99.0% of cases, irrespective of the opti-
las are directly translated into systematic shifts of the re- cal length conversion used.
spective prediction errors REF true–calc (cf. Table 6 and With immersion ultrasound, minimally better results
Fig. 6). The rule of thumb stating that 1 mm axial change (by 1%) were obtained in the ±1D group – the difference
is equivalent to a 3 D change in IOL power, of which was not statistically significant (cf. Table 6).
70–80% (≈2–2.5 D) shows up in the spectacles’ plane, Ideally, a mean zero prediction error is independent of
can be observed in action in Table 6 and Fig. 6. The GBS AL, although it may well be produced by equal errors,
(–0.17±0.71 D) and AL4 (–0.06±0.72 D) results lie clos- albeit of opposite sign for short and long eyes. In
est to each other, which again is to be expected since the Fig. 10, the individual deviations REF true–calc are plot-
AL4 calibration originally stems from GBS data. The ted against ultrasound AL values for the calculations
fact that both results are pretty close to zero must not be based on GBS as well as AL4 data. As can be seen, there
overrated and is considered coincidental; in general, IOL is virtually no AL dependence, the slopes of the respec-
constants need to be individualized in order to be free of tive regression lines being statistically indistinguishable
systematic errors. from zero. The reason for this behavior is due to the fact
Optimization of constants as described above pro- that with three constants rather than with just one classi-
duced individual triples (a0, a1, a2) for all conversion cal lens constant it is possible to model the AL depen-
formulas, listed in Table 5. By definition, these constants dence of the optical ACD [11, 16].
are derived thus as to produce a mean zero prediction er- It follows from this study that with the AL4 conver-
ror. Therefore, the box plots of Fig. 7 are all centered sion (built into the Zeiss IOLMaster) optical PCI biome-
round zero. The distribution of prediction errors seems to try is directly comparable to high-precision immersion
be slightly smaller for the optical axial lengths; an outlier ultrasound measurements. This holds especially for the
in the GBS results with an absolute deviation REF quality of refractive outcome after cataract surgery.
true–calc=2.3 D is identified to stem from a myopic Drexler et al. [2, 3] found an improvement of 27% in
eye with an acoustic AL of 25.87 mm. With the AL4 absolute postoperative refractive error of 85 cases if IOL
calibration, the respective optical AL was measured at calculations were based on AL values from their experi-
25.33 mm with a smaller resultant deviation REF mental PCI instrument at λ=855 nm instead of ultra-
true–calc of 1.1 D. This coincides with our general expe- sound data. Their acoustic measurements, however, had
rience that in myopic eyes optical ALs seem to be more been performed with the Alcon OcuScan in applanation
reliable than acoustic data. This may be explained by technique. Acoustic results based on this method – as
myopic fundus changes affecting the macula position: discussed above and commonly accepted – are of lesser
with PCI measuring along the visual axis there is a better quality than immersion results. In addition, Drexler and
chance of obtaining the relevant AL value than with ul- colleagues used the SRK II formula, which – although
trasound. still widely applied – is well known to have poorer per-
The effect of optimizing constants can be seen from formance than theoretical formulas. Thus, it seems not
Table 6 and from Figs. 8 and 9, which depict the percent- surprising that optical biometry increased their percent-
772

age of correct refraction predictions within ±1 D from biometries that will be replaced by laser interferometry.
72.9% to 85%, within ±2 D from 96.4% to 100%. “Normal” cataract eyes with visual acuities ≥0.1 [23]
It has already been mentioned above that no optical and without additional pathologies will be ideal candi-
measurements could be obtained from 16 (12%) of 134 dates. However, further research is necessary, especially
eyes. It was not the aim of this study, to analyze the rea- with respect to the optical properties of the ocular media,
sons for and the circumstances of these failures. A sepa- e.g. their group refractive indices. Analogous to ultra-
rate project devoted exclusively to this problem is pres- sound properties of the lens, it is expected that different
ently under way. In a former study [21], we had found cataract forms will give rise to different lens refractive
that 58 (9%) of 678 eyes could not be measured optical- indices which affect the OPL. Allowance may be made
ly. Among the reasons were inability to cooperate (fix- for these effects in future PCI instrument versions de-
ate), tremor, respiratory distress, severe tear film prob- signed for segmental optical measurements. Further re-
lems, keratopathy, corneal scarring, mature cataract, nys- search has to deal with the origin and incidence of addi-
tagmus, lid abnormalities, vitreous hemorrhage, mem- tional signals which sometimes show up in the optical
brane formation, maculopathy and retinal detachment. A-scans. Also, influences of pathologic structures such
Schrecker and Strobel [23] report that 10 (11.1%) of 90 as membranes on the PCI measurement need to be as-
eyes were not suitable for LIB. Thus, it seems that with sessed in more detail.
present technology the eyes of 9–12% of the patients of a PCI biometry today is a user- and patient-friendly
university eye clinic cannot be measured by laser inter- method for AL determination and IOL planning in
ferometry. In these cases, ultrasound biometry will con- the preparation of cataract surgery. Its accuracy is supe-
tinue to be indispensable. rior to that of the commonly used applanation method
On the other hand, if an optical signal has been ob- and is directly comparable to that of high-precision im-
tained, it is due to the fact that the interference condi- mersion ultrasound. In fact, the Zeiss IOLMaster simu-
tions were met. So, a signal if present represents a cor- lates precise segmental immersion ultrasound measure-
rect AL measurement. LIB thus allows reliable, high- ments by means of its inbuilt conversion relations. The
precision measurements to be performed even by un- new optical biometry technique has the potential to be-
trained users. come a routine method for IOL calculation in cataract
With the commercial version of PCI in the Zeiss IOL- surgery in cases of otherwise “normal” cataract eyes
Master having been available for only a short time, it without additional pathologies and with visual acuities
is too early to predict the proportion of ultrasound ≥0.1.

References
6. Fercher AF, Mengedoht K, Werner W 11. Haigis W (1996) Einfluß der Optik-
1. Decker D, Trier HG, Irion KM, Lepper (1988) Eye length measurement by in- form auf die individuelle Anpassung
RD, Reuter R (1980) Rechnergestützte terferometry with partially coherent von Linsenkonstanten zur IOL-
Gewebsdifferenzierung am Auge. light. Optics Lett 13:186 Berechnung. In: Rochels R, Duncker
Ultraschall Med 1:284–296 7. Findl O, Drexler W, Menapace R, GIW, Hartmann C (eds): 9.Kongress d.
2. Drexler W, Findl O, Menapace R, Bittermann S, Fercher AF (1999) Deutsch. Ges. f. Intraokularlinsen
Rainer G, Vass C, Hitzenberger CK, Hochpräzisionsbiometrie pseudophaker Implant., Kiel 1995, Springer, Berlin
Fercher AF (1998) Partial coherence Augen mittels Teilkohärenz-Laser- Heidelberg, pp 183–189
interferometry: a novel approach to interferometrie. Ohrloff C, Kohnen T, 12. Haigis W (1998) IOL calculation
biometry in cataract surgery. Duncker G (eds) 11. Kongreß der DGII according to Haigis.
Am J Ophthalmol 126:524–534 1997. Springer, Berlin Heidelberg, http://www.augenklinik.uniwuerzburg.
3. Drexler W, Findl O, Menapace R, pp 120–124 de/uslab/ioltxt/haie.htm, last revision:
Rainer G, Vass C, Hitzenberger CK, 8. Haigis W (1995) Biometrie. In: Straub 7 December 1998
Fercher AF (1999) Teilkohärenz-Laser- W, Kroll P, Küchle HJ (eds) Augen- 13. Haigis W, Gross A (1997) Modell-
interferometrie: eine neue hochpräzise ärztliche Untersuchungsmethoden. rechnungen zur Vorhersage von IOL-
Biometrie-Methode zur Verbesserung Enke, Stuttgart, pp 255–304 Konstanten. In: Vörösmahrty D,
der Refraktion nach Kataraktchirurgie. 9. Haigis W (1995) Biometrie. In: Duncker G, Hartmann C (eds):
In: G.Duncker, C.Ohrloff, F.Wilhelm Kampik A (ed) Jahrbuch der Augen- 10. Kongress d. Deutsch. Ges. f. Intra-
(eds), 12. Kongress der DGII 1998. heilkunde 1995: Optik und Refrak- okularlinsen-Implantation und Re-
Springer, Berlin Heidelberg, p 471 tion. Biermann, Zülpich, pp 123– fraktive Chirurgie, Budapest 1996.
4. Fercher AF (1996) Optical coherence 140 Springer, Berlin, Heidelberg, pp 288–
tomography. J Biomed Opt 1:157–173 10. Haigis W (1996) Biometrie bei kom- 294
5. Fercher AF, Roth E (1986) Ophthalmic plizierten Ausgangssituationen. In:
laser interferometer. Proc SPIE Rochels R, Duncker GIW, Hartmann C
658:48–51 (eds) 9. Kongress d. Deutsch. Ges. f.
Intraokularlinsen-Implant., Kiel 1995.
Springer, Berlin Heidelberg, pp 17–
26
 773

14. Haigis W, Lege BAM (2000) Ultra- 17. Hitzenberger CK (1991) Optical 21. Lege BAM, Haigis W (2000) Erste
schallbiometrie und optische Biome- measurement of the axial eye length klinische Erfahrungen mit der opti-
trie. In: Kohnen T, Ohrloff C, Wenzel by laser Doppler interferometry. schen Biometrie. In: Kohnen T, Ohrloff
M (eds): 13. Kongress d. Deutschspr. Invest Ophthalmol Vis Sci 2:616– C, Wenzel M (eds), 13. Kongress d.
Ges. f. Intraokularlinsen-Implant. und 624 Deutsch. Ges. f. Intraokularlinsen-
Refraktive Chirurgie, Frankfurt 1999. 18. Hitzenberger CK (1992) Measurement Implant. u. Refrakt. Chirurgie, Frank-
Biermann, Cologne, pp 180–186 of corneal thickness by low-coherence furt 1999. Biermann, Cologne, 175–179
15. Haigis W, Waller W, Duzanec Z, interferometry. Appl Opt 31:6637– 22. Retzlaff JA, Sanders DR, Kraff M
Voeske W (1990) Postoperative bio- 6642 (1990) Lens implant power calculation,
metry and keratometry after posterior 19. Hitzenberger CK, Mengedoht K, 3rd edn. Slack, Thorofare, NJ, USA
chamber lens implantation. Fercher AF (1989) Laseroptische 23. Schrecker J, Strobel J (2000) Optische
Eur J Implant Ref Surg 2:191–202 Achsenlängenmessung am Auge. Achsenlängenmessung mittels
16. Haigis W, Duzanec Z, Kammann J, Fortschr Ophthalmol 86:159–161 Zweistrahlinterferometrie. In: Kohnen
Grehn F (1999) Benefits of using 20. Hitzenberger CK, Drexler W, Dolezal T, Ohrloff C, Wenzel M (Hrsg): 13.
three constants in IOL calculation. C, Skorpik F, Juchem M, Fercher AF, Kongress d. Deutschspr. Ges. f. Intra-
http://aao5.aao.org/am98_99/am/am/99 Gnad HD (1993) Measurement of the okularlinsen-Implant. u. Refrakt.
/sciprog/Detail3.cfm?EVENTS__EID axial length of cataract eyes by laser Chirurgie, Frankfurt 1999. Biermann,
NO=7713 doppler interferometry. Cologne, pp 169–174
Invest Ophthalmol Vis Sci 34:1886–
1893