Sunteți pe pagina 1din 24

1

IMAGE Meeting 2013 Vienna



PROGRAMME AND ABSTRACTS

Venue and Hotel for the meeting
The venue is the olu phaimacy in the olu (foimei) vienna ueneial Bospital (Allgemeines
Kiankenhaus, AKB). It is nicely locateu in the 1st yaiu of the olu AKB, S-1u min walk foim
the hotel, anu also S-1u min walk to the piesent builuing of the AKB. The venue is calleu
"Aula am Campus" (see map below).
The hotel is nameu Botel Regina, Rooseveltplatz 1S, A-1u9u Wien (+4S 1 4u4 46 u)


2

WEDNESDAY: get-together-dinner
Meet at 19.40 in hotel lobby (we will have a 15 min walk to the restaurant)
For those who missed to meet in the lobby
Dinner venue: Restaurant Schnattl
Address: 1080 Wien, Lange Gasse 40 (+43 1 4053400)



S

THURSDAY:
9.00 13.00: meeting
13.00 14.00: lunch
14.00 17.30: meeting
19.30: dinner (meet at 19.00 in hotel lobby)
We will have a 25 min walk to the restaurant

For those who missed to meet in the lobby
Dinner venue: Restaurant Le Ciel, Grand Hotel Wien
Address: 1010 Wien, Krntner Ring 9 (+43 1 515 80 9100)




4

or by public transport: Take the tramway Nr. 1 or 71 at Schottenring, exit at the opera,
then 5 minutes walk to the restaurant along the Ringstrae.

S


FRIDAY:
8.30 12.30: meeting
13.00 14.00: lunch
14.30 18.00: Guided Vienna City Tour (optional)
Meet at 14.20 in hotel lobby

Our host in Vienna will be
Clemens Vass (clemens.vass@meduniwien.ac.at; mobile phone: + 43676 5104678)

You may also try to contact
Hans Lemij (hlemij@me.com; mobile phone +31655863578).
6




7
Comparison of birefringence-related data between scanning laser polarimetry
(GDx) and polarization sensitive optical coherence tomography (PS-OCT).

Clemens Vass
1
, Ivania Pereira
1
, Stefan Zotter
2
, Stephan Holzer
1
, Hemma Resch
1
, Michael
Pircher
2
; Christoph Hitzenberger
2


1. Dept of Ophthalmology, Medical University of Vienna, Vienna, Austria.
2. Center for Medical Physics and Biomedical Engineering, Medical University of Vienna,
Vienna, Austria.

Purpose: To compare the retardation of the retinal nerve fiber layer (RNFL) measured
with the scanning laser polarimeter GDx-VCC and GDx-ECC (Carl Zeiss Meditec Inc.)
with the retardation as measured with polarization sensitive optical coherence
tomography (PS-OCT).

Methods: A sample of 8 healthy volunteers, 13 glaucoma suspects and 5 early glaucoma
patients underwent complete ophthalmologic examination, including visual field testing,
GDx-VCC and GDx-ECC and PS-OCT. Early glaucoma was defined as glaucomatous
optic neuropathy with a visual field mean deviation (VF-MD) of -4 dB or better. The
definition of glaucoma suspect was based on the suspect appearance of the optic disc
while VF was normal. PS-OCT was measured with a system operating at 70kHz A-scan
rate (scan angle 30x30 or 40x40, scan pattern 1024x250 A-scans). We compared the
average retardation of the superior and inferior sectors. The thickness values of the GDx
were converted to degree of retardation for comparison using the published fixed
conversion factor. When converting the data we furthermore took into account for the
different wavelengths of the GDx and our PS-OCT.

Results: For the mean values (deg.) and standard deviations of the three different
measurements please see the table. For GDx-ECC, GDx-VCC and PS-OCT we found
statistically significant differences between the healthy and the suspect groups (p<0.05).
There was also a small but statistically significant difference between the retardation
found with the GDx and the PS-OCT, the latter giving slightly higher values (p<0.05).

Healthy Suspects Early glaucoma
Age 49.5 60.2 59.9
PS-OCT superior 20.84 (2.35) 17.36 (3.68) 16.71 (6.60)
PS-OCT inferior 22.36 (1.80) 18.37 (3.30) 16.94 (4.40)
GDx-ECC superior 19.01 (1.20) 16.06 (1.60) 14.87 (4.05)
GDx-ECC inferior 18.57 (2.07) 15.34 (2.08) 13.14 (2.64)
GDx-VCC superior 19.53 (0.93) 16.22 (1.71) 14.98 (4.05)
GDx-VCC inferior 19.22 (2.23) 15.53 (2.26) 12.86 (3.02)


Conclusions: The retardation values given by the GDx and the PS-OCT are comparable
as are the differences between the groups.

8
Doppler OCT
Leopold Schmetterer
Department of Clinical Pharmacology, Center of Medical Physics and Biomedcial
Engineering, Medical University of Vienna, Vienna, Austria

Optical coherence tomography (OCT) has become a standard tool for assessing optic
disc morphology and retinal nerve fiber layer thickness in glaucoma. In the recent
years several functional extensions of the technique have been proposed. One
approach is Doppler OCT, which combines the principles of laser Doppler velocimetry
with those of OCT. This yields in principle information on both, retinal blood velocity
and retinal vessel diameter. We developed a technique based on diameter
measurements with a fundus-camera based system and bi-directional Fourier Domain
Doppler OCT. Using this technique it is possible to measure total blood flow in both
arteris and veins. Comparison of velocity data as obtained with bi-directional Fourier
Domain Doppler OCT showed good correlation with laser Doppler velocimetry data
during both normoxic and hyperopic conditions. In addition, data measured at retinal
bifurcations indicate the validity of the technique. Finally, measurements at retinal
arterial and venous vessels show that absolute retinal blood flow can be extracted with
high precision. The technique may be suitable to test the hypothesis that reduced
retinal blood flow is a risk factor for primary open angle glaucoma.

9
valiuation of 0CT attenuation coefficient estimation by phantom
measuiements

!"#"$%&'(&&'
)#
*$+"+"#"$,&-.
/
*$+"$01
/
*$2"3"$4&(56
)7
*$+"8"$-&$71&'
)#*/$

A
Rotteiuam 0phthalmic Institute,
B
ulaucoma Seivice,
1
Rotteiuam Eye Bospital, Rotteiuam, The Netheilanus;
2
Bept. of Physics anu Astionomy, v0 0niveisity, Amsteiuam, The Netheilanus.


!"#$%&'()
We have pieviously intiouuceu a novel methou to ueteimine the attenuation coefficients
(ACs) of the ietinal neive fibei layei anu othei ietinal layeis fiom 0CT uata. Bowevei, the
tiue ACs of these tissues aie not known. We peifoimeu contiolleu phantom stuuies to
valiuate the accuiacy of the AC estimation algoiithm.

*'+,%-()
Two methous weie evaluateu:
1. The conventional cuive fitting appioach, wheie the 0CT signal fiom a segmenteu
layei is appioximateu by an exponentially uecaying function, wheie the
exponential paiametei pioviues an estimate of the AC.
2. 0ui novel methou that iteiatively solves the AC fiom a foiwaiu single-scatteiing
mouel of light scatteiing in tissue. It enables the calculation of the AC of eveiy pixel
in the scanneu volume. These values weie then aveiageu ovei the iegion of
inteiest.

./0"#')12 0CT scan of the layeieu phantom
(left) anu coiiesponuing AC image (iight).
1u
Phantoms weie cieateu fiom uiffeient
concentiations of Ti02 paiticles in silicone.
0nifoim samples with uiffeient concentiations
(up to 1 w%) weie built. In auuition, a multi-
layeieu phantom containing five 17S m thin
layeis of vaiying concentiations (u.S w%, u.uS
w%, u.2S w%, u.u2S w%, u.1 w%), mounteu on
a thick substiate (u.2S w%), was cieateu.
All phantoms weie imageu on an expeiimental
1Suu nm swept-souice 0CT system anu AC
images weie calculateu (see Figuie 1 foi an
example of the layeieu phantom). Nean uepth-
piofiles weie calculateu fiom these images anu
aveiage ACs weie then estimateu by both
methous on all phantoms (see Figuie 2).

3'&"4+&()
The estimateu aveiage ACs foi both methous foi the unifoim samples aie shown in Figuie
S. The measuiements aie inuicateu by ciosses (blue foi the cuive fitting, ieu foi oui novel
methou). The uasheu lines aie fits of a mouel that assumes a lineai ielation between Ti02
concentiation anu attenuation, with an offset foi scatteiing in the silicone
( ( ) ( ) bx a y + = log log ). The fitteu values weie a =1.S anu b =7.S foi the 0CT slope methou
anu a =1.u anu b =7.6 foi the pioposeu methou.
The aveiage ACs weie estimateu foi eveiy layei in the layeieu phantoms. The
ieconstiucteu ACs foi the layeieu phantom aie shown in Figuie 4 (blue foi the cuive
fitting, ieu foi oui methou). In this case, the fitteu mouel paiameteis weie a =1.S anu
b =S.2 foi the 0CT slope methou anu a =u.1 anu b =1S foi the pioposeu methou.


./0"#')52 AC estimation on unifoim phantoms. ./0"#')62 AC estimation on a layeieu phantom.


)
./0"#')72)Nean uepth-piofiles foi 0CT uata
(top) anu ieconstiucteu ACs (bottom). The ieu
lines inuicate the fits foi estimating the
aveiage AC of eveiy layei.
11
8%9:4"&/%9&()
In unifoim phantoms, AC estimation by oui novel methou is consistent with the 0CT cuive
fitting methou foi concentiations laigei than u.1 w%. Foi smallei concentiations, the
cuive fitting methou iesulteu in a constant AC, while ACs estimateu by the pioposeu
methou coiielates with concentiations uown to u.u2S w%.
0n the layeieu phantom, the pioposeu methou piouuceu ACs that coiielate well with the
Ti02 concentiations of the layeis.

12
;9+'#&"<=':+):%##'4>+/%9)<'+?''9):/#:"@$>$/44>#A)-/&+#/<"+/%9)%B)#'+/9>4)9'#C')B/<'#)4>A'#)
>9-)#'+/9>4)C'&&'4&():%@$>#/&%9)<'+?''9)%$+/:>4):%,'#'9:')+%@%0#>$,A)>9-)&:>99/90)4>&'#)
$%4>#/@'+#A)
Ivania Peieiia
1
, Stephanie Webei
1
, Stephan Bolzei
1
, Bemma Resch
1
, Baibaia Kiss
1
, ueoig Fischei
2
,
Clemens vass
1


1. Bepaitment of 0phthalmology & 0ptometiy, Neuical 0niveisity vienna, vienna, Austiia
2. Centei foi Neuical Statistics Infoimatics anu Intelligent Systems, Section foi Neuical
Infoimation Nanagement anu Imaging, Neuical 0niveisity vienna, vienna, Austiia

!"#$%&')
To assess ciicumpapillaiy ietinal vessel thickness (cRvT) piofiles anu coiielate them with ietinal
neive fibei layei (RNFL) thickness measuieu by high iesolution optical coheience tomogiaphy
(BR-0CT) anu scanning lasei polaiimetiy (SLP).

*'+,%-&)
A sample of 1u6 healthy volunteeis unueiwent complete ophthalmologic examination, incluuing
BR0CT scanning (Ciiius Cail Zeiss Neuitec Inc.) anu SLP (uBx-vCC anu uBx-ECC Cail Zeiss
Neuitec Inc.). A piopiietaiy softwaie was uevelopeu in NATLAB (veision R2uu9b, The
Nathwoiks Inc.) foi inuiviuual manual assessment of the optic uisc (0B) boiuei anu cRvT, using
the SL0 image fiom Ciiius BR-0CT centeieu in the 0B. Foi each measuiable vessel, we establisheu
a line connecting its centei point at the 0B boiuei with the 0B centei. The angle between this line
anu a hoiizontal line passing thiough 0B centei was ueteimineu. Inuiviuual cRvT measuiements
weie convoluteu with a uaussian winuow, geneiating a cRvT piofile uepenuent on the numbei of
sectois measuieu in each uevice (2S6 sectois foi BR-0CT oi 64 sectois foi both uBxvCC anu
uBxECC). Inteisubject vaiiability was ueteimineu using NATLAB.

3'&"4+&)
The meuian values foi inteisubject coiielation between cRvT anu RNFL thickness anu the
numbeis of significant sectois as measuieu with Ciiius BR-0CT, uBx-ECC anu uBx-vCC weie
u.2SS (1862S6 sectois statistically significant), u.192 (S264) anu u.19S (SS64). 9u%
peicentiles of the inteisubject coiielation weie u.4uu, u.SSS anu u.SS6, iespectively.

8%9:4"&/%9&)
Both SLP methous (uBx-vCC anu uBx-ECC) yielu compaiable iesults iegaiuing the inteisubject
coiielation of the cRvT piofile anu the RNFL with about Su% of the sectois showing significant
coiielation. Bowevei, iegaiuing the BR-0CT, oui uata showeu a tenuency towaius a bettei
1S
coiielation which was statistically significant in 7u% of the sectois. Baseu on oui iesults, up to
2S% of the inteiinuiviuual vaiiation of the RNFL thickness might be explaineu by the vaiiation of
the cRvT.
14
DEEFGDH)8EI!DGJ)DGKGD)FGL).E3)3GL;MND)DNHG3)FGO*GMLNL;EM)
}elena Novosel
1, 2
, Koeniaau A. veimeei
1
, uijs Thepass
1
, Bans u. Lemij
S
, Lucas }. van vliet
2
1
Rotteiuam 0phthalmic Institute, Rotteiuam Eye Bospital, Rotteiuam, The Netheilanus
2
Quantitative Imaging uioup, Faculty of Applieu Science, Belft 0niveisity of Technology,
Belft, The Netheilanus
S
ulaucoma Seivice, Rotteiuam Eye Bospital, Rotteiuam, The Netheilanus


!"#$%&')
A novel level-set methou foi ietinal layei segmentation is piesenteu anu its accuiacy anu
iepiouucibility aie evaluateu. The methou exploits anatomical infoimation specific to the
ietina such as the pieuefineu oiuei of the layeis anu incoipoiates piioi knowleuge about
the intensity uistiibution anu thickness of the layeis.
)
*'+,%-)
The level-set methou is applieu to the attenuation maps obtaineu fiom in-vivo human
ietinal 0CT uata. As an optical piopeity of the tissue, the attenuation coefficient is not
affecteu by common imaging aitefacts such as shauing. The bettei consistency is theiefoie
expecteu to impiove the iobustness of segmentation methous.
0ui methou peifoims simultaneous localization of the inteifaces between ietinal layeis
using a level set appioach, wheie piioi knowleuge anu image uata aie incoipoiateu in a
piobabilistic fiamewoik by application of Bayes' theoiem. The pieuefineu oiuei of the
layeis is incoipoiateu via a novel level set coupling appioach. Equation (1) uesciibes the
upuate step foi the level set function () that uefines the 5
9:$
inteiface. The expiession
consists of multiple teims wheie the fiist two ensuie piopei oiueiing of the layeis. The
thiiu teim uesciibes the piobability of a pixel belonging to a ceitain layei anu the last
teim contiols the smoothness of the iesulting inteiface.

0ne eye foi each of the 24 noimal subjects was imageu with a Spectialis 0CT system. Ten
eyes weie useu foi tiaining of the algoiithm, the othei fouiteen eyes weie useu to assess
the accuiacy. Six eyes weie imageu again on the next uay to evaluate the iepiouucibility.
0ne B-scan of eveiy scan was manually segmenteu.
Thiee inteifaces weie consiueieu: the vitieous - ietinal neive fibei layei (RNFL)
inteiface, the RNFL-ganglion cell layei (uCL) inteiface anu the innei plexifoim layei (IPL)
- innei nucleai layei (INL) inteiface. Baseu on these thiee inteifaces, two layeis (the
RNFL anu the uCC, consisting of the RNFL, uCL anu IPL) weie uefineu anu theii
thicknesses weie analyseu. The ioot mean squaie eiioi (RNS), mean absolute ueviation
(NAB) anu Bice coefficient weie useu as eiioi metiic. Bloou vessels weie excluueu fiom
the evaluation.
3'&"4+&)
An example of the segmentation iesult obtaineu on the attenuation coefficient image of a
noimal eye is shown in figuie 1 anu one obtaineu on a glaucoma eye in figuie 2. All
evaluation iesults aie listeu in table 1. The accuiacy expiesseu by the RNS eiioi vaiies
between 4.2 - 9.S m (1-2.S pixels). Repiouucibility eiiois foi automateu (RNS S.9 - 8.S
m) anu manual (RNS 4.9 - 11.4 m) segmentation aie in the same iange as the accuiacy.
1S
When compaiing the iepiouucibility eiiois foi the manual anu automatic segmentation,
automatic segmentation shows on aveiage a 1.7 m smallei RNS eiioi. Eiioi measuies
foi the accuiacy aie up to 2 m bettei than the iepiouucibility of manual annotations.
)
8%9:4"&/%9)
The automateu segmentation shows a goou agieement with the segmentation peifoimeu
manually by a meuical uoctoi. The automateu methou pioviues a moie consistent
segmentation than manual labelling. The accuiacy of the automateu segmentation anu the
intia-obseivei iepiouucibility of manual annotations aie similai, suggesting that the
accuiacy of the automatic segmentation is at least as goou as the manual annotations.

Table 1: Accuracy of the automated segmentation and reproducibility of automated (A) and manual (M)
segmentation for the first three interfaces and two layers.
Accuiacy Repiouucibility
RNS
(m)
NAB
(m)
Bice
RNS (m) NAB
(m)
Bice
A N A N A N
vitieous -
RNFL
4.2 S.1 na S.9 4.9 S.2 S.8 na na
RNFL -uCL 8.9 S.6 na 8.1 1u.S S.7 6.8 na na
IPL - INL 6.4 4.7 na 6.4 7.S 4.S S.7 na na
RNFL
thickness
9.S 6.6 u.9S 8.S 11.4 S.9 7.8 u.9S u.9S
uCC
thickness
6.7 6.2 u.96 7.2 8.S 4.9 6.S u.97 u.96







Figure 2: Automated segmentation obtained on a B-scan superior to the optical nerve head
of a glaucoma eye.
Figure 1: Automated (yellow) and manual (red) segmentation obtained on a B-scan
superior to the optical nerve head of a normal eye.
16
Structural and functional monitoring of glaucoma patients: a comparison using regression
analysis

Gijs Thepass
1
, Koenraad A. Vermeer
1
, Josine van der Schoot
1
, Hans G. Lemij
1,2


1
Rotterdam Ophthalmic Institute, Rotterdam Eye Hospital, Rotterdam, The Netherlands
3
Glaucoma Service, Rotterdam Eye Hospital, Rotterdam, The Netherlands

Purpose
Structural and functional measurements play a key role in the follow-up of glaucoma patients.
Unfortunately the agreement between structural and functional measurements in glaucomatous
progression is generally poor. For Scanning Laser Polarimetry (SLP), the agreement with functional
measurements has been explored relatively little.

Methods
All eligible eyes of glaucoma patients from the Rotterdam Glaucoma Imaging Study with at least 5
follow-up measurements were enrolled (n=90) in this study. The average time between the first and last
visit was 28 months (22-32). The proprietary progression programs for both SLP (GDx ECC, Carl Zeiss
Meditec) and Standard Automated Perimetry (SAP) (Humphrey Field Analyzer II, 24-2 white on white
SITA test program, Carl Zeiss Meditec) were used (GDx GPA Fast mode and HFA GPA, respectively).
The GDx GPA flags any progression either as no progression, as possible or as likely. As an
comparison we also calculated linear regression coefficients for each series of measurements of the
SAP Visual Field Index (VFI) and Mean Deviation (MD), and the GDx peripapillary retinal nerve fibre
layer thickness (TSNIT average).

Results
All results of the regression analyses are presented in Figs. 1 & 2. None of the tested eyes showed
statistically significant progression by HFA GPA, but 5 (6%) eyes showed likely progression by GDx
GPA (marked by red triangles in Figs. 1&2). Eight eyes (9%) showed TSNIT average increases (marked
by green diamonds in figures 1&2). In our own regression analyses, 52 eyes (58%) showed disease
progression by HFA VFI, 60 eyes (67%) by HFA MD and 49 eyes (54%) by GDx TSNIT average. In 31
eyes (34%), there was both structural and functional progression by HFA VFI and TSNIT average (see
Figs.). Both structural and functional regression occurred in 18 eyes (20%).
17

Conclusions
Glaucomatous progression may be detected by SLP and by SAP. There was poor agreement between
GDx and HFA progression, regardless of the used analysis. HFA GPA was more conservative than
linear regression analysis. One of the limitations of this study is the relatively short average follow-up
time. Plots that combine data from both structural and functional test methods (such as in Figs. 1 & 2)
may help clinicians appreciate whether structural and functional measurements agree. In case of such
agreement, they may feel more confident in their glaucoma management.












18
Nomdo M. Jansonius,
1,2
Julia Schiefer,
3
Jukka Nevalainen,
4
Jens Paetzold,
5
Ulrich Schiefer
5


1.Department of Ophthalmology, University of Groningen, University Medical Center Groningen, Groningen, The Netherlands
2.Department of Epidemiology, Erasmus Medical Center, Rotterdam, The Netherlands
3.University Eye Hospital Munich, Ludwig Maximilians University, Munich, Germany
4.University Eye Hospital Oulu, Oulu, Finland
5.Centre for Ophthalmology, University of Tbingen, Tbingen, Germany

Relationship between Retinal Vessel Course and Retinal Nerve Fiber Bundle
Trajectories in the Human Eye

Purpose: Jansonius et al. developed a mathematical model for the description of retinal nerve
fiber bundle trajectories in the human eye (Vision Res 2009; Exp Eye Res 2012). Refraction
and possibly optic disc inclination explained some of the observed variability in the superior-
temporal region but not in the inferior-temporal region. Aim of the current study was to
determine the relationship between the retinal vessel course and the retinal nerve fiber bundle
trajectories.
Methods: The previous dataset of high-quality fundus photographs was used, comprising 28
eyes of 28 subjects (Exp Eye Res 2012). In this dataset, refraction ranged from -8.75 to +6.25
D, with a mean value (standard deviation) of -0.4 (3.3) D. Optic disc inclination ranged from
0.8 to 10.6 deg, with a mean value of 5.1 (2.4) deg. In total 625 retinal nerve fiber bundle
trajectories were traced. For all individual trajectories, the departure from the model was
calculated. Subsequently, we calculated, per subject, a 'mean departure' for the superior-
temporal and inferior-temporal regions. To quantify the retinal vessel course, we measured the
vertical distance between the fovea and the superior-temporal and inferior-temporal veins
(venous arcades). These distances were normalized to the distance between fovea and optic
nerve head center. Multiple linear regression analysis was performed for the superior-temporal
and inferior-temporal regions separately, with the mean departure as dependent variable and
refraction, optic disc inclination and retinal vessel course as independent variables.
Results: In the superior-temporal region, refraction was significantly associated with mean
departure (P=0.03) but optic disc inclination (P=0.13) and retinal vessel course (P=0.28) were
not. The addition of the retinal vessel course to the regression analysis increased the explained
variance from 0.26 to 0.29. In the inferior-temporal region, non of the independent variables
were significantly associated with mean departure (P=0.20, 0.87 and 0.13 for refraction, optic
19
disc inclination and retinal vessel course, respectively). The addition of the retinal vessel course
to the regression analysis increased the explained variance from 0.07 to 0.16.
Conclusions: The wiring of the human retina displays a considerable variability. Refraction,
optic disc inclination and the position of the venous arcades explain some of this variability.
2u

Optimizing the correspondence between visual field defects and
mathematical models of retinal nerve fiber morphometry
N.S. Erler
1
, S.R. Bryan
1
, P.H.C. Eilers
2
, E.M.E.H. Lesaffre
2
, H.G. Lemij
3
, K.A. Vermeer
1

1
Rotterdam Ophthalmic Institute, Rotterdam Eye Hospital, Rotterdam, The Netherlands
2
Erasmus Medical Center, Rotterdam, The Netherlands
3
Glaucoma Service, Rotterdam Eye Hospital, Rotterdam, The Netherlands

Purpose:
Visual fields, as determined by automated perimetry, are notoriously noisy. The purpose of this
study was to evaluate the correspondence of visual field defects with a model of the retinal
nerve fiber layers morphology and to develop a way to adjust the model parameters for optimal
correspondence. Visual fields can then be smoothed according to the nerve fiber bundles'
orientation for noise reduction.

Figure 1: Correlation vs. distance, for the conventional visual fields
We applied a model that describes the paths of retinal nerve fibers (Airaksinen et al, 2008) to
the 24-2 threshold test locations. This model associates every test location with the angle of the
nerve fiber at the optic nerve head and the nerve fiber's length. The model's two shape
parameters (A and B) steer the spatial distribution of nerve fibers.
The correlation coefficient between every combination of two visual field locations were
calculated from threshold deviation values from 104 visual fields with moderate glaucoma (MD
between -12 dB and -6 dB) from an ongoing study at the Rotterdam Eye Hospital.
Figure 2: Correlation vs. distance, for the nerve fibre model with A=0.8 and B=0.015

Results:
A scatter plot of the correlation coefficients versus the Euclidean distance in the conventional
visual field is shown in Figure 1. The correlation is very inconsistent, i.e., for the same distance
between two locations the correlation coefficient varies considerably. This is expressed by the
large root-mean-squared error (RMSE) which is determined from the residuals.
A similar plot is shown in Figure 2 for distances in nerve fiber angles, with shape parameters
A=0.8 and B=0.015. The consistency of correlation coefficients is much smaller than for the
Euclidean distance. Optimal consistency (minimal RMSE) was found for A=0.4 and B=0.03, as
shown in Figure 3.

Conclusion:
We developed a method to optimize the shape parameters of a mathematical nerve fiber model
to visual field defects. The optimal model considerably improves the consistency of the
correlation between visual field locations. The estimated optimum is in agreement with the
parameter values that were determined by Airaksinen et al based on fundus photos.

21

Advancements in high resolution retinal microangiography and blood flow assessment
using Doppler OCT

R.A. Leitgeb,
Center for Med Physics & Biomed Engineering,
Medical University Vienna, Vienna, Austria.

Purpose: To assess the blood flow dynamics quantitatively and to contrast the microvasculature
network label free of the posterior segment in healthy and diseased human subjects using a
high-speed Doppler OCT platform.

Methods: Traditional Doppler OCT is highly sensitive to motion artifacts due to the
dependence on the Doppler angle, limiting its accuracy in clinical practice. To overcome this
limitation, we use a bidirectional dual beam platform with flexible aligning of the incidence
plane allowing reconstruction of the true flow velocity in the range of 5 to 500mm/s. The angle
independent quantitative flow dynamics are extracted from specific vessel cross-sections of
arteries and veins, acquired from circumpapillary and segment DOCT scan series over time at
100kA-scans/s. Microangiography is used to assess the vascular network branching and
integrity. Contrast of flow towards static tissue is achieved by using an intensity difference
algorithm between successive B-scans. Novel High Speed OCT platforms operating in the MHz
A-scan rate range allow for 45x45 degree wide field imaging of retinal and choroidal
microvasculature.

Results: The quantitative analysis profits from the intrinsic stability with respect to motion. The
use of 1060nm center wavelength shows enhanced penetration into choroidal structures. The
flow values in retinal vessels of healthy subjects of 20-50mm/s and 10-20mm/s for arteries and
veins respectively fit well with previous findings. Highly sensitive flow contrasting has been
performed on healthy subjects over a wide field of view of 45degree using novel laser
technology. The images show impressive contrast down to individual capillaries in both retina
and choroid. Choroidal imaging is optimal due to the better penetration at the operating
wavelength of 1060nm.

Conclusions: We apply a dual beam bidirectional Doppler OCT system to accurately quantify
flow dynamics together with high speed OCT for non-invasive 3D microangiography of the
posterior segment in the human eye with unpredecented filed of view and vascular contrast. The
measurements performed in healthy subjects allow deriving characteristic vascular patterns of
the pathological tissue. The motion artifact stable and accurate flow quantification and
visualization may therefore lead ultimately to a better understanding and an enhanced early
diagnosis of major retinal diseases.

I acknowledge my group members Cedric Blatter, Branislaw Grajcar, Amardeep Singh, Tilman
Schmoll, and Severine Couquoz, collaboration with Dr. Christoph Mitsch, Dr. Sonja Prager,
and Prof. Ursula Schmidt-Erfurth from the Department of Ophthalmology and Optometry at the
Medical University Vienna, Dr. Robert Huber from the LMU Munich, as well as financial
support from European Commision (FP7-HEALTH FUN-OCT) and the Austrian Christian
Doppler Association.

22

L ReLoaded: A Metamorphosis and Transfiguration.
Application to Data from the Heidelberg Retina Tomograph.



Martin Diller, Heidelberg; Paul Artes, Halifax

Anything that can be said can be said clearly. What cannot be said, thereof one must be silent.
L Wittgenstein (1889 1951). www.youtube.com/watch?v=57PWqFowq-4

God is in the details.
L Mies van der Rohe (1886 1969)

Ophthalmic imaging technologies yield complex data data that are
difficult to interpret without more or less sophisticated statistical tools.
Yet clinicians, technicians, and patients often do not have a
background in statistics or probability theory, and thus there is a
danger that results are misinterpreted. L is our yet incomplete attempt
to find a generalisable solution to this dilemma, applied to data from
the HRT.
L aims to translate progression data from series of HRT images a) into
a simple graphical representation, and b) into clear, jargon-free yet
technically accurate, verbal statements that provide clinically relevant
guidance to non-experts. L makes statements on the most important
dimensions (speed, clinical and statistical significance of change, and
the power to detect it; validity of linearity assumptions made by the
analysis; image quality). Each dimension is quantified by an appropriate
combination of indices (for example, image quality ~ mean pixel height
standard deviation [MPHSD] + uniformity of illumination + +
etc.), providing a hierarchy of statements that users are invited but not
forced to explore. These dimensions are scaled into an [0, 100] interval
by comparison with reference quantiles from a previously published
population (Reis et al, Ophthalmology 2012).
We look forward to discussing the L approach with the critical
audience of IMAGE.


2S
Confocal Laser Scanning Tomography to predict Visual Field Conversion
in Glaucoma Suspects and Ocular Hypertension
Schrems-Hoesl LM*, Schrems WA*, Horn FK, Mardin CY, Kruse FE,
Juenemann AG, Laemmer R.
Department of Ophthalmology, University of Erlangen-Nuremberg, Erlangen,
Germany
*These authors contributed equally to this work.

Purpose: To compare Moorfields regression analysis (MRA), Glaucoma
probability score (GPS) and different discriminant functions to predict
future visual field conversion of glaucoma suspects and ocular
hypertensive subjects.
Methods: 120 patients with ocular hypertension and 110 patients with
suspected glaucoma were recruited from the Erlangen glaucoma registry for this
study. Annually, all patients underwent standard automated perimetry, 24-hour
intraocular pressure profile, optic disc photography and HRT (Heidelberg Retina
Tomograph; Heidelberg Engineering) measurements. One randomly selected
eye of each patient was used in the study. The cohort was divided into two
groups based on the development of repeatable glaucomatous visual
fields during follow-up. Positive (PPV) and negative predictive values
(NPV) were compared for MRA, GPS and the classification of Bathija,
Mikelberg and Mardin at baseline. Kaplan-Meier Survival curves and
logrank tests were used to evaluate equality of survival distributions for
different test results.
Results: Mean follow-up time was 9.47 +/- 2.81 years (minimum = 3.42 years;
maximum = 14.33 years). 26 eyes (11.3 %) demonstrated glaucomatous visual
field loss in the follow-up period. MRA temporal-superior and temporal-
inferior outside normal limits were predictive of future visual field loss with
PPVs of 33.3% and 28.6%. Normal GPS Temporal Sector demonstrated a
NPV of 96.4% and normal results in discriminant functions between 94.7%
and 95.5%.
Conclusions: HRT is an useful tool to predict future visual field conversion.
Development of visual field defects in 10 years is highly unlikely, if GPS
classification and/ or Classification of discriminant analysis at
baseline were normal. MRA temporal-superior and temporal-inferior
outside normal limits are associated with future VF conversion.


24
Additional topics
Mystery cases In previous years, it was interesting to discuss mysterious features
that we had come across in various images. Please bring along and present any of
those mystery cases. Maybe we can come up with reasonable solutions, or good
suggestions to further explore the mysteries.

OCT standards The differences between the various OCT imaging devices may be
large and subject to continual change. This limits comparative studies across
instruments as well as longitudinal studies with devices that undergo repeated
upgrades. It would therefore be useful to set specific standards; one could think of
scanning protocols, (pre)processing procedures, data display standards, selection of
normative databases, etc., etc. Are we ready for setting any such standards? Will
those standards hamper or promote the further defvelopment of OCT technology in
glaucoma management?

Next IMAGE meeting The turn-out and number of abstracts have been lower this year
than in previous years. Is this due to timing, the venue, bad luck, or to other, yet
unknown factors? How does this impact on setting the next date and venue for future
IMAGE meetings? What can we learn, and should we do, to promote future meetings?
Where and when will we have IMAGE 2014? Any candidate cities?

S-ar putea să vă placă și