USE OF THE KONAN NONCON-USE OF THE KONAN NONCON-ROBO SPECULAR MICROSCOPE ROBO SPECULAR MICROSCOPE

IN CLINICAL RESEARCHIN CLINICAL RESEARCH

Henry F. Edelhauser, Ph.D.Ramzy G. Azar, MPH

Emory University Eye CenterAtlanta, Georgia

PurposePurpose

Understand variability issues with Understand variability issues with specular microscopy that may bias specular microscopy that may bias resultsresults

ObjectivesObjectives

Provide examples of good and poor Provide examples of good and poor photographyphotography

Illustrate variability in specular Illustrate variability in specular microscopy photography and microscopy photography and analysisanalysis

Illustrate variability within a single Illustrate variability within a single imageimage

What is a Good Image?What is a Good Image? Distinct cellsDistinct cells Can identify at least Can identify at least

150 cells150 cells Cells can be Cells can be

grouped in a grouped in a uniform areauniform area

What may be good What may be good for clinical purposes for clinical purposes may not be good may not be good researchresearch

Things to Consider That Things to Consider That Affect Quality of ImageAffect Quality of Image

Dry eyeDry eye Contact lens useContact lens use Wrong Specular Manual SettingsWrong Specular Manual Settings KeratoconusKeratoconus Patient CompliancePatient Compliance AgeAge Training, experience of Training, experience of

photographerphotographer

Poor Quality ImagesPoor Quality Images

Poor Quality Images Poor Quality Images Continued…Continued…

Poor Quality Images Poor Quality Images Continued…Continued…

Conditions that Potentially Conditions that Potentially Increase VariabilityIncrease Variability

Guttata (Fuch’s dystrophy)Guttata (Fuch’s dystrophy) Polymegethism/PleomorphismPolymegethism/Pleomorphism InjuryInjury Low Cell density (Huge cells)Low Cell density (Huge cells)

Guttata (Fuch’s dystrophy)Guttata (Fuch’s dystrophy)

Capturing the Best Capturing the Best Image PossibleImage Possible

Make sure Pt is comfortableMake sure Pt is comfortable Instruct Pt to blinkInstruct Pt to blink Instruct Pt not to move and to open eyes Instruct Pt not to move and to open eyes

widewide Instruct Pt to focus on the green lightInstruct Pt to focus on the green light Be patientBe patient Use Manual setting to improve quality when Use Manual setting to improve quality when

cornea is unusually thicker than normalcornea is unusually thicker than normal

Things to Consider When Things to Consider When Analyzing ImagesAnalyzing Images

Locate the best and most representative Locate the best and most representative areaarea– Number of cellsNumber of cells– Quality of CellsQuality of Cells

No shadowsNo shadows DiseaseDisease

– Use area with the fewest distortionsUse area with the fewest distortions BlurringBlurring Washed-out imagesWashed-out images ShadowsShadows

Locating the Best Analysis Locating the Best Analysis Area (Sample Images)Area (Sample Images)

Dotting CellsDotting Cells

Dot all Cells at the CenterDot all Cells at the Center– Remain accurate and consistent Remain accurate and consistent

throughoutthroughout Dot 150 cellsDot 150 cells Grouping is importantGrouping is important

Where to Group the Where to Group the Analysis?Analysis?

What is Wrong With This What is Wrong With This Analysis?Analysis?

Analysis is not Analysis is not representative representative

Introducing Introducing BiasBias

Not likely to Not likely to repeatrepeat

Not enough Not enough cells countedcells counted

Grouping DetailsGrouping Details

EasyEasy ClearClear No No

shadowshadowss

Dot 150 Dot 150 cellscells

Need Need Good rep.Good rep.

Take Take more more timetime

Dot > 150Dot > 150

NormalNormal PolymegathismPolymegathism

Grouping your AnalysisGrouping your Analysis

Correct GroupingCorrect Grouping– ConcentricConcentric– EvenEven– UniformUniform

Incorrect GroupingIncorrect Grouping– linearlinear– unevenuneven– WindingWinding

Cell Grouping - GuttataCell Grouping - Guttata

Group only in one areaGroup only in one area

To Analyze the Cells:To Analyze the Cells: You need to be able to visualize You need to be able to visualize

cellscells Find a patternFind a pattern IdentifyingIdentifying

– Cells vsCells vs– Damage vsDamage vs– ShadowsShadows

Where an Image is Where an Image is Analyzed Can Create Analyzed Can Create

VariabilityVariability

Examples of VariabilityExamples of Variability

CD = 2873CD = 2873

SD = 170SD = 170

CV = 48CV = 48

6A = 536A = 53

CD = 2976CD = 2976

SD = 113SD = 113

CV = 33CV = 33

6A = 536A = 53

ΔCD = 103

(4%)

Examples of Variability Examples of Variability Within Readers Within Readers

CD = 2531CD = 2531

SD = 139SD = 139

CV = 35CV = 35

6A = 556A = 55

CD = 2358CD = 2358

SD = 222SD = 222

CV = 52CV = 52

6A = 566A = 56

ΔCD = 173

(7%)

Examples of Variability Examples of Variability Between ReadersBetween Readers

AnalysisAnalysisRepeated Repeated

4x4x#1 - 2631#1 - 2631

#2 - 2557#2 - 2557

#3 - 2531#3 - 2531

#4 - 2570#4 - 2570

#5 - 2624#5 - 2624

Range 2531 - 2631Range 2531 - 2631

Consequences of Under or Consequences of Under or Over CountingOver Counting

Analysis N p -valueMean SD (Cells) %

Regular 10 3233 24SKIP

One 8 3213 49 0.3174 -20 -0.6Two* 8 3188 20 0.0006 -45 -1.4Three 4 3137 9 <0.0001 -96 -3.0Four 4 3091 30 <0.0001 -142 -4.4Five 4 3069 25 <0.0001 -164 -5.1

COUNTED TWICE

One 10 3240 30 0.5741 7 0.2Two 7 3255 19 0.0613 22 0.7

Three* 6 3257 13 0.0429 24 0.7Four 6 3265 35 0.0457 32 1.0Five 6 3281 36 0.0062 48 1.5

Study by Nauman Rashid

CD (Cells/mm2) CD

Endothelial Cell DensityEndothelial Cell Density

0

1000

2000

3000

4000

5000

10-19 20-29 30-39 40-49 50-59 60-69 70-79 80-89

Age (years)

Cel

l Den

sity

(cel

ls/m

m2 )

non-contact ROBO

contact (Yee et al.)

Asian population

pre-LASIK (Jones et al.)

Japanese population (Matsuda et al.)

Precision of 36 Robo corneal endothelial Precision of 36 Robo corneal endothelial specular images of each eye (OD, OS) specular images of each eye (OD, OS)

taken on 18 different days and analyzed taken on 18 different days and analyzed with the Robo softwarewith the Robo software

NN Cell DensityCell Density PrecisioPrecisionn

ODOD 1818 2545 ± 45 2545 ± 45 cells/mmcells/mm22 (1.7%)(1.7%)

OSOS 1818 2600 ± 41 2600 ± 41 cells/mmcells/mm22 (1.5%)(1.5%)

(From AJO 125:465-471, 1998 LASIK Paper)

Age Dependent Cell Density Age Dependent Cell Density Variation Within 3 Different Variation Within 3 Different

Corneal RegionsCorneal Regions

y = -18x + 3454

y = -14x + 3622

y = -13x + 4310

1000

2000

3000

4000

5000

6000

10 20 30 40 50 60 70 80

Age (Years)

Cel

l Den

sity

(cel

ls/m

m2)

Central

Paracentral

Limbal

(10)(11)

(9)

(9)

(14)

(11) (9)

Sources of Variability Sources of Variability SummarySummary

Difficult to return to same location Difficult to return to same location (1 mm = (1 mm = 56 cells/mm 56 cells/mm2 2 - - 2.0%2.0%))

Poor image quality (minimal # of Poor image quality (minimal # of analyzable cells = analyzable cells = 100100) )

Technician error Technician error (Training/consistency)(Training/consistency)

Reader analysis (Training/consistency) Reader analysis (Training/consistency) Equipment calibration/alignmentEquipment calibration/alignment

Data Flow ChartData Flow Chart

Reading Reading CenterCenter Data Process Data Process

CenterCenter

Specular Specular SitesSites

Technology CoTechnology Co