La cytologie digitalisée: avantages et limites. · comprendre les developments en pathologie...

La cytologie digitalisée: avantages et limites.

R. Dina, MD, FIAC, FRCPathImperial College London, UK

COI

▪Declaration d’interet

• Je suis un simple cytopathologiste qui essaie de

comprendre les developments en pathologie digitale

et moleculaire

• Ma femme ne crois pas que j’aille une qualilification

medicale

• En effet, même si instruit je ne trouve rien dans le

frigidaire.

• Je n’ai acun interêt commercial

Avantages▪ (1) Diagnostic primaire (télécytologie)

▪ (2) Consultation specialiste à distance

▪ (3) Activité éducative dans la même institution ou à distance

▪ (4) Archivage des cas intéressants et medico-légaux (réplication de lames de cytologienumérique)

▪ (5) assurance de la qualité

▪ (6) Conférences éducatives telles que des panneaux de tumeur (localement ou àdistance)

▪ (7) Test de compétence cytologique

▪ (8) Examen ou certification à distance

▪ (9) Analyse d'image détaillée et cytomorphométrie

▪ (10) Annotation électronique de diverses entités sur les lames à des fins d'enseignement

▪ (11) Acquisition facile d'images statiques à partir d'images de la lame entière

▪ (12) Fournir des services de cytopathologie aux hôpitaux sans cytopathologie

▪ (13) Accèder à l'expertise en cytologie à distance

▪ (14) Évaluation et triage à distance des lames (ROSE)

▪ (15) consultation synchrone

Limitations

▪ (1) Coûteux: configuration initiale et memoire▪ (2) Fonctions de mise au feu limitée au moment▪ (3) temps de digitalisation▪ (4) Stockage: grande taille de fichier▪ (5) Exigences de formation▪ (6) études de validation limitées▪ (7) Absence de normalisation: fournisseurs multiples,

logiciels et manque d'interopérabilité▪ (8) Soutien à l'infrastructure informatique (limitation

de la bande passante des réseaux)▪ (9) réticence des professionnels à adopter

Patholog Res Int. 2011; 2011: 264683.

Walid E. Khalbuss, 1, 2 * Liron Pantanowitz, 1, 2 and Anil V. Parwani 1

1Division of Pathology Informatics, Department of Pathology, University of Pittsburgh Medical Center, Pittsburgh,

PA 15232, USA2Division of Cytology, UPMC Shadyside Hospital, 5150 Centre Avenue, POB2, Suite 201, Pittsburgh, PA 15232,

USA

Review Article:Screening for Cervical Cancer Using Automated Analysis of PAP-SmearsEwert Bengtsson and Patrik Malm

Division of Visual Information and Interaction, Department of Information Technology, Centre for Image Analysis,Uppsala University, Box 337, 751 05 Uppsala, SwedenComputational and Mathematical Methods in MedicineVolume 2014

Development of automated screening devices

Type Tools Issues

Cytoanalyzer 1956 2D histograms of nuclear size vs density

Too many false positives

CYBEST 1974 Nuclear area, density, cytoplasmic area, N/C ratio

Could not measure nuclearshape and chromatin pattern

PAPNET (NSI) 1994 first to introduce interaction into automated screening first by an algorithmic classifier and then by a neural network classifier

centralization of scanning, litigation issues,

LBC (Cytic,AutoCyte) early90s

Liquid based cytology samples facilitate digital scanning

More expensive

Development of automated screening devices

Tools Issues

AutoPap 300-Tripath-BDFirst FDA approved 1998

25% of the slides as normal for no further review;theother 75% are ranked into five categories at risk forabnormality.

Very expensive; could also allow review of fieldsMedico-legal issues

Cytyc- ThinPrep Imaging System 2003 -from 2007 Hologic

Increasing detectionof abnormalities by improved specimen preparation andscreening both visually and bymachine

Expensive- Not clearly shown to improve detection in a screening setting

Use of Digital Imaging

At Imperial College we have been using in the past digital images for

▪ cytology tests during our MSc in Cytopathology▪ Cytology mock exams during our Advanced Courses

which prepare for the MRCPath examination ▪ WSI for research purposes and testing.▪ We currently informally review WSI of cases but we

do not issue a formal report on them. This is because although technology on the different platforms available on the market has markedly improved we do not feel that there is an agreed standardised practice for it.

Conclusions

▪ WSI is here to stay and is fast improving and getting cheaper

▪ It is an important teaching and training tool

▪ It is used for EQA schemes and Quality Assurance

▪ It is used in MDT meetings (Tumour Boards)

▪ It helps retaining a screening component to all assessment tests

▪ BUT…… it is one of the many tools!

1) The research has used the Google search engine: www.google.com;

2) Searched nouns as keyword: nouns had to be the most concise as possible. The used keywords are: cytology web sites, cytology atlas, cytology and cytopathology journal, and cytology societies;

How many web sites use digital cytology?

1) Sponsor, scientific society, personal web page, academic institution or commercial site: whether a website is sponsored by a Society, a particular product or interest group, the owner of the web site. Personal web page web sites can list the author of the information and biographical information.

2) Society: the name of the involved Society.

3) Purpose: to provide educational information, professional advice, promoting the profession of cytologists, encouraging the science of cytology. Many web sites provide information on topics of interest to the owner, as well as tutorials or opinions.

4) Topic: FNA, gynaecologic or non-gynaecology cytology.

5) Target groups: whether the web site is recommended to cytologists, cytotechnologists, cytology trainees or students, laboratory personnel.

Criteria

6) Access: public, only registered members, any payment fees required.

7) Educational resources: each web site has been checked whether with or without educational purpose or to improve academic success.

8) Imaging: static or dynamic as virtual slides.

9) Passive or interactive: some web sites have just slides to look but no possibility to have an interactive approach. Other web sites allow the visitors to take quizzes or view solutions previously hidden, in order to test trainees or students.

Criteria

▪ The number of web sites is about 671,000 results for each keyword. Sites with only histopathology have been excluded.

▪ Based on the above mentioned criteria, the number of web sites considered adequate is 31.

Results

▪ There are numerous web sites available

▪ Aims are different

▪ Few are available in multiple languages

▪ Cytology is notoriously more difficult to comprehensively scan

▪ Too few web sites are completely free to use

▪ Few offer interactive e-training

▪ However it is getting better all the time!

Conclusions

Average monthly traffic 2009-14

0

10000

20000

30000

40000

50000

60000

year 2009 year 2010 year 2011 year 2012 year 2013 year 2014

average visits

average page downloads

Observations after 5 years of use▪ The platform is used more during the

academic months

▪ Page download is higher in the academic months

▪ After 5 years the content and IT architecture is obsolete

▪ This is reflected by the decrease in page downloads

Eurocytology 2013-15

▪ Successful in applying for another EU grant▪ Move platform from static to dynamic architecture

▪ EFCS is a partner (already linked to website)▪ Linked to IAC website▪ Translate into Turkish, Czech, Portuguese and German

▪ Incorporate translations and adaptations into national curricula (Turkey, Czech Republic)

▪ Increase interactivity by devising online assessment and testing tools

▪ Maintain an editorial committee

▪ Validate it as a European Accreditation Scheme Platform Through QUATE

What is peculiar to this project?

▪ Transnationality

▪ Validation through a European Network

▪ Dinamic tests through the use of a specific software and image database

▪ Use of virtual slides

▪ Translated in several languages

AnalyticsEurocytology 2.0

All Web SiteDataGo to report

Audience Overview

Mar 1, 2018 - Feb 28,2019

Overview

2

50,000

Users

100,000

… May

Users

018 July 2018 September 2018 November 2018 January 2019

New Visitor Returning Visitor

New Users Sessions

771,338 756,828 919,524 11%

Number ofSessions per User

1.19

Pageviews

1,254,510

Pages / Session

1.36

89%

Avg. Session Duration

00:01:07

Bounce Rate

85.94%

Language Users %Users

1. pt-br 126,478 16.56%

2. es-es 98,820 12.94%

3. en-us 93,158 12.20%

4. es-419 69,146 9.05%

5. es 30,262 3.96%

6. el-gr 29,190 3.82%

7. tr-tr 26,793 3.51%

8. it-it 26,700 3.50%

9. en-gb 25,437 3.33%

10. es-us 22,556 2.95%

© 2019 Google

All Users100.00%Users

AnalyticsEurocytology 2.0

All Web SiteDataGo to report

Rows 1 -10 of212

Location

Mar 1, 2018 - Feb 28,2019

Map Overlay

Summary

Country

Acquisition Behavior Conversions

Users New Users Sessions Bounce Rate Pages / Session Avg. Session Duration Goal Conversion Rate GoalCompletions GoalValue

771,338

% of Total:

100.00%

(771,338)

757,234

% of Total:

100.05%

(756,828)

919,524

% of Total:

100.00%

(919,524)

85.94%

Avg forView:

85.94%

(0.00%)

1.36

Avg forView:

1.36

(0.00%)

00:01:07

Avg forView:

00:01:07

(0.00%)

0.00%

Avg forView:

0.00%

(0.00%)

0

% of Total:

0.00%

(0)

$0.00

% of Total:

0.00%

($0.00)

1. Brazil 134,803

(17.69%)

134,619

(17.78%)

159,697

(17.37%)

87.35% 1.22 00:01:03 0.00% 0

(0.00%)

$0.00

(0.00%)

2. Mexico 68,192

(8.95%)

67,615

(8.93%)

79,743

(8.67%)

88.06% 1.24 00:00:58 0.00% 0

(0.00%)

$0.00

(0.00%)

3. Greece 49,675

(6.52%)

49,352

(6.52%)

61,564

(6.70%)

80.90% 1.40 00:01:22 0.00% 0

(0.00%)

$0.00

(0.00%)

4. Colombia 46,083

(6.05%)

45,799

(6.05%)

53,007

(5.76%)

89.64% 1.18 00:00:46 0.00% 0

(0.00%)

$0.00

(0.00%)

5. Italy 43,123

(5.66%)

43,032

(5.68%)

53,323

(5.80%)

86.56% 1.39 00:01:06 0.00% 0

(0.00%)

$0.00

(0.00%)

6. Turkey 40,455

(5.31%)

40,235

(5.31%)

46,076

(5.01%)

88.58% 1.20 00:00:41 0.00% 0

(0.00%)

$0.00

(0.00%)

7. Spain 38,050

(4.99%)

37,773

(4.99%)

48,677

(5.29%)

84.01% 1.60 00:01:23 0.00% 0

(0.00%)

$0.00

(0.00%)

8. Romania 31,471

(4.13%)

31,267

(4.13%)

40,468

(4.40%)

86.94% 1.27 00:01:06 0.00% 0

(0.00%)

$0.00

(0.00%)

9. United States 27,833

(3.65%)

27,349

(3.61%)

30,718

(3.34%)

88.02% 1.37 00:00:46 0.00% 0

(0.00%)

$0.00

(0.00%)

10. Poland 27,548

(3.62%)

27,154

(3.59%)

32,348

(3.52%)

86.27% 1.39 00:00:49 0.00% 0

(0.00%)

$0.00

(0.00%)

1 134,803

© 2019 Google

All Users100.00%Users

Digital Histology vs digital Cytology

▪ Digital Histology and digital Cytology need a different technical approach for many reasons

Digital Histology vs digital Cytology▪ Dimension:

Digital Histology vs digital Cytology▪ Dimension:

Digital Histology vs digital Cytology▪ The nature of the material is different:

Histology

Cytology

Digital Histology vs digital Cytology▪ The scanning technique is different:

If the scanner autofocus works well, a single layer virtual slide allows a high quality screen of a histological preparation.

A multi level scanning is compulsory to get an acceptable cytological virtual slide.

Digital Cytology: a 3D problemApparently it is like a short blanket:

Low number of levels = not acceptable focusing wideness.

Short distance between levels = increased number of levels.

Many levels with short distance between them = very big files.

In any case: sensibly longer scanning time.

Digital Cytology: a 3D problemThere are at least 3 possible ways

to try to find a solution.

Digital Cytology: a 3D problemThe first is the one used in the virtual slides library of some teaching web sites.

For example:

www.eurocytology.eu

www.cytology.cloud/gk

Digital Cytology: a 3D problemThis technique consists 3 steps:

1. dividing in small areas (tiles) the image resulting from the scanning of each level

1. taking the best-focused tile from each layer

1. building a new virtual slide where all the objects result in focus

Leve

l 1

Leve

l 2

Leve

l 3

New single level

image

Digital Cytology: a 3D problemThe final result is a single level virtual slides where all the tiles are perfectly in focus.

Pros: - small dimension of the file

- good “visual” results

Cons: - long processing time

- a lot of unnecessary data generated

Digital Cytology: a 3D problem

Digital Cytology: a 3D problem

A second interesting method is proposed in

Digital Cytology: a 3D problem

A specific software generates during the scanning a three

dimensional focus map of the cells in the slide.

Following this map the scanner takes only the images of the cells

avoiding the generation of unnecessary and unwanted data.

Digital Cytology: a 3D problem

A third interesting solution is based on the

“tilted focus plane” and/or “tilted lens”.

Digital Cytology: a 3D problemThe idea to tilt the camera lens or the focus plane is quite old

Digital Cytology: a 3D problem

The idea to tilt the camera lens or the focus plane is quite old

Digital Cytology: a 3D problem

Tilting of the focus plane or the lens

this kind of camera can take sharp pictures

of objects located at different distances

from the camera without the need to close

the camera iris to increase the depth of focus

Digital Cytology: a 3D problem

Digital Cytology: a 3D problem

Digital Cytology: a 3D problem

By means of a scanner equipped with tilted lens or focus plane it is possible to scan with a single pass all the cells in the thickness of the smear saving time and disk space (with the appropriate software)

Summarizing

If time and disk space are not a problem any

method suits well.

This is true for teaching, research, documentation

and second opinion.

For real primary diagnostic purposes new

technologies must be developed.

Thank you!