FGAI4H-F-012-A01-R1 Zanzibar, 3-5 September 2019 - PowerPoint Presentation

Slide1

FGAI4H-F-012-A01-R1

Zanzibar, 3-5 September 2019

Source:TG-Ophthalmo topic driverTitle:TDD update: TG-Ophthalmo (Ophthalmology)Purpose:Discussion

Contact:Arun ShroffE-mail: arun@xtend.ai

Abstract:

This PPT summarizes the content of

F-012

with the TDD for the

TG on ophthalmology

, for presentation and discussion during the meeting.

Meeting F Topic Group UpdateOphthalmology (TG-Ophthalmo

)Zanzibar, Sep 3 – 5, 2019Arun Shroff,

Topic Driver, TG-Ophthalmology

Topic Group – Ophthalmology

Topics in this group:

Diabetic Retinopathy (DR) Age Related Macular Degeneration (AMD)Glaucoma (GC) Pathological Myopia (PM)Topic Group Description Document (FGAI4H-F-012) Topic Group Call for Participation (FGAI4H-F-005-A07)

The Health Challenge

Diabetic Retinopathy (DR)

At risk population - 422M people with diabetes worldwide (2014)35%, 148M have DR / 11%, 48M have Vision Threatening DR (64M by 2040) Leading cause of blindness among adults worldwideAge Related Macular Degeneration (AMD) Damages macula and impairs central vision 196M by 2020Third leading cause of vision loss overall, leading cause for those over 50

The Health Challenge

Glaucoma (GC)

Damages optic nerve & leads to vision loss 80M by 2020Pathological Myopia (PM) Global Prevalence is 0.9% to 3.1%35% of people with myopia have High Myopia, which can develop into PMNeed for AI: Acute shortage of specialists globally to screen everyone at risk – specially in LMICs.

Benchmarking: DR Classifications

Multi-class Classification: [0 (Nongradable Image) ] 1(No DR) 2 (Mild) 3 (Moderate NPDR) 4 (Severe NPDR) 5 (PDR) Binary : [0 (Nogradable Image)] 1 (Nonreferable Retinopathy = No DR or Mild) 2 (Referable Retinopathy = Moderate, Severe, PDR)

Benchmarking: AMD, GC, PM Classifications

AMD:

[0 (Image Nongradable)]1 (No/early stage AMD 2 (Intermediate/advanced stage AMD) GC: [0 (Image Nongradable.]1 (No GC)2 (GC)PM: [0 (Image Nongradable)]1 (No PM/HM)2 (HM: high myopia)3 (PM)

Available Public Datasets - DR

EyePACS

dataset: Approx 90,000 fundus images, 5 levels of severity Kaggle: (derived from EyePACS)Approx 35,000 images : 5 levels of severity MESSIDOR dataset:

1,200 images, 4 levels of severity DiaRetDB dataset: ~ 200 images marked with lesions etc

Available Public Datasets - AMD, GC

AMD:

AREDS dataset: (AgeRelated Eye Disease Study )Images from ~4700 patients : KORA dataset: (Cooperative Health Research in the Region of Augsburg (KORA) dataset,)Approx

2840 patient recordsGC: ORIGA, 650 fundus imagesRetinal fundus images for glaucoma analysis (RIGA, 760 images) ACHIKO-K (258 images)DRISHTI-GS (100 images)

Benchmarking Metrics

Sensitivity:

% of positive (disease) cases correctly classified True Positive/(True Positive + False Negative)Specificity: % of negative (normal) cases correctly classified True Negative/(True Negative + False Positive)AUC (Area Under ROC);Sensitivity Vs (1-Specificity) plotted at different points of the model

Other Metrics: Precision/Accuracy, F1 Score, Confusion Matrix

Use-Case & Topic Group History

Meeting B - New York, 15-16 November 2018AI for Ophthalmology Use case submitted in response to the Call for Proposals

“Using AI for Early Detection of DR to Prevent Vision Loss” accepted as a use caseMeeting C - Lausanne, Switzerland, 22-25 January 2019Status report on the use case “Using AI for Early Detection of DR”Topic Group “Ophthalmology” established 2 Members : Medindia.net / Xtend.ai Baidu, China.

Use-Case & Topic Group History

Meeting D- Shanghai, April 2-5, 2019Topic Description Document (TDD) – Version 1 completed Topic Group Status Update

Meeting E - Geneva, May 30 – June 1, 2019Topic Description Document (TDD) Updated Edits / Corrections madePathological Myopia (PM) added (by Xingxing Cao, Baidu) Reviewed and validated by topic group membersNew topic group members: Ashley Kras, M.D. M. S., Ophthalmologist & Bioinformatician

Progress Since Meeting E

Call For Participation: Outreach via email / social mediaSeveral inbound emails with interest in joining/contributing to group

New Topic group members:Dr Covadonga Bascaran, PHEC MSc Programme Director, International Centre for Eye Health (ICEH), London School of Hygiene & Tropical Medicine Inês Sousa , Head of Intelligent Systems, Fraunhofer PortugalOnline Meetings/Calls: Prof Leo Celi, Clinical Research Director, Harvard MIT Division of Health Science and Technology and Ash Krasley: (June 22, 2019) Details about MIT Open Access ProjectPotential collaboration with FGAI4H / Contribution of DataDr. Jorge Cuadros, EyePACS (July 22, 2019) Database of over 5 million imagesDiscussion about contributing datasets (ongoing)

Progress Since Meeting E

Online Meetings/Calls: Topic Group Meeting, (Jul 31, 2019): Participants:

Dr. Covadonga Bascaran & Inês SousaDiscussion about DR-Net Possibility of getting undisclosed data sets for testingContribution of data from different countries to make data representativeImages are not currently annotated/labeled – this would need to be doneOther meetings/projects: Meeting in Geneva with ITU/WHO, Sanofi, Aivision.health(France), Minister of Health, Senegal Discussion about pilot project in Senegal - AI for DR detection, target start date Nov 2019

Topic Group – Ophthalmology Members

Arun Shroff,

Xtend.AI and Medindia.net Topic Driver for TG-OphthalmologyYanwu XU, Artificial Intelligence Innovation Business, Chief Scientist, Baidu, ChinaXingxing Cao, Artificial Intelligence Group, Baidu, China Jingyu WANG, Artificial Intelligence Group, Baidu, China Shan Xu, CAICT, China Ashley Kras, M.D. M. S., Ophthalmologist & Bioinformatician (Harvard Medical School)Covadonga Bascaran, PHEC MSc Programme Director, International Centre for Eye Health (ICEH), London School of Hygiene & Tropical Medicine, U.K. Inês Sousa, Head of Intelligent Systems, Fraunhofer Portugal AICOS

Next Steps

Topic Description Document Continue to improve TDD for accuracy and completenessTDD validation from experts

Call For ParticipationContinue outreach to increase members and get more experts involvedDatasets/Benchmarking: Follow-up with DR-Net, EyePACs, Moorefields, Open Eye & others for collaboration and procurement of undisclosed, labeled datasets for benchmarking. Start working on benchmarking process & protocols.

Thank you!