Topcon and Orbis partner to expand DR screenings in sub-Saharan Africa

Orbis International and Topcon Healthcare are collaborating with the Rwanda International Institute of Ophthalmology (RIIO) to provide expanded diabetic retinopathy (DR) screenings in the country.

Let’s start with these players.

Orbis is a global non-governmental organization (NGO) dedicated toward preventing and treating avoidable blindness via a network of corporate partners, NGOs, and local hospitals.

The charity’s influence currently extends across Africa, Asia, Latin America, and the Caribbean.

Topcon Healthcare, a division of the Japanese optical equipment manufacturer Topcon Corporation, is a global provider of medical devices and software solutions to the ophthalmic community.

Take note: This current collaboration isn’t the first for the company; in May 2024, it announced two new partnerships with:
- Microsoft Corporation
  - Purpose: To utilize their respective technologies and provide an artificial intelligence (AI)- and cloud-based healthcare platform.
- RetiSpec
  - Purpose: To accelerate the commercialization of RetiSpec’s diagnostic AI technology, designed to non-invasively detect early Alzheimer’s disease.

And RIIO?

Founded in 2011 and based in Kigali, the capital of Rwanda, the Rwanda International Institute of Ophthalmology (RIIO) is an eye care center and constituent institution of the College of Ophthalmology of Eastern, Central and Southern Africa (COECSA).

Its main purpose: Enhancing eye health education, providing technical advice, and supporting the delivery of ophthalmic services and promoting eye health within Rwanda.

And how is the eye care center making a difference?

Per RIIO, via four main objectives and activities it operates with in mind:

Implementing innovative eye care delivery projects based in equity and excellence
1. Programs targeting geographically, social, and economically-disadvantaged communities with limited access to care
Supporting academic and professional advancements in eye health
1. Training and producing the region's first internationally-certified eye technicians
2. Establishing a 4-year ophthalmology residency program
Influencing Rwanda and the region’s eye health policies via evidence and advocacy
1. Supporting global initiative Vision 2020: The Right to Sight
Leading regional developments via a network of innovative activities, collaborations, and partnerships
1. Our topic of focus.

Alrighty, now dive into this partnership!

Topcon has donated two robotic retinal cameras for use in DR screenings among diabetic patients:

NW400
- Fully-automated, non-mydriatic retinal camera for acquiring high-resolution color images of the retina and anterior segment
NW500
- Fully-automated, non-mydriatic retinal camera for acquiring sharp-quality, consistent fundus imaging with slit scan illumination and rolling shutter mechanism

How is this technology being utilized?

By using these fundus imaging cameras, clinicians can upload diabetic patients’ images to Orbis’s free, open-access, telemedicine and e-learning platform: Cybersight AI.

About this platform: Cybersight’s capabilities extend to offering:
- Virtual consultations from 100+ ophthalmic experts
- online, Expertly-curated courses with CME/CPD credit opportunities
- Live online training events led by international clinical expertS

And as a refresh: Cybersight is also being utilized in Orbis’s previously-announced partnership with Heidelberg Engineering to offer virtual vision services.

Gotcha. And its use in these DR screenings?

The AI tool performs an automated interpretation of the collected fundus images to detect and visualize DR.

Note: It can also detect glaucoma and macular diseases.

How fast can this be done?

In mere seconds, giving patients immediate results and enabling a referral for further evaluations or treatment, if needed.

Why this is key: Typically, patients may wait days or weeks to receive screening results from a human grader, resulting in a delay in treatment or other follow-up care that may result in preventable vision loss—particularly for those with treatable diseases.

So why focus on DR?

It all leads back to recent findings from Orbis’s Rwanda-focused research published in the British Journal of Ophthalmology that analyzed the usage of AI to screen for DR among a region of patients.

But first: With DR as the leading cause of vision loss among working-age individuals across the globe, the largest increase in the prevalence of DR has been projected in sub-Saharan Africa (Rwanda).

Now this research.

The details: A total of 827 diabetes (type 1 and type 2) patients (aged 18+; 59.6% females) were screened for DR via retinal imaging with an AI-based system at four diabetes clinics in Rwanda between March and June 2021.

About the AI system used: Per the study, the model was based on the Inception-ResNet-v2 Convolutional Neural Network (CNN) architecture trained to classify fundus photographs into one of five categories based on the International Clinical DR Disease Severity Scale
- Designed to detect optic nerve and macula anomalies outside of DR

And what were the findings?

A total of 33.2% of patients were referred for follow-up following screenings, with a 99.5% satisfaction rate associated with the AI-based system.

In fact: 63.7% of participants reported preferring AI over human grading.

Did they compare the AI grading to human grading?

They did. And the findings were in the AI favor, per the study.

The sensitivity of the AI for referable DR was 92% (Confidence interval [CI]: 0.863, 0.968), with a specificity of 85% (CI: 0.751, 0.882).

Plus: Of the participants referred by AI:
- 88% were for DR only
- 39.6% were for DR and an anomaly
- 23.6% were for an anomaly only
- 4.7% for other reasons

Further: Adherence to their referrals was the highest among those referred for DR (53.4%).

So what were the takeaways from this?

This was the first clinical study to assess the use of AI for DR screenings in Rwanda—described as a “low-middle income country” with an “under-studied setting.”

The investigators noted that the research demonstrated the “feasibility of and high satisfaction with AI-based screening for DR.”

Any specific reasons for this high satisfaction?

Indeed. The following factors were attributed:

Patients receiving the exam (screening) during their diabetes appointment
- In lieu of separate scheduling
Reduced number of appointments and travel time needed
- Noted as an historically key barrier to DR screenings
AI screening model was integrated into the diabetes clinical specialists’ workflow
No patient refused to participate in the study

As a bonus: The study also incorporated referrals for eye diseases other than DR, which ended up increasing the referral number by 33%.

Any notable limitations?

In terms of AI grading vs human garding, the AI detected “any deviation from the healthy images on which it was trained, whereas human graders review a checklist of features for disease.”

For example: A human grader would ignore light reflections observed in the macula of younger patients; however, an AI grader would flag that as an anomaly.
- What this means: It made it challenging for the researchers to quantify and compare the outcomes of the two gradings.

See here for other limitations, including how the type of diabetes among participants and the cost (or, rather, lack thereof) of such screenings for long-term sustainability may have impacted the data.

And the authors’ final conclusions?

Ultimately, they reasoned that the use of AI for DR screenings may be deemed a “practical and acceptable” method for diabetes patients in high-need settings such as Rwanda and the sub-Saharan African region.