Skip Ribbon Commands
Skip to main content
Menu

Technology Development & Commercialisation Office

Technology Portfolio - Software



AVIRI
(Referral for disease-related visual impairment using retinal photograph-based deep learning: a proof-of-concept, model development study)
Visual impairment is a major public health problem. It is associated with reduced quality of life and increased risk of frailty and mortality. Globally, an estimated 553 million people had a visual impairment and 43 million were blind in 2020. To address the burden of visual impairment, WHO recommends annual vision screening for individuals aged 60 years and older, and some high-income countries have already implemented annual vision screening for older people in the community. Nevertheless, strategies and models for simple and efficient screening and referral remain key challenges for sustainable implementation of these screening programmes. The emergence of deep learning technology offers new opportunities to revolutionise this clinical referral pathway. We developed a deep learning algorithm for detection of disease-related visual impairment.


The Technology
Using retinal fundus images from 15 175 eyes with complete data related to best-corrected visual acuity or pinhole visual acuity from the Singapore Epidemiology of Eye Diseases Study, we developed and tested a single-modality deep learning algorithm based on retinal photographs alone for detection of any disease related visual impairment (defined as eyes from patients with major eye diseases and best-corrected visual acuity of <20/40), and moderate or worse disease-related visual impairment (eyes with disease and best-corrected visual acuity of <20/60).


Target Market
Globally in 2020, an estimated 553 million people had a visual impairment and 43 million were blind. 40% of visual impairment is related to refractive error (typically myopia) that requires the provision of spectacles in community settings; however, the remaining 60% of cases cannot be corrected with spectacles and require assessment, diagnosis, treatment, and possibly surgery in eye-care settings led by ophthalmologists. These 60% of people with visual impairment can be referred to as having disease-related visual impairment (ie, substantial loss of vision caused by an eye disease, and unrelated to refractive error). The leading causes of disease-related visual impairment (eg, cataract, diabetic retinopathy, age-related macular degeneration, and glaucoma) are typically age related and are thus increasing in numbers globally.


Current Stage of Development
We validated the performance of the algorithm using retinal images from internal and external dataset of more than 10 thousand participants. In the internal test dataset, AUC for detection of any disease-related visual impairment was 94·2% (95% CI 93·0–95·3), with sensitivity of 90·7% (87·0–93·6) and specificity of 86·8% (85·6–87·9). The AUC for detection of moderate or worse disease related visual impairment in the internal test dataset was 93·9% (95% CI 92·2–95·6), with sensitivity of 94·6% (89·6–97·6) and specificity of 81·3% (80·0–82·5).



Categories: Software, AI, Visual Impairment


Intellectual Property: Proprietary software, knowhow and clinical data



BONSAI
(Artificial Intelligence to Detect Papilledema from Ocular Fundus Photographs)
Examination of the optic nerves is a fundamental component of the clinical examination, but direct ophthalmoscopy is usually avoided or poorly performed by general physicians and non-ophthalmic specialists. Detection of papilledema, defined as optic-nerve edema from intracranial hypertension, and the ability to determine that the optic disk is normal are valuable in the evaluation of patients with headache and other neurologic symptoms. The findings on ophthalmoscopy influence diagnostic strategy and treatment options. Failure to detect papilledema may result in visual loss and neurologic complications.


The Technology
We trained, validated, and externally tested a deep-learning system to classify optic disks as being normal or having papilledema or other abnormalities from 15,846 retrospectively collected ocular fundus photographs that had been obtained with pharmacologic pupillary dilation and various digital cameras in persons from multiple ethnic populations. Of these photographs, 14,341 from 19 sites in 11 countries were used for training and validation, and 1505 photographs from 5 other sites were used for external testing.


Current Stage of Development
The algorithm has been tested in validation and external datasets. In the external-testing data sets, the AUCs were 0.98 (95% CI, 0.97 to 0.98), 0.96 (95% CI, 0.95 to 0.97), and 0.90 (95% CI, 0.88 to 0.92) for the classification of normal disks, disks with papilledema, and disks with other abnormalities, respectively.



Categories: Software, AI, Neuro-Ophthalmology


Intellectual Property: Proprietary software, knowhow and clinical data



CATARACT AI
(Detecting visually significant cataract using retinal photograph-based deep learning)
Age-related cataracts are the leading cause of visual impairment among older adults. Many significant cases remain undiagnosed or neglected in communities, due to limited availability or accessibility to cataract screening. There is a critical need to facilitate access to cataract screening for earlier surgical intervention. This is also important given that cataract surgery is a highly cost-effective intervention. The traditional ophthalmologist-dependent model has limited reach and screening capacity. An automated, deep-learning algorithm that can detect visually significant cataracts based on retinal photographs can help to address this issue.


The Technology
We developed and validated a retinal photograph-based, deep-learning algorithm for automated detection of visually significant cataracts, using more than 25,000 images from population-based studies. This algorithm can serve as a more efficient cataract-screening tool in the community. Furthermore, given the increasing availability of retinal cameras and their increasing use in community eye-screening programs, this new algorithm could be adopted and integrated into existing screening programs.


Target Market
Age-related cataracts are the leading cause of disease-related visual impairment globally, accounting for 94 million adults aged >50 years who experienced low vision or blindness in 2020. Although a cataract is easily treatable, a significant number of patients with a visually significant cataract (that is, a cataract with severe visual loss) remain undiagnosed in communities, especially in rural areas, due to the limited availability of, or accessibility to, cataract screening. Based on a previous report in an Asian population, up to 68.8% of older adults with visually significant cataracts were not aware of having the condition.


Current Stage of Development
We validated the performance of the algorithms using retinal images from more than 9 thousand individuals from internal and external data sets. In the internal test set, the area under the receiver operating characteristic curve (AUROC) was 96.6%. External testing performed across three studies showed AUROCs of 91.6–96.5%.



Categories: Software, AI, Cataract


Intellectual Property: Proprietary software, knowhow and clinical data