To assess the performance of a deep learning algorithm for detecting referable glaucomatous optic neuropathy (GON) based on color fundus photographs.
A deep learning system for the classification of ...GON was developed for automated classification of GON on color fundus photographs.
We retrospectively included 48 116 fundus photographs for the development and validation of a deep learning algorithm.
This study recruited 21 trained ophthalmologists to classify the photographs. Referable GON was defined as vertical cup-to-disc ratio of 0.7 or more and other typical changes of GON. The reference standard was made until 3 graders achieved agreement. A separate validation dataset of 8000 fully gradable fundus photographs was used to assess the performance of this algorithm.
The area under receiver operator characteristic curve (AUC) with sensitivity and specificity was applied to evaluate the efficacy of the deep learning algorithm detecting referable GON.
In the validation dataset, this deep learning system achieved an AUC of 0.986 with sensitivity of 95.6% and specificity of 92.0%. The most common reasons for false-negative grading (n = 87) were GON with coexisting eye conditions (n = 44 50.6%), including pathologic or high myopia (n = 37 42.6%), diabetic retinopathy (n = 4 4.6%), and age-related macular degeneration (n = 3 3.4%). The leading reason for false-positive results (n = 480) was having other eye conditions (n = 458 95.4%), mainly including physiologic cupping (n = 267 55.6%). Misclassification as false-positive results amidst a normal-appearing fundus occurred in only 22 eyes (4.6%).
A deep learning system can detect referable GON with high sensitivity and specificity. Coexistence of high or pathologic myopia is the most common cause resulting in false-negative results. Physiologic cupping and pathologic myopia were the most common reasons for false-positive results.
Convolutional neural networks have recently been applied to ophthalmic diseases; however, the rationale for the outputs generated by these systems is inscrutable to clinicians. A visualization tool ...is needed that would enable clinicians to understand important exposure variables in real time.
To systematically visualize the convolutional neural networks of 2 validated deep learning models for the detection of referable diabetic retinopathy (DR) and glaucomatous optic neuropathy (GON).
The GON and referable DR algorithms were previously developed and validated (holdout method) using 48 116 and 66 790 retinal photographs, respectively, derived from a third-party database (LabelMe) of deidentified photographs from various clinical settings in China. In the present cross-sectional study, a random sample of 100 true-positive photographs and all false-positive cases from each of the GON and DR validation data sets were selected. All data were collected from March to June 2017. The original color fundus images were processed using an adaptive kernel visualization technique. The images were preprocessed by applying a sliding window with a size of 28 × 28 pixels and a stride of 3 pixels to crop images into smaller subimages to produce a feature map. Threshold scales were adjusted to optimal levels for each model to generate heat maps highlighting localized landmarks on the input image. A single optometrist allocated each image to predefined categories based on the generated heat map.
Visualization regions of the fundus.
In the GON data set, 90 of 100 true-positive cases (90%; 95% CI, 82%-95%) and 15 of 22 false-positive cases (68%; 95% CI, 45%-86%) displayed heat map visualization within regions of the optic nerve head only. Lesions typically seen in cases of referable DR (exudate, hemorrhage, or vessel abnormality) were identified as the most important prognostic regions in 96 of 100 true-positive DR cases (96%; 95% CI, 90%-99%). In 39 of 46 false-positive DR cases (85%; 95% CI, 71%-94%), the heat map displayed visualization of nontraditional fundus regions with or without retinal venules.
These findings suggest that this visualization method can highlight traditional regions in disease diagnosis, substantiating the validity of the deep learning models investigated. This visualization technique may promote the clinical adoption of these models.
The purpose of this study is to evaluate the feasibility and patient acceptability of a novel artificial intelligence (AI)-based diabetic retinopathy (DR) screening model within endocrinology ...outpatient settings. Adults with diabetes were recruited from two urban endocrinology outpatient clinics and single-field, non-mydriatic fundus photographs were taken and graded for referable DR ( ≥ pre-proliferative DR). Each participant underwent; (1) automated screening model; where a deep learning algorithm (DLA) provided real-time reporting of results; and (2) manual model where retinal images were transferred to a retinal grading centre and manual grading outcomes were distributed to the patient within 2 weeks of assessment. Participants completed a questionnaire on the day of examination and 1-month following assessment to determine overall satisfaction and the preferred model of care. In total, 96 participants were screened for DR and the mean assessment time for automated screening was 6.9 minutes. Ninety-six percent of participants reported that they were either satisfied or very satisfied with the automated screening model and 78% reported that they preferred the automated model over manual. The sensitivity and specificity of the DLA for correct referral was 92.3% and 93.7%, respectively. AI-based DR screening in endocrinology outpatient settings appears to be feasible and well accepted by patients.
Artificial intelligence technology has advanced rapidly in recent years and has the potential to improve healthcare outcomes. However, technology uptake will be largely driven by clinicians, and ...there is a paucity of data regarding the attitude that clinicians have to this new technology. In June-August 2019 we conducted an online survey of fellows and trainees of three specialty colleges (ophthalmology, radiology/radiation oncology, dermatology) in Australia and New Zealand on artificial intelligence. There were 632 complete responses (n = 305, 230, and 97, respectively), equating to a response rate of 20.4%, 5.1%, and 13.2% for the above colleges, respectively. The majority (n = 449, 71.0%) believed artificial intelligence would improve their field of medicine, and that medical workforce needs would be impacted by the technology within the next decade (n = 542, 85.8%). Improved disease screening and streamlining of monotonous tasks were identified as key benefits of artificial intelligence. The divestment of healthcare to technology companies and medical liability implications were the greatest concerns. Education was identified as a priority to prepare clinicians for the implementation of artificial intelligence in healthcare. This survey highlights parallels between the perceptions of different clinician groups in Australia and New Zealand about artificial intelligence in medicine. Artificial intelligence was recognized as valuable technology that will have wide-ranging impacts on healthcare.
Abstract
This study investigated the diagnostic performance, feasibility, and end-user experiences of an artificial intelligence (AI)-assisted diabetic retinopathy (DR) screening model in real-world ...Australian healthcare settings. The study consisted of two components: (1) DR screening of patients using an AI-assisted system and (2) in-depth interviews with health professionals involved in implementing screening. Participants with type 1 or type 2 diabetes mellitus attending two endocrinology outpatient and three Aboriginal Medical Services clinics between March 2018 and May 2019 were invited to a prospective observational study. A single 45-degree (macula centred), non-stereoscopic, colour retinal image was taken of each eye from participants and were instantly screened for referable DR using a custom offline automated AI system. A total of 236 participants, including 174 from endocrinology and 62 from Aboriginal Medical Services clinics, provided informed consent and 203 (86.0%) were included in the analysis. A total of 33 consenting participants (14%) were excluded from the primary analysis due to ungradable or missing images from small pupils (n = 21, 63.6%), cataract (n = 7, 21.2%), poor fixation (n = 2, 6.1%), technical issues (n = 2, 6.1%), and corneal scarring (n = 1, 3%). The area under the curve, sensitivity, and specificity of the AI system for referable DR were 0.92, 96.9% and 87.7%, respectively. There were 51 disagreements between the reference standard and index test diagnoses, including 29 which were manually graded as ungradable, 21 false positives, and one false negative. A total of 28 participants (11.9%) were referred for follow-up based on new ocular findings, among whom, 15 (53.6%) were able to be contacted and 9 (60%) adhered to referral. Of 207 participants who completed a satisfaction questionnaire, 93.7% specified they were either satisfied or extremely satisfied, and 93.2% specified they would be likely or extremely likely to use this service again. Clinical staff involved in screening most frequently noted that the AI system was easy to use, and the real-time diagnostic report was useful. Our study indicates that AI-assisted DR screening model is accurate and well-accepted by patients and clinicians in endocrinology and indigenous healthcare settings. Future deployments of AI-assisted screening models would require consideration of downstream referral pathways.
To investigate the prevalence and incidence of presbyopia in an urban Chinese population.
1817 subjects aged ≥35 years were identified by random cluster sampling in Yuexiu District, Guangzhou, China, ...at baseline in 2008, and all were invited for the follow-up examination in 2014. Distance and near visual acuity (VA) tests, as well as non-cycloplegic automated refraction were performed at each examination as per standardised protocol. Participants with presenting near VA ≤20/40 were further tested with add power at a standard distance of 40 cm to obtain their best-corrected near VA. Functional presbyopia was defined as near VA under presenting distance refraction correction of <20/50 and could be improved by at least one line with add power.
A total of 1191 (83.5% of the 2014 follow-up) participants were included in the current analysis with a mean (SD) age of 50.4 (9.7) years, and 52.9% were female. Prevalence of functional presbyopia at baseline was 25.2% (95% CI 21.5 to 28.9) and the 6-year incidence was 42.8% (95% CI 39.4 to 50.1). Older and more hyperopic subjects had both higher prevalence and incidence of presbyopia (P<0.001). Average presbyopic correction coverage (PCC) was 87.7% at baseline and was significantly lower in myopic participants (P=0.006).
Prevalence of functional presbyopia in urban China is relatively lower along with a higher PCC compared with previous population-based rural cohorts. We identified a high presbyopia incidence, and further studies are needed to understand longitudinal presbyopia progression as well as the urban-rural gap in presbyopia to throw light on future strategic planning.
The eye care workforce, particularly in lower resource settings, face challenges of limited integration into the health system, limited workforce capacity, mismatch of workforce to population need ...and poor quality of care. In recognition of these challenges, coupled with a gap in existing tools, provides a strong rationale for the development of the Eye care competency framework (ECCF).
A mixed methods approach was utilised to develop and validate the ECCF. Content was developed by extracting relevant components of existing frameworks used both within and outside of eye care. A diverse technical working group provided feedback and guidance on the structure, design, and content to create a preliminary draft. Competencies and activities were validated using a modified-Delphi study, and the framework was then piloted at four sites to understand how the tool can be implemented in different settings.
The final version of the ECCF included eight outcomes, nine guiding principles, and content of each of the key elements, including the six domains, 22 competencies, 21 activities, 193 behaviours and 234 tasks, and the knowledge and skills that underpin them. 95/112 participants from the six WHO regions completed the modified-Delphi study, yielding an average of 96% agreement across the competencies and activities in the ECCF. The pilot showcased the versatility and flexibility of the ECCF, where each of the four sites had a different experience in implementing the ECCF. All sites found that the ECCF enabled them to identify gaps within their current workforce documentation.
The ECCF was developed using a collaborative approach, reflecting the opinions of participants and stakeholders from all around the world. The comprehensive competencies and activities developed in the ECCF encompass the diverse roles of eye care workers, and thus encourage multi-disciplinary care and better integration into the health system. It is recommended that eye care workforce planners and developers use the ECCF, and adapt it to their context, to support workforce development and focus on the quality and scope of eye care service provision.
Purpose
It is widely thought that excess pulsatile pressure from increased stiffness of large central arteries (macro-vasculature) is transmitted to capillary networks (micro-vasculature) and causes ...target organ damage. However, this hypothesis has never been tested. We sought to examine the association between macro- and micro-vasculature waveform features in patients with type 2 diabetes (i.e., those with elevated stiffness; T2D) compared with non-diabetic controls.
Methods
Among 13 T2D (68 ± 6 years, 39% male) and 15 controls (58 ± 11 years, 40% male) macro-vascular stiffness was determined via aortic pulse wave velocity (aPWV) and macro-vascular waveforms were measured using radial tonometry. Forearm micro-vascular waveforms were measured simultaneously with macro-vascular waveforms via low power laser Doppler fluxmetry. Augmentation index (AIx) was derived on macro- and micro-vascular waveforms. Target organ damage was assessed by estimated glomerular filtration rate (eGFR) and central retinal artery equivalent (CRAE).
Results
aPWV was higher among T2D (9.3 ± 2.5 vs 7.5 ± 1.4 m/s,
p
= 0.046). There was an obvious pulsatile micro-vascular waveform with qualitative features similar to macro-vasculature pressure waveforms. In all subjects, macro- and micro-vasculature AIx were significantly related (
r
= 0.43,
p
= 0.005). In T2D alone, micro-vasculature AIx was associated with eGFR (
r
= − 0.63,
p
= 0.037), whereas in controls, macro-vasculature AIx and AP were associated with CRAE (
r
= − 0.58,
p
= 0.025 and
r
= − 0.61,
p
= 0.015).
Conclusions
Macro- and micro-vasculature waveform features are related; however, micro-vasculature features are more closely related to markers of target organ damage in T2D. These findings are suggestive of a possible interaction between the macro- and micro-circulation.
We assessed the validity and reliability of self-report of eye disease in participants with unilateral vision loss (presenting visual acuity worse than 6/12 in the worse eye and equal to or better ...than 6/12 in the better eye) or bilateral vision loss (presenting visual acuity worse than 6/12 in the better eye) in Australia's National Eye Health Survey. In total, 1738 Indigenous Australians and 3098 non-Indigenous Australians were sampled from 30 sites. Participants underwent a questionnaire and self-reported their eye disease histories. A clinical examination identified whether participants had cataract, age-related macular degeneration, diabetic retinopathy and glaucoma. For those identified as having unilateral or bilateral vision loss (438 Indigenous Australians and 709 non-Indigenous Australians), self-reports were compared with examination results using validity and reliability measures. Reliability was poor for all four diseases (Kappa 0.06 to 0.37). Measures of validity of self-report were variable, with generally high specificities (93.7% to 99.2%) in all diseases except for cataract (63.9 to 73.1%) and low sensitivities for all diseases (7.6% in Indigenous Australians with diabetic retinopathy to 44.1% of non-Indigenous Australians with cataract). This study suggests that self-report is an unreliable population-based research tool for identifying eye disease in those with vision loss.