Ophthalmology is currently one of the specialties at the forefront of technology in medicine. Numerous members of our field are pioneers in the areas of big data, AI, machine learning, and deep learning, all of which will enable us to develop more efficient and sustainable medical practices today and in the future.
WHAT IS BIG DATA?
The term big data refers to large data sets (structured or unstructured) that, due to their volume, variety, and velocity, can be complex to analyze, manage, or process with traditional technologies or tools.1 In order to make big data more reliable, we must also keep in mind the veracity and value of these data, as doing so results in more trustworthy and applicable information.
When it comes to managing such a high volume of data, the key is to determine the rational and intelligent use of the information. This involves analyzing, structuring, and optimizing the data in a way that enables better decision making, reduces costs, and produces more effective results in health systems.
BIG DATA APPLICATIONS
Currently, big data has applications in all health sectors, ranging from administrative management to telemedicine, epidemiology to genomics, and clinical processes to trials. Given these potential uses, it is safe to assume that, in the future, big data will play a role in our ability to prevent ophthalmic disease and to follow patients’ clinical conditions remotely and in real time. These types of advances are already starting to occur, thanks to an avalanche of available data.
IRIS registry. Developed by the AAO, the Intelligent Research In Sight (IRIS) registry is the largest registry of specialized clinical data in the world. As of September, the database included information from more than 252.95 million consultations and 60.78 million patients, obtained from more than 15,000 US eye care providers.
The IRIS registry includes data on patient demographics, medical and ocular histories, clinical examination findings, diagnoses, procedures, and medications. Access to this information may allow ophthalmologists to become more aware of less common pathologies, medication and pathology relations, etiological hypotheses, and early disease detection and predispositions.2 With the IRIS registry, for example, users can access information on the rates of endophthalmitis after cataract surgery and anti-VEGF injections, visual outcomes and specific risk factors for endophthalmitis, the prevalence of myopic choroidal neovascularization and practice patterns for treatment, and more.3
Allergic eye disease in India. A recent study used big data analytics to evaluate the demographics, clinical presentation, and risk factors of allergic eye disease (AED) in children and adolescents in India.4 The purpose of the study was to gain further knowledge into this pathology, and it was achieved thanks to the use of electronic medical records and big data, which facilitated the structuring and analysis of 259,969 patients in an ophthalmology hospital network. Relying on organized and analyzed information, the authors were able to conclude that about one-tenth of the children and adolescents seeking eye care in India are affected by AED. They noted that the condition commonly affects boys with atopy from middle-to-high-income families during early to middle childhood and shows a self-limiting trend by late adolescence.
This study provides a clear example of how big data can be used to better understand the demographics, clinical manifestations, and risk factors of a particular condition and thus enable the design of more suitable preventive and therapeutic strategies in response.
Diabetic retinopathy. Another example of how big data and deep learning can be applied in ophthalmology involves the detection of diabetic retinopathy (DR). In one study, a deep convolutional neural network was trained using a retrospective development data set of 128,175 retinal images, which were graded three to seven times for DR, diabetic macular edema, and image gradability by a panel of 54 US licensed ophthalmologists and ophthalmology senior residents.5 The investigators found that the resultant deep learning–trained algorithm had high sensitivity and specificity for detecting DR and macular edema from retinal fundus photographs—97.5% and 96.1% sensitivity and 93.4% and 93.9% specificity in two validation sets.
CHALLENGES FOR THE FUTURE
The linkage challenge. Griffin M. Weber, MD, PhD, Associate Professor of Medicine and Biomedical Informatics at Harvard University, posed an interesting challenge in a 2014 article about how linking big data will allow physicians and researchers to test new hypotheses and identify possible areas of intervention.6 Dr. Weber posited: If we determined the shopping patterns of patients in grocery stores, could we predict the rates of obesity and type 2 diabetes (and, for our purposes, better prevent DR)? Does patients’ distance from hospitals and pharmacies affect their behavioral patterns? The ability to assess individual behaviors and link them to patients’ medical histories could open new horizons for effective disease management and prevention.
Explainable AI. Explainable AI (EAI) is an emerging field of machine learning with the purpose of making AI more transparent and understandable for users. Needless to say, EAI will be key to the application of this technology in ophthalmology. EAI will guarantee its users clarity and understanding of how and why decisions are made by the algorithms used in machine learning and deep learning.7
The applications of big data, AI, machine learning, and deep learning are increasingly evident as well as necessary in health care and medicine. Cost reductions, the accessibility of wearable devices, and improved decision making are only a few of the advantages afforded by an almost unimaginable mass of data. But the technological challenges ahead will not be easy to navigate. To provide more personalized, preventive, and predictive ophthalmic care, it will be the duty of all agents involved to seek the correct, ethical, and efficient use of the information available for the benefit of patients.
1. Sosa Escudero W. Big Data. Buenos Aires, Argentina: Siglo XXI Editores; 2019.
2. Parke Ii DW, Lum F, Rich WL. The IRIS Registry: purpose and perspectives. Ophthalmologe. 2017;114(Suppl 1):1-6.
3. Chiang MF, Sommer A, Rich WL, Lum F, Parke DW 2nd. The 2016 American Academy of Ophthalmology IRIS Registry (Intelligent Research In Sight) database: characteristics and methods. Ophthalmology. 2018;125(8):1143-1148.
4. Das AV, Donthineni PR, Prashanthi S, Basu S. Allergic eye disease in children and adolescents seeking eye care in India: electronic medical records driven big data analytics report II. Ocul Surf. 2019;17(4):683-689.
5. Gulshan V, Peng L, Coram M, et al. Development and validation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs. JAMA. 2016;316(22):2402-2410.
6. Weber GM, Mandl KD, Kohane IS. Finding the missing link for big biomedical data. JAMA. 2014;311(24):2479-2480.
7. Gunning D. Explainable artificial intelligence (XAI). November 2017. www.darpa.mil/attachments/XAIProgramUpdate.pdf. Accessed December 11, 2019.