Future of Ophthalmology

Abstract

In 2015, out of a global population of 7.3 billion an estimated 34.3 million people were blind, an additional 24.3 million had severe vision impairment, 214 million had moderate vision impairment, and 663 million had near vision impairment. Combined, vision loss accounted for 24.5 million years lived with disability (YLD’s) and therefore was the third largest impairment, globally. In 2015, cataract was the most common cause for blindness, followed by age-related macular degeneration (AMD), glaucoma, uncorrected refractive error, diabetic retinopathy (DR) and cornea-related disorders. Early diagnosis is of upmost importance in reducing the YLD’s and impairment due to many ophthalmological diseases, therefore novel screening methods will play a major role for the future of ophthalmology. In 2021 every 73 days medical data is estimated to be doubled. This plethora of information in combination with the advance of computer science and artificial intelligence (AI) poses a great potential, but also hurdles for the early diagnostics of ophthalmological diseases. Especially ophthalmology with its ever-increasing use of imaging modalities, such as fundus photography, optical coherence tomography (OCT), visual field or objective refraction, has a great potential to evolve by the usage of AI and deep learning methods. Wide coverage of screening programs, reduced barriers to access and improved patient outcomes by providing early detection and treatment in a cost and workforce effective matter can be considered leading implications for the usage of deep learning methods. On the one hand, this article will detail wherein the potential lies and to what extend it is and may in the future be realized. And on the other hand, show and explain the multitude of hurdles in the way of revolutionizing ophthalmology. The most common ophthalmological diseases will be discussed and evaluated according to the state-of-the-art research and the status quo of AI implementation. In the following the four ophthalmological diseases with the biggest perceived interest in the literature for CNN development are compared and evaluated. In the process DR, AMD, Retinopathy of Prematurity (RoP) and Glaucoma are sorted and discussed in a decreasing order in accordance with their perceived importance and impact of AI. This article presents an overview of AI adjunctive to new developments relevant for ophthalmology. Methodologically a literature search in PubMed and Google Scholar for data sources reporting the use of AI in ophthalmology using the search terms: ‘Diabetic Retinopathy’ or ‘Age Related Macular Degeneration’ or ‘Retinopathy of Prematurity’ or ‘Glaucoma’ and ‘deep learning’ or ‘artificial intelligence’ and ‘bioethical dilemma of AI’ was conducted. The published articles that were deemed relevant were summarized.

In 2015, out of a global population of 7.3 billion an estimated 34.3 million people were blind, an additional 24.3 million had severe vision impairment, 214 million had moderate vision impairment, and 663 million had near vision impairment. Combined, vision loss accounted for 24.5 million years lived with disability (YLD’s) and therefore was the third largest impairment, globally. In 2015, cataract was the most common cause for blindness, followed by age-related macular degeneration (AMD), glaucoma, uncorrected refractive error, diabetic retinopathy (DR) and cornea-related disorders. Early diagnosis is of upmost importance in reducing the YLD’s and impairment due to many ophthalmological diseases, therefore novel screening methods will play a major role for the future of ophthalmology. In 2021 every 73 days medical data is estimated to be doubled. This plethora of information in combination with the advance of computer science and artificial intelligence (AI) poses a great potential, but also hurdles for the early diagnostics of ophthalmological diseases. Especially ophthalmology with its ever-increasing use of imaging modalities, such as fundus photography, optical coherence tomography (OCT), visual field or objective refraction, has a great potential to evolve by the usage of AI and deep learning methods. Wide coverage of screening programs, reduced barriers to access and improved patient outcomes by providing early detection and treatment in a cost and workforce effective matter can be considered leading implications for the usage of deep learning methods. On the one hand, this article will detail wherein the potential lies and to what extend it is and may in the future be realized. And on the other hand, show and explain the multitude of hurdles in the way of revolutionizing ophthalmology. The most common ophthalmological diseases will be discussed and evaluated according to the state-of-the-art research and the status quo of AI implementation. In the following the four ophthalmological diseases with the biggest perceived interest in the literature for CNN development are compared and evaluated. In the process DR, AMD, Retinopathy of Prematurity (RoP) and Glaucoma are sorted and discussed in a decreasing order in accordance with their perceived importance and impact of AI. This article presents an overview of AI adjunctive to new developments relevant for ophthalmology. Methodologically a literature search in PubMed and Google Scholar for data sources reporting the use of AI in ophthalmology using the search terms: ‘Diabetic Retinopathy’ or ‘Age Related Macular Degeneration’ or ‘Retinopathy of Prematurity’ or ‘Glaucoma’ and ‘deep learning’ or ‘artificial intelligence’ and ‘bioethical dilemma of AI’ was conducted. The published articles that were deemed relevant were summarized.