Ocular Imaging Standardization for Artificial Intelligence Applications in Ophthalmology: the Joint Position Statement and Recommendations from the Asia-Pacific Academy of Ophthalmology and the Asia-Pacific Ocular Imaging Society
Daniel S. W. Ting,Tien Y. Wong,Ki Ho Park,Carol Y. Cheung,Clement C. Tham,Dennis S. C. Lam
DOI: https://doi.org/10.1097/apo.0000000000000421
2021-01-01
Asia-Pacific Journal of Ophthalmology
Abstract:This Joint Position Statement is supported by Asia Cornea Society, Asia Dry Eye Society, Asia-Pacific Glaucoma Society, Asia-Pacific Society of Ocular Oncology and Pathology, Asia-Pacific Society of Ophthalmic Plastic and Reconstructive Surgery, Asia-Pacific Strabismus and Paediatric Ophthalmology Society, Asia-Pacific Tele-Ophthalmology Society, and Asia-Pacific Vitreo-retina Society. The American Academy of Ophthalmology (AAO) recently published a statement on the recommendation for standardization of images in ophthalmology that has been supported and endorsed by a few ophthalmology societies globally, including the Asia-Pacific Academy of Ophthalmology (APAO).1 The Asia-Pacific region, consisting of both developed and developing countries, is the home to 60% of the world's population, accounting for up to 4.3 billion.2 Given the aging population and the COVID-19 pandemic crisis, there is a pressing need for digital transformation in health care.3 In ophthalmology, the standardization of ocular imaging can facilitate the push for artificial intelligence (AI), deep learning, and big data analytics to enhance patients’ care and experience for global eye health.4–6 The Digital Imaging and Communications in Medicine (DICOM) is the standard for communication and management of medical imaging information that is widely adopted by different medical specialties for data storage, exchange, and transmission.7 It is an imaging standard that was developed by the National Electrical Manufacturers Association (NEMA) and is currently managed by the Medical Imaging and Technology Alliance division in NEMA. Broadly speaking, the DICOM has various application areas, including 1) network image management; 2) network image interpretation management; 3) network print management; 4) imaging procedure management; and 5) offline storage media management.7 In radiology, DICOM standard, implemented in different radiological imaging modalities such as computed tomography, magnetic resonance, nuclear medicine, etc,8 has enabled the digital transformation of the workflow and sharing of radiological images in the large consortium for AI and machine learning. In ophthalmology, the ocular imaging consists of many different modalities that capture the anterior and posterior segment imaging, ranging from ultrasound, color fundus photos, optical coherence tomography (OCT) for macula and optic nerve, keratometers and autorefractors, optical biometry devices, visual field perimetry testing equipment, cornea topography, and widefield ophthalmology photography. Given the low compliance of ophthalmic imaging modalities, the hospital systems, the imaging device companies, and the Picture Archiving and Communications Systems (PACS) need to apply the DICOM standard for all the ocular imaging based on the AAO recommendation. The PACS is a modality of imaging technology that helps image transmission from the site of image acquisition to multiple physically disparate locations. It consists of a combination of an image acquisition device, storage device/server, transmission network, display stations, camera, and integration with the hospital information system. The American Academy of Ophthalmology Recommendations as a Reference For the ophthalmic imaging device manufacturers, it is important to comply with the AAO standards, including:1 Ophthalmic photography devices (DICOM Supplement 91) Optical coherence tomography for anterior and posterior segment (glaucoma and retina) (DICOM Supplement 110) Visual field perimetry testing equipment (DICOM Supplement 146) Optic nerve head topography and retinal thickness mapping devices (DICOM Supplement 152) Corneal topography equipment (DICOM Supplement 168) Widefield ophthalmology photography devices (DICOM Supplement 173) Optical coherence tomography angiography (DICOM Supplement 197) Key measurements in encapsulated PDF (DICOM Correction Proposal 1811) In the recent position statement, the 2 AAO recommendations are:1 1. Provide machine-readable, discrete data for user-selected reports of ophthalmic imaging or functional testing 2. Use of lossless compression for pixel or voxel data to encode the same raw data as used by manufacturers Recommendations From the APAO and the Asia-Pacific Ocular Imaging Society (APOIS) on AI Applications in Ophthalmology 1. Comply and support the AAO recommendations on ocular image standardization 2. Ensure the data privacy and patients’ confidentiality for data and image deidentification during the research and development phase, approved by the institutional research ethics committee 3. Generate robust ground truth for specific use cases and ensure no overlaps between the training, validation, and testing datasets 4. Report the AI algorithm performance using the international reporting standards (eg, CONSORT-AI, SPIRIT-AI, STARD, etc) 9,10 based on the intended use clinical environment 5. Refer to the US Food and Drug Administration (FDA) recommendation on AI and machine learning software as a medical device action plan on tailored regulatory framework, good machine learning practice, patient-centered approach with transparency, regulatory science methods related to algorithm bias and robustness and real-world performance,11 tailoring that to specific country regulatory and clinical needs within the Asia-Pacific region 6. Consider health economics and cost-effectiveness in low-resource and resource-rich settings for AI implementation 7. Future directions in adversarial attack,12 blockchain,13,14 and federated learning15 domains to improve AI transparency and cybersecurity To ensure interoperability between imaging devices and electronic health records, the image standardization is one of the first steps that are critical to facilitate seamless communication and comprehension between the parties. With the advent of deep learning, this could potentially create a powerful imaging database and multiple robust AI algorithms that could be generalized to many parts of the world. For the Asia-Pacific region, the standard of ophthalmic care varies widely among the different countries. To improve the outcome and cost-effectiveness of global eyecare, the standardization of the image using DICOM standard could potentially create a win-win situation between many stakeholders, including patients, service providers (eg, ophthalmologists, optometrists, etc), and manufacturers. In resource-constraint low-to-middle income countries, this becomes even more relevant where more financial and manpower resources can be redirected toward early screening to reduce the incidence of blindness. Following image standardization, more effort could focus on having an agnostic platform to deidentify and share data more effectively between different countries. For AI algorithms development, testing, and deployment, there is also increasing research interest in the areas of federated learning, blockchain technology, and continuous AI learning using different platforms. In Chinese saying, the crisis is called “Wei-Ji” – “Wei” signifies danger and “Ji” signifies opportunity. In other words, the opportunity will always follow danger. The COVID-19 pandemic crisis has catalyzed the need to digitally transform in medical care and the regulators, funders, patients, clinicians, and government are becoming more receptive to the use of AI and digital solutions. To push for AI applications in ophthalmology, image standardization is merely the start. To democratize AI, there are still many areas that require further work, including the development of reporting standards, regulatory and ethical guidelines. The APAO will continue to support the use of AI and digital innovation, powered by image standardization, with the aim to create more innovative solutions for screening, diagnosis, treatment stratification, and disease surveillance to reduce the incidence of blindness in the Asia-Pacific region.