Dr. Hopkins explained that the first step was to get the data into a format that could be interrogated by the AI tools.
An article published in March was the first published as part of the ophthalmology personalized healthcare initiative, and suggests that artificial intelligence could be used to provide widespread, cost-effective eye screenings.1
Dr. Hopkins said the next step is finding how to validate these algorithms on larger, more diverse datasets, which may be best accomplished through collaboration.
One of the challenges is to determine how to build and utilize collaborations to start to build larger collections of data, not only in terms of existing disease but also in earlier stages of disease. The organization sees a need for unprecedented levels of collaboration across the healthcare space to get enough data to really be able to answer key scientific questions and to cover enough patients to be generalizable.
One possible model could be a consortia framework, where a combination of pharmaceutical companies, academic institutions, and clinical groups could get together to look at collecting data in an ongoing fashion. The company is currently looking for the best ways to make this data useable, while always remaining sensitive to data privacy, and any other such issues.
The organization sees potential for personalized healthcare solutions—tools that could predict progression of disease and response to treatment as a means of delivering the right therapy at the right time, in the right dose, and with the right delivery system.
As algorithms are developed, they must be tested and validated. As the company builds collaborations to build datasets, it will become possible to test an algorithm on a set of tens of thousands of patients.
Dr. Hopkins said because it is a new field, regulatory agencies are thinking carefully about what this will look like in terms of their framework. Roche-Genentech and other companies who are doing work in similar areas are working with the regulatory agencies to determine what makes sense moving forward regarding AI, machine learning, and algorithms to bring meaningful impact to patient care, she said.
Jill Hopkins, MD
E: [email protected]
Dr. Hopkins is global head, ophthalmology personalized health care at Roche-Genentech. She has no other financial considerations to disclose.
1. Arcadu F, Benmansour F, Maunz A, et al. Deep Learning Predicts OCT Measures of Diabetic Macular Thickening From Color Fundus Photographs. Invest Ophthalmol Vis Sci. 2019 Mar 1;60(4):852-857. doi: 10.1167/iovs.18-25634.