A new approach to AI and your medical data
Robert Chang, MD, associate professor of ophthalmology, normally stays busy prescribing drops and performing eye surgery. However, six years ago during an ophthalmology data science competition he was introduced to the capabilities of deep learning. He immediately joined the vanguard pioneering machine learning and artificial intelligence in the ophthalmology field and his glaucoma specialty.
Now, coupled with recent FDA-cleared automated interpretation of ophthalmic photos and other online remote vision testing advances, Chang is focusing on AI deployment and telemedicine, including partnering in an exciting entirely new approach to medical data ownership and sharing, to train AI.
Access to high quality data is a top priority to achieving the full benefits of artificial intelligence, along with other current challenges such as explanatory AI, full generalizability, handling outliers, care integration, and potential confounders or biases.
Chang relies on longitudinal and multimodal eye imaging data to track the development of conditions like glaucoma. And given how big data makes AI stronger, he reasoned with enough scans he might find patterns to help him better interpret test results. That is, if he could get his hands on enough data.
Chang embarked on a journey that is familiar to many medical researchers looking to integrate AI and machine learning into their research. He started with his own patients, but that wasn’t nearly enough, since training AI algorithms can require thousands or even millions of data points. He filled out grants and appealed to collaborators at other universities. He went to donor registries, where people voluntarily bring their data for researchers to use. “I was basically begging for data,” Chang says.
But soon he hit a wall. The data he needed was tied up in complicated rules for sharing. Data ownership, HIPAA compliance, and treatment of inventions and intellectual property were all barriers to large-scale data sharing.
Then, Chang identified a workaround to the data problem: patients.
He started working with Dawn Song, a professor at the University of California-Berkeley and founder of startup Oasis Labs, to create a secure way for patients to share their data with researchers.
Chang and Song agree that for data to be controlled by users, but still usable to others, the whole system needs a rethink.
The approach relies on a cloud computing network and is designed so that researchers don’t have to copy or take possession of the shared data, even when it’s used to train AI. A decentralized and secure hardware platform allows computation on encrypted data, and because all the computations happen within the secure hardware, the researchers never actually access the unencrypted data they are analyzing. To encourage patient participation, the patients could eventually benefit from the outcome of the algorithms or even share in any profit that derives from their applications.
That design has implications well beyond healthcare.
In California, Governor Gavin Newsom recently proposed a so-called “data dividend” that would transfer wealth from the state’s tech firms to its residents, and US Senator Mark Warner (D-Virginia) has introduced a bill that would require firms to put a price tag on each user’s personal data. These initiatives reflect a growing belief that the tech industry’s power is rooted in its vast stores of user data. The initiatives would upset that system by declaring that you own your data, whether it’s your genome or your Facebook ad clicks, and companies should pay you to use it.
Chang and Song agree that for data to be controlled by users, but still usable to others, the whole system needs a rethink. Health research, Song says, is a good way to start testing those ideas, in part because people are already often paid to participate in clinical studies.
Currently, companies mostly get to choose how they store user data, and their business models mostly depend on holding it directly. Companies including Apple have embraced differential privacy as a way to privately gather data from your iPhone and enable features like Smart Replies, without revealing individual personal data, but Facebook’s core ad business doesn’t work that way.
Chang thinks that privacy-conscious design could help open medicine’s data silos, which prevent data from being shared across institutions. Patients and their doctors might be more willing to upload their data knowing it won’t be visible to anyone else, and the new approach would also prevent the sale of patient data to a pharmaceutical company without permission.
“It’s a new approach to medical data AI deployment and telemedicine that may not only benefit ophthalmology and glaucoma research, but also address broader questions of data ownership and privacy,” Chang says.
With thanks to “AI Needs Your Data – And You Should Get Paid For It” by Gregory Barber, Wired.com 8-8-2019, https://www.wired.com/story/ai-needs-data-you-should-get-paid/