October 27, 2021 - Podcast

Episode 195—Privacy concerns about smart glasses, and applying AI to healthcare

Imagine putting on a pair of glasses each morning that gives you an electronically enhanced view of the world. That technology may not be too far off into the future. Many companies, including Facebook, are researching new uses for smart glasses, which capture audio and video so wearers can record their experiences and interactions and display useful information within the lenses. But Apu Kapadia, a professor of computer science at IU, says there are privacy concerns associated with this technology and that it is important for companies to proceed with caution and consider the security and privacy risks of augmented reality. While many people are used to being photographed in public when others use their smartphones, augmented reality glasses fundamentally disrupt or violate this sense of normalcy, he says, which can be viewed as a major invasion of privacy. Kapadia and colleagues have found that many people want a clearer sense of when their privacy is being compromised because they find it difficult to know whether they are being recorded. He says we need better ways to convey whether a camera or microphone is recording people. One key application of smart glasses is as a memory aid, allowing people to record or “lifelog” their entire day. Kapadia and colleagues have studied volunteers who wore lifelogging cameras for several days. They discovered several privacy concerns for the camera wearer, and found these people were especially concerned about the places being recorded, as well as recordings of their computer and phone screens, which formed a large portion of their lifelogging history. Kapadia says AI may be utilized to identify digital memories to delete, and his research team has developed AI-based algorithms to detect sensitive places like bathrooms and computer screens, which can then be selectively deleted from a person’s digital memories. Because smart glasses have the potential to do more than simply record video, Kapadia says it is important to prepare for the possibility of a world in which smart glasses use facial recognition, analyze people’s expressions, look up and display personal information, and even record and analyze conversations. As Facebook and other companies move forward with augmented reality, Kapadia says it is critical that they address privacy and security concerns associated with the technology.

In other news, researchers at the Regenstrief Institute are making significant investments and advancements in applying artificial intelligence to healthcare. Dr. Shaun Grannis, vice president for data and analytics at Regenstrief, says AI will continue to be a significant focus of the organization over the next five years due to the tremendous potential AI has for improving the delivery of patient care. Researchers at Regenstrief are focused on machine learning, data mining, and natural language processing, and Grannis says current research projects show encouraging proof of concepts. Evidence suggests that at least one in four adults, and possibly as many as one in two, have a need driven by a social determinant of health, he says. Regenstrief’s award-winning machine learning project called Uppstroms, led by IU’s Josh Vest, addresses patients’ socioeconomic, behavioral and financial needs by combining social determinants of health information with electronic health records. Ultimately, the algorithm identifies primary care patients who may need wraparound services, such as those provided by a social worker or counselor, allowing care providers to make referrals before a situation turns into a crisis. Uppstroms has been deployed in nine clinics within a health system in Indianapolis and can be integrated into electronic health records and used in various healthcare settings to address social determinants of health. Grannis says natural language processing can also be used to tap into data within electronic health records, providing a great tool for researchers, clinicians and healthcare administrators to identify cohorts and analyze trends to inform clinical and administrative decisions. In fact, researchers at Regenstrief, led by Dr. Mike Weiner and Kun Huang, recently showed the promise of this approach by creating the largest chronic cough cohort to date. Grannis says AI can also be used to aid clinical decision making, such as the Health Dart app, whose development was led by Regenstrief’s Dr. Titus Schleyer and partners. The app, Grannis says, sorts through the electronic health record and identifies relevant tests and information related to seven of the most common patient complaints in the emergency department. This novel search algorithm can save clinicians valuable time, allowing them to spend more time with the patient and work more efficiently. Grannis says these AI technologies have been shown to work in real-world settings, and Regenstrief teams will continue to refine and improve these tools and devise new ones in the future. Regenstrief is also investing in the broader ecosystem needed to sustain advanced AI and machine learning methods. In the same way that clinical decision-makers, including physicians and other care providers, undergo regular training updates and certification due to healthcare’s evolving nature and potential for bias, advanced algorithms will need frequent updates and certification. The need for frameworks for overseeing algorithms and analytics is emerging. Developing and evaluating approaches to accurately and efficiently monitor AI and machine learning will become increasingly important in the future of healthcare analytics.