Image: Jingtong Hu, Associate Professor of Electrical and Computer Engineering at the University of Pittsburgh
outlook more
Photo credit: University of Pittsburgh
Artificial intelligence (AI) can help bring important medical screenings to people around the world. University of Pittsburgh engineering researcher Jingtong Hu is working to ensure screening is effective and fair, regardless of whose hand is holding it.
AI has been used in a wide range of healthcare applications, e.g. B. skin or cancer detection, emotion detection, monitoring of vital signs and other medical imaging and diagnostics. However, neural networks are only as good as the dataset they are trained on, and minorities are generally underrepresented in those datasets, leading to a particularly insidious form of technological inequality.
Hu and his team at the University of Pittsburgh’s Swanson School of Engineering are building a distributed, integrative data collection and learning framework based on smartphone apps that will facilitate participation while protecting user privacy. The National Institutes of Health (NIH) recently awarded Hu $1,744,696 for this work.
“Existing and easily accessible data sets are inherently biased. It is not always easy for people in marginalized communities to engage in data collection and research, and these communities may also lack medical professionals,” said Hu, associate professor and William Kepler Whiteford Faculty Fellow of Electrical and Computer Engineering. “AI could make critical healthcare more accessible to these communities; But without a dataset that accurately reflects the diversity of the population, AI could misdiagnose people who are underrepresented during the data collection phase, thereby widening healthcare disparities.”
Hu’s project would help avoid these differences by developing an on-device learning framework that continuously learns from new users’ data when using a mobile application. It will take advantage of Federated Learning (FL), which uses multiple devices to train a common model together while keeping the data on the devices. In FL, the models are shared with the cloud instead of user data to protect user privacy.
“By using this method, not only can we make the global model fairer by including more evenly represented data, but we can also personalize the model for each individual. After all, the most important metric for any user is accuracy for themselves,” Hu said. “A user could use our app to diagnose their skin condition, for example to see if a skin problem is skin cancer or just normal eczema. In the meantime, our algorithm learns locally from the new images. Patient images are not uploaded to the server; They are analyzed on their own phones.”
Unlike existing frameworks, this framework would rely on unsupervised learning with data from a variety of smartphone models and other devices, which would allow more people to participate in the study. The framework would also need to consider the fairness of different machine learning models. The team will develop a machine learning framework that will automatically scan existing learning models and use the best architectures for datasets with diverse data.
The project “Achieve Fairness in AI-Assisted Mobile Healthcare Apps through Unsupervised Federated Learning” will be carried out for four years in cooperation with Dr. Alaina James from the Department of Dermatology at the University of Pittsburgh School of Medicine and Dr. Yiyu Shi of the University funded by Notre Dame and Dr. Lei Yang of George Mason University.
Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of the press releases published on EurekAlert! by contributing institutions or for the use of information about the EurekAlert system.
