Artificial intelligence fairness technology has great significance in saving lives

Daphne Yao, a professor of computer science at Virginia Tech, hopes to improve the predictive accuracy of machine learning models in medical applications. Inaccurate predictions can have life-threatening consequences. These prediction errors can lead to miscalculations of a patient's likelihood of dying or surviving their cancer during an emergency room visit.

#Her findings were recently published in the journal Medical Communications, a journal dedicated to publishing high-quality research, reviews, and papers covering all clinical, Translational and public health research areas.

Many clinical data sets are unbalanced because they are dominated by majority population samples, Yao said. In the typical one-size-fits-all machine learning model paradigm, racial and age differences are likely to exist but may be ignored.

Yao and her research team collaborated with Charles B. Nemeroff, a member of the National Academy of Medicine and associate professor of psychiatry and behavioral sciences at the Dell School of Medicine at The University of Texas at Austin Professor in the department studies how bias in training data affects prediction outcomes, particularly for underrepresented patients, such as younger patients or patients of color.

"I am extremely excited to be working with Yao, who is a world leader in advanced machine learning," said Nemeroff. "She and I discussed a concept that machines The new advances in learning could be applied to a very important problem that clinical researchers often encounter, namely the relatively small number of minorities who typically participate in clinical trials.”

This results in medical conclusions being drawn primarily for the majority group (white patients of European descent), which may not apply to minority ethnic groups.

Nemeroff said: "This new report provides a way to improve the accuracy of predictions for minority groups." "Clearly, these findings have important implications for improving the treatment of minority patients." Clinical care is of great importance.”

Yao’s Virginia Tech team consists of doctoral students Sharmin Afrose and Wenjia Song in the Department of Computer Science and Chang Lu in the Department of Chemical Engineering, Composed by Professor Fred W. Bull. To conduct the study, they conducted experiments on four different prognostic tasks on two datasets, using a novel dual-priority (DP) bias correction method to train customized models for specific ethnic or age groups.

"Our work demonstrates a new AI fairness technique that can correct prediction errors," said Song, a fourth-year doctoral student whose research areas include digital health and cybersecurity machine learning in . "Our DP method improves performance in minority classes by up to 38% and significantly reduces prediction differences between different demographic groups, 88% better than other sampling methods."

## The #Surveillance, Epidemiology, and End Outcomes dataset was used by Song for tasks on breast cancer and lung cancer survival, while fifth-year doctoral student Afrose used a dataset from Beth Israel Deaconess Medical Center in Boston for in-hospital mortality prediction and lapse compensation prediction task.

"We're excited to have found a solution to reduce bias," said Afrose, whose research focuses include machine learning in healthcare and software security. "Our DP bias correction technology will reduce potentially life-threatening prediction errors for minority groups."

As these findings are published and publicly accessible, the team is eager to collaborate with other researchers Collaborate to use these methods in their own clinical data analyses.

#Song said: "Our method is easily deployed on a variety of machine learning models and can help improve the performance of any prognostic task with representation bias."

The above is the detailed content of Artificial intelligence fairness technology has great significance in saving lives. For more information, please follow other related articles on the PHP Chinese website!