Artificial Intelligence News

AI in medical imaging could magnify health inequalities, research finds


Artificial intelligence (AI) technologies in the medical field have the possibility to automate diagnoses, reduce the workload of doctors and even bring specialized healthcare to people in rural areas or developing countries. However, with the odds come potential pitfalls.

Artificial intelligence (AI) technologies in the medical field have the possibility to automate diagnoses, reduce the workload of doctors and even bring specialized healthcare to people in rural areas or developing countries. However, with the odds come potential pitfalls.

Analyzing the multi-source datasets used to create AI algorithms from medical images, University of Maryland School of Medicine (UMSOM) researchers found that most did not include patient demographics. In research published April 3 in Natural Medicine, the researchers also found that the algorithm also does not evaluate inherent bias. That means they have no way of knowing whether these images contain samples that are representative of populations such as blacks, Asians, and Native Americans.

According to the researchers, many drugs in the US are rife with bias toward a certain race, gender, age, or sexual orientation. Small biases in individual data sets can be massively amplified when hundreds or thousands of these data sets are combined in these algorithms.

“These deep learning models can diagnose things doctors can’t see, like when someone might die or detect Alzheimer’s disease seven years earlier than any test we know of — superhuman tasks,” said senior researcher Paul Yi, MD, Assistant Professor of Diagnostic Radiology and Nuclear Medicine at UMSOM. He is also the Director of the University of Maryland Center for Medical Intelligent Imaging (UM2ii). “Because these AI machine learning techniques are so good at finding a needle in a haystack, they can also determine sex, sex, and age, meaning these models can then use those features to make biased decisions.”

Most of the data collected in large studies tends to come from wealthy people who have relatively easy access to healthcare. In the US, this means the data tends to lean toward men versus women, and toward white people over other races. Since the US tends to do more imaging than the rest of the world, this data is compiled into an algorithm that has the potential to skewed results worldwide.

For the current study, the researchers chose to evaluate a data set used in a data science competition where computer scientists and doctors collect data from around the world and try to develop the best and most accurate algorithms. These competitions tend to have leaderboards that rank each algorithm and award cash prizes, motivating people to make the best of it. Specifically, researchers investigate medical imaging algorithms, such as those evaluating CT scans to diagnose brain tumors or blood clots in the lungs. Of the 23 competition data analyzed, 61 percent did not include demographic data such as age, gender or race. None of the competitions has biased evaluations towards underrepresented or disadvantaged groups.

“We hope that by addressing this issue in this data competition – and if implemented in the right way – there is great potential to overcome this bias,” said lead author Sean Garin, Program Coordinator at the UM2ii Center.

The study authors also encourage future competition not only requiring high accuracy, but also fairness among different groups of people.

“As AI models become more common in medical imaging and other fields of medicine, it is important to identify and address potential biases that could exacerbate existing health inequalities in clinical care – an important priority for any academic medical institution,” said UMSOM Dean Mark T Gladwin, MD, Vice President of Medical Affairs, University of Maryland, Baltimore, and John Z. and Akiko K. Bowers Honorary Professor.

About the University of Maryland School of Medicine

Now in its third century, the University of Maryland School of Medicine was chartered in 1807 as the first public medical school in the United States. It continues today as one of the world’s fastest growing top-level biomedical research companies – with 46 academic departments, centers, institutes, and programs, and a faculty of more than 3,000 physicians, scientists, and related health professionals, including members from the National Academy of Medicine and the National Academy of Sciences, and a two-time winner of the Albert E. Lasker Prize in Medical Research. With an operating budget of over $1.3 billion, the School of Medicine works closely with the University of Maryland Medical Centers and Medical Systems to provide research-based, academic, and clinical-based care to nearly 2 million patients each year. The School of Medicine has nearly $600 million in extramural funding, with most of its academic departments ranking highly among all medical schools in the nation in research funding. As one of seven professional schools that make up the University of Maryland, Baltimore campus, the School of Medicine has a total population of nearly 9,000 faculty and staff, including 2,500 students, trainees, residents, and associates. The combined College of Medicine and Medical Systems (“University of Maryland Medicine”) has an annual budget of more than $6 billion and an economic impact of nearly $20 billion in state and local communities. Faculty of Medicine, which is ranked as 8th highest Among public medical schools in research productivity (according to Association of American Medical Colleges profile) is an innovator in translational medicine, with 606 active patents and 52 new companies. In the latest US News & World Reports the ranking of the Best Medical Faculties published in 2021 UM Faculty of Medicine is rank #9 among 92 public medical schools in the US, and above 15 percent (#27) out of all 192 public and private US medical school. The School of Medicine operates locally, nationally and globally, with research and treatment facilities in 36 countries around the world. Visit medschool.umaryland.edu




Source link

Related Articles

Back to top button