Study Reveals Why AI Models that Analyze Medical Images can be Biased

Artificial intelligence (AI) models often play a role in medical diagnoses, especially when it comes to analyzing images such as X-rays. However, studies have found that these models don’t always perform well across all demographic groups, usually faring worse on women and people of color.

These models have also been shown to develop some surprising abilities. In 2022, MIT researchers reported that AI models can make accurate predictions about a patient's race from their chest X-rays - something that the most skilled radiologists can’t do.

That research team has now found that the models that are most accurate at making demographic predictions also show the biggest "fairness gaps" - that is, discrepancies in their ability to accurately diagnose images of people of different races or genders. The findings suggest that these models may be using "demographic shortcuts" when making their diagnostic evaluations, which lead to incorrect results for women, Black people, and other groups, the researchers say.

"It's well-established that high-capacity machine-learning models are good predictors of human demographics such as self-reported race or sex or age. This paper re-demonstrates that capacity, and then links that capacity to the lack of performance across different groups, which has never been done," says Marzyeh Ghassemi, an MIT associate professor of electrical engineering and computer science, a member of MIT’s Institute for Medical Engineering and Science, and the senior author of the study.

The researchers also found that they could retrain the models in a way that improves their fairness. However, their approached to "debiasing" worked best when the models were tested on the same types of patients they were trained on, such as patients from the same hospital. When these models were applied to patients from different hospitals, the fairness gaps reappeared.

"I think the main takeaways are, first, you should thoroughly evaluate any external models on your own data because any fairness guarantees that model developers provide on their training data may not transfer to your population. Second, whenever sufficient data is available, you should train models on your own data," says Haoran Zhang, an MIT graduate student and one of the lead authors of the new paper. MIT graduate student Yuzhe Yang is also a lead author of the paper, which will appear in Nature Medicine. Judy Gichoya, an associate professor of radiology and imaging sciences at Emory University School of Medicine, and Dina Katabi, the Thuan and Nicole Pham Professor of Electrical Engineering and Computer Science at MIT, are also authors of the paper.

As of May 2024, the FDA has approved 882 AI-enabled medical devices, with 671 of them designed to be used in radiology. Since 2022, when Ghassemi and her colleagues showed that these diagnostic models can accurately predict race, they and other researchers have shown that such models are also very good at predicting gender and age, even though the models are not trained on those tasks.

"Many popular machine learning models have superhuman demographic prediction capacity - radiologists cannot detect self-reported race from a chest X-ray," Ghassemi says. "These are models that are good at predicting disease, but during training are learning to predict other things that may not be desirable." In this study, the researchers set out to explore why these models don’t work as well for certain groups. In particular, they wanted to see if the models were using demographic shortcuts to make predictions that ended up being less accurate for some groups. These shortcuts can arise in AI models when they use demographic attributes to determine whether a medical condition is present, instead of relying on other features of the images.

Using publicly available chest X-ray datasets from Beth Israel Deaconess Medical Center in Boston, the researchers trained models to predict whether patients had one of three different medical conditions: fluid buildup in the lungs, collapsed lung, or enlargement of the heart. Then, they tested the models on X-rays that were held out from the training data.

Overall, the models performed well, but most of them displayed "fairness gaps" - that is, discrepancies between accuracy rates for men and women, and for white and Black patients.

The models were also able to predict the gender, race, and age of the X-ray subjects. Additionally, there was a significant correlation between each model’s accuracy in making demographic predictions and the size of its fairness gap. This suggests that the models may be using demographic categorizations as a shortcut to make their disease predictions.

The researchers then tried to reduce the fairness gaps using two types of strategies. For one set of models, they trained them to optimize "subgroup robustness," meaning that the models are rewarded for having better performance on the subgroup for which they have the worst performance, and penalized if their error rate for one group is higher than the others.

In another set of models, the researchers forced them to remove any demographic information from the images, using "group adversarial" approaches. Both of these strategies worked fairly well, the researchers found.

"For in-distribution data, you can use existing state-of-the-art methods to reduce fairness gaps without making significant trade-offs in overall performance," Ghassemi says. "Subgroup robustness methods force models to be sensitive to mispredicting a specific group, and group adversarial methods try to remove group information completely."

However, those approaches only worked when the models were tested on data from the same types of patients that they were trained on - for example, only patients from the Beth Israel Deaconess Medical Center dataset.

When the researchers tested the models that had been "debiased" using the BIDMC data to analyze patients from five other hospital datasets, they found that the models' overall accuracy remained high, but some of them exhibited large fairness gaps.

"If you debias the model in one set of patients, that fairness does not necessarily hold as you move to a new set of patients from a different hospital in a different location," Zhang says.

This is worrisome because in many cases, hospitals use models that have been developed on data from other hospitals, especially in cases where an off-the-shelf model is purchased, the researchers say.

"We found that even state-of-the-art models which are optimally performant in data similar to their training sets are not optimal - that is, they do not make the best trade-off between overall and subgroup performance - in novel settings," Ghassemi says. "Unfortunately, this is actually how a model is likely to be deployed. Most models are trained and validated with data from one hospital, or one source, and then deployed widely."

The researchers found that the models that were debiased using group adversarial approaches showed slightly more fairness when tested on new patient groups that those debiased with subgroup robustness methods. They now plan to try to develop and test additional methods to see if they can create models that do a better job of making fair predictions on new datasets.

The findings suggest that hospitals that use these types of AI models should evaluate them on their own patient population before beginning to use them, to make sure they aren’t giving inaccurate results for certain groups.

The research was funded by a Google Research Scholar Award, the Robert Wood Johnson Foundation Harold Amos Medical Faculty Development Program, RSNA Health Disparities, the Lacuna Fund, the Gordon and Betty Moore Foundation, the National Institute of Biomedical Imaging and Bioengineering, and the National Heart, Lung, and Blood Institute.

Yang Y, Zhang H, Gichoya JW, Katabi D, Ghassemi M.
The limits of fair medical imaging AI in real-world generalization.
Nat Med. 2024 Jun 28. doi: 10.1038/s41591-024-03113-4

Most Popular Now

MEDICA 2024 + COMPAMED 2024: Adapted Hal…

11 - 14 November 2024, Düsseldorf, Germany. The final preparations for MEDICA 2024 and COMPAMED 2024 in Düsseldorf have begun. A total of more than 5,500 exhibitors from approximately 70 countries...

AI does Not Necessarily Lead to more Eff…

The use of artificial intelligence (AI) in hospitals and patient care is steadily increasing. Especially in specialist areas with a high proportion of imaging, such as radiology, AI has long...

Why the NHS is Seeking to Make Media Ser…

Opinion Article by Dean Moody, Healthcare Services Director, Airwave Healthcare. Tim Kelsey and Martha Lane Fox called for WiFi to be made available free of charge throughout the NHS back in...

Commission Joins Forces with Venture Cap…

The Commission has launched a Trusted Investors Network bringing together a group of investors ready to co-invest in innovative deep-tech companies in Europe together with the EU. The Union's investment...

An AI-Powered Pipeline for Personalized …

Ludwig Cancer Research scientists have developed a full, start-to-finish computational pipeline that integrates multiple molecular and genetic analyses of tumors and the specific molecular targets of T cells and harnesses...

Wearable Cameras Allow AI to Detect Medi…

A team of researchers says it has developed the first wearable camera system that, with the help of artificial intelligence (AI), detects potential errors in medication delivery. In a test whose...

Philips and Medtronic Advocacy Partnersh…

Royal Philips (NYSE: PHG, AEX: PHIA), a global leader in health technology, and Medtronic Neurovascular, a leading innovator in neurovascular therapies, today announced a strategic advocacy partnership. Delivering timely stroke...

AI could Transform How Hospitals Produce…

A pilot study led by researchers at University of California San Diego School of Medicine found that advanced artificial intelligence (AI) could potentially lead to easier, faster and more efficient...

New AI Tool Predicts Protein-Protein Int…

Scientists from Cleveland Clinic and Cornell University have designed a publicly-available software and web database to break down barriers to identifying key protein-protein interactions to treat with medication. The computational tool...

Great Start for Ideas and Innovations: D…

8 - 10 April 2025, Berlin, Germany. From 15 October to 15 November 2024, the DMEA invites experts from business, science, politics and practice to actively participate in shaping the congress...

Start-Ups will Once Again Have a Starrin…

11 - 14 November 2024, Düsseldorf, Germany. The finalists in the 16th Healthcare Innovation World Cup and the 13th MEDICA START-UP COMPETITION have advanced from around 550 candidates based in 62...

G-Cloud 14 Makes it Easier for NHS to Bu…

NHS organisations will be able to save valuable time and resource in the procurement of technologies that can make a significant difference to patient experience, in the latest iteration of...