November 29, 2022 – Artificial intelligence could help improve diversity, equity and inclusion in clinical trials and drug development by overcoming some traditional human biases in these fields. But we're not there yet, experts say. The technology could also help doctors with data insights to make diagnosis and treatment more precise.
It starts with quality. Artificial intelligence (AI) relies on big data to create algorithms—or computer instructions—that develop best practices and predictions. But the instructions are only nearly as good as the information used to create them. And humans are those who create the information.
“The basis for the development of AI technologies is people, and these people have their own biases,” says Naheed Kurji, chairman of the Alliance for Artificial Intelligence in Healthcare. “Consequently, the algorithms will have their own biases.”
An example of this can be a technology that uses language to diagnose diseases.
“There are many cases, examples, where companies have failed to recognize the differences in language across cultures,” says Kurji. If technology relies on language patterns of a limited population, “then that model fails when applied in the real world to a different population with a different accent.”
“That’s why it’s not representative.”
Another example is genetic and genomic data.
“More than 90 percent of the genetic and genomic data comes from people of European descent. It does not come from people from Africa, Southeast Asia, Asia or South America,” says Kurji, who can also be president and CEO of Cyclica Inc., a data-driven drug discovery company based in Toronto.
Therefore, “many studies conducted at this data level are inherently biased,” he says.
To be fair
Creating data that takes under consideration diversity, equity and inclusion of individuals and cultures world wide isn’t a hopeless challenge. But it can take time, experts say. Once achieved, AI must be freed from human and systemic biases.
Greater awareness is important.
“The solution to the problem is for people to innately understand that this bias exists,” Kurji says, after which only include fair and balanced data that passes a diversity test.
Vote more properly?
Another promising approach for AI is to optimize the drug development process, narrow down potential drug candidates and increase the cost-effectiveness of clinical trials.
“If the source data has challenges and limitations, AI will simply propagate those limitations further,” agrees Sastry Chilukuri, co-CEO of data-driven clinical trials company Medidata and founder and president of Acorn AI. “The source data needs to become more representative and fair so that AI can reflect what is happening.”
When it involves human or systemic bias in drug development, “it would be too simplistic to say that AI or machine learning can fix it,” says Dr. Angeli Moeller, head of information and integrations that generate insights at Roche in Berlin. “But the responsible use of AI and machine learning can help us identify bias and find ways to mitigate any negative impacts.”
Silent partners?
While AI is designed to streamline drug development, experts say the technology may help all doctors do their jobs higher. For example, AI could help disseminate knowledge and expertise widely and share best practices from doctors with a whole lot of experience with more complex patients. This would help those that treat only a couple of such patients per 12 months.
In New York City or Delhi, the variety of operations is perhaps within the a whole bunch of patients per 12 months, says Chilukuri. “But in the interior of the United States, such as Nebraska, the surgeon simply does not see that many patients.”
AI will help doctors “by giving them the tools to deliver the same high-quality care to all patients much faster,” he says.
Increased efficiency
AI could make therapy more targeted by utilizing data to discover patients at highest risk. The technology could also eliminate some bottlenecks in medicine, similar to the time needed to interpret radiological images, Kurji says.
There's an AI company “whose business model is not to replace radiologists, but to make radiologists better,” he notes. One of the corporate's goals is to “prevent deaths or serious illnesses from X-rays that are missed or left in the pile and simply not acted on quickly enough for the patient.”
Radiologists are so busy that they might only have 30 seconds or less to interpret each scan, Chilukuri says. AI can flag potentially concerning lesions, but in addition compare a picture to previous scans of the identical patient. This AI-enabled vision applies not only to radiology, but to all data-driven areas of medication.
Advancing personalized medicine
AI could also support a personalised approach to surgery, “because people don't come in small, medium and large,” says Chilukuri. The technology could help surgeons determine exactly where to operate on a person patient.
Moeller agrees that AI has the potential to advance personalized medicine.
“AI can help with diagnosis and risk prediction, which can enable earlier intervention,” says Moeller, who can also be vice chair of the board of the Alliance for Artificial Intelligence in Healthcare. “For example, if you look at a diabetic patient, what is the likelihood that he or she will develop eye problems due to diabetic macular edema?”
Technology could also help us see the larger picture.
“Machine learning can look for patterns in a population that may not be in your medical textbook,” says Moeller.
Beyond diagnosis and treatment, AI could also aid recovery by tailoring rehabilitation to every patient, Chilukuri predicts.
“It's not like every person goes through rehab the same way. So there are highly individualized AI plans that allow you to actually stay on track and predict where you're going.”