Artificial Intelligence: Transforming the future of the medical field
By: Diana Bello Aristizábal
Artificial intelligence (AI) has been gradually transforming the world in multiple industries being the medical field the one that has been most impacted by these tools that, although they improve processes and services, also pose opportunities and challenges.
General advantages include reducing the frequency of medical errors, improving diagnostic accuracy and saving time by automating repetitive activities. But how widely are they being used in healthcare today, how useful are they, and what will they look like in the future?
“We have more than 90 active AI projects spanning health system operations and clinical diagnostics,” James Lindgren, Executive Director of Optimization at University of Miami Health System, said at the “Artificial Intelligence: Navigating Our Future” forum organized in June by the South Florida Hispanic Chamber of Commerce.
Dr. Fernando Collado-Mesa, Professor of Clinical Radiology at University of Miami Health System, told within the same panel discussion that there are around 800 software and medical devices that have been authorized by the FDA for use in healthcare in the U.S. “76% of them are in radiology. The second specialty that follows is cardiology with merely 10%.”
Their help is undeniable. “There are artificial intelligence systems that in the screening to detect breast cancer have been able to find very small lesions that the radiologist may not have seen,” Dr. Collado-Mesa said in an interview for Doral Family Journal.
Additionally, some systems tell the radiologist whether or not what they are seeing poses a high risk. This way, the patient doesn’t have to return for more imaging, which reduces costs and allows for an early diagnosis.
Artificial intelligence is not only contributing to more accurate and early detection of diseases but is going further. “With AI we are detecting the architecture of a female breast that will lead to breast cancer so we can prevent it,” Dr. Allan Stewart, Chairman of Surgery and Medical Director of Cardiac Surgery at HCA Florida Mercy Hospital, stated in an interview with this journal.
Concerns and opportunities
Despite AI’s positive impact, specialists also recognize its downsides and the challenges they bring. James Lindgren, for example, explained in the forum that after deploying a natural language Bot (its purpose is to generate natural conversations with greater understanding and accuracy for the parties involved) in the UHealth telephone systems, he found a strong bias in the language.
University of Miami Health System offers support in English, Spanish and Creole for scheduling appointments. However, when doing an accuracy test, it was discovered that the Bot does a much better job flipping from Spanish to English than the opposite scenario. The accuracy was 99% against 50% respectively.
“This happened not because the technology is flawed but because the data it was trained on is inherently biased towards the English language,” Lindgren explained. Cases like the previous one have made the focus shift from understanding everything the technology does to how to apply the technology.
Trust and control are two essential parts of this equation. “We don’t trust blindly. We take a sample and verify it to make sure the given value is accurate,” Lindgren said, adding that he set up an artificial intelligence tool that monitors all of the health system’s computers and data ports and sends notifications if something out of the ordinary happens.
“Testing is too important because that is in part how trust is built. To move in that direction, we have to educate ourselves, learn how the most sophisticated programs work, why it is worth using them or not and oversee the outcomes since there must always be a human involved in the decisions,” Dr. Collado-Mesa pointed out.
This also applies to the diagnosis of diseases, as the radiologist cannot become too confident in artificial intelligence to the extent of not verifying the information provided. “If the system tells you that something is suspicious and you don’t check it, it may not actually be suspicious.”
Considering all that AI can do, according to Dr. Allan Stewart, in the future the technology will be more geared towards wellness rather than focusing on how to treat diseases. “We want to move out of the space of how to best treat cancer or heart disease and instead learn how to avoid them.”
It will also help to more quickly verify the effectiveness of a medical treatment. “Right now, we can’t determine whether or not a treatment works without following it for a while, but in the future, we will have a massive amount of data to know this faster and without including the placebo group in clinical trials.”
In addition, AI will continue to be key in cardiovascular and coronary artery disease, helping to prevent more strokes and mitigate heart attacks. In addition, it is estimated that these technologies will have an impact not only on breast cancer but also on lung and colon cancer as it will serve to increasingly better detect these three types of cancer.
“If you look at Tesla, there are autonomous driving programs that are not perfect yet, but with all the data that is being collected they will be over time. The same will happen in the medical field. The more data we get, the more information we will have that will eliminate errors.”