Artificial intelligence, or AI, is a contentious topic in today’s world, but both health care professionals and researchers are enthusiastic about its potential to revolutionize patient care — from diagnosis and treatment to how doctors correspond with their patients.
While recent advancements have catapulted AI into the spotlight, the technology isn’t by any means new. It’s been around for more than 70 years, with its first medical applications emerging in the 1970s. Since then, the technology has helped doctors diagnose conditions and identify appropriate treatments. It’s also assisted radiologists in interpreting microscopic abnormalities on imaging scans, including mammograms and other routine cancer screenings.
Fast forward to now: AI is more sophisticated than ever before, and it will only continue to become more knowledgeable. Just recently, the first AI-designed drug, Rentosertib, made its way through several clinical trials, showing promise as an effective treatment for idiopathic pulmonary fibrosis (IPF), a chronic, progressive lung disease that makes it challenging to breathe.
AI also assists scientists with drug repurposing, or identifying which readily available medications can be used, alone or in combination with others, to treat rare diseases, including certain cancers and neurological conditions. In radiology, AI offers radiologists a second set of eyes that surpass human capabilities.
Yet despite the technology’s impressive prospects, AI — much like humans — can make mistakes, which is why it’s best used (for now) as a collaborative tool for physicians and researchers. In other words, providers and scientists alike still need to keep a careful eye on its decision-making abilities.
[READ: How to Find a Primary Care Doctor]
How Is AI Used in Medicine?
The use cases for AI in medicine are rapidly expanding. In fact, your doctor likely uses AI during your office visits.
Administrative tasks
“One area that has made a big difference in patient care is ambient documentation, in which AI listens to a conversation between the doctor and the patient and produces notes in real time,” says Dr. David Westfall Bates, who’s the co-director of the Center for Artificial Intelligence and Bioinformatics and the Learning Healthcare System at Mass General Brigham.
In some instances, the physician can have key takeaways from the appointment and next steps drafted and ready to send by the end of the visit.
“AI scribes provide some immediate quick wins, from an organizational perspective, and also decrease the epidemic of clinician burnout that we’re seeing today,” says Brenton Hill, head of operations and general counsel for the Coalition for Health AI (CHAI).
Data suggest that tools like these can reduce a physician’s time spent documenting a patient’s electronic health record (EHR) by as much as 16%. That said, AI scribes can record notes incorrectly sometimes, which is why the physician needs to review the script before logging it into the patient’s medical records or sending follow-up notes via the patient portal.
AI can also be used to generate letters and scripts for insurance companies, Bates says. Essentially, AI can help physicians execute the administrative elements of their job in a more timely and efficient manner.
Disease detection and diagnosis
AI offers clinical support by aiding in diagnosis. It analyzes patterns in patient data and presents them to health care professionals to help guide informed decisions.
“AI algorithms are being tested for their ability to consolidate and analyze a variety of patients’ data, such as medical histories, lab and imaging results, and genetic information, to predict when disease might arise,” says Dr. Andrew Hantel, a health services researcher, hematologic oncologist and ethics consultant at Dana-Farber Cancer Institute and Harvard Medical School. “AI can enhance interpretation of images used in diagnosis or treatment. When used by trained physicians, AI tools can improve the speed and accuracy of some diagnoses.”
For example, tools like Google’s DeepMind have successfully identified breast cancer in mammograms, says Thomas Swalla, CEO of Dotmatics, a software company that connects science, data and decision-making.
According to a study published in a 2020 issue of Nature, Google’s DeepMind AI system outperformed six human radiologists when it came to identifying breast cancer on the screen. This suggests that the tool can be used to help reduce false negatives and false positives.
Bates adds that AI is particularly strong at reading pathology slides, which are tiny pieces of tissue placed on glass for doctors to examine under a microscope and check for signs of disease.
“This is the way most cases of cancer are diagnosed,” notes Bates.
Personalized treatments
“AI will be used a lot in the future, I believe, to personalize treatments for patients,” Bates says.
When treating a specific condition, such as hypertension, a doctor will likely start you on a medication that most people respond well to. But each person has their own unique genetic makeup, meaning how you fare on a medication could be vastly different from another person.
Using AI to make personalized treatment recommendations based on a patient’s genome can help physicians make better choices for patients, potentially saving them the distress of having to rule out multiple medications or therapies through trial and error.
“AI algorithms have the potential to help synthesize different data sources to recommend treatments that account for genetic data, co-occurring conditions and estimated risk,” Hantel says.
These algorithms are still in the testing phase, so they aren’t widely used yet in the clinical setting.
Robotics
Robots can assist physicians in a variety of ways. For example, AI-powered robots are currently being tested for use in surgery to enhance precision and reduce the need for invasive procedures when possible.
“Robotics can also help surgeons better assess their operations in real time,” Hantel says.
Robotics can also be employed to help with logging notes in the patient portal, cleaning hospital rooms and examining images such as MRIs and X-rays.
Drug development
AI is helping design life-saving drugs, such as Rentosertib.
“AI-based algorithms are being used to predict the structure of different human proteins and how they interact with a variety of chemical compounds to identify promising candidates, reduce research costs and shorten the time to clinical trials,” Hantel says.
At Dotmatics, Swalla is seeing AI accelerate early-stage drug discovery by modeling and predicting how molecules will behave before they’re synthesized in the lab. “Scientific intelligence platforms, like our own Dotmatics Luma, enable scientists to simulate, iterate and analyze compound designs faster and more intelligently, reducing both time and cost across the design-make-test-decide cycle.”
[READ: How Do You Know If a Clinical Trial Is Right for You?]
Benefits of AI for Doctors and Patients
The benefits for doctors and patients are endless. Currently, many patients are having their clinical notes created with AI using “ambient documentation;” the AI listens to the interaction and generates a note, which the doctor edits.
Perhaps most important: AI can help physicians support their patients better. When AI becomes intelligent enough to safely make suggestions on which medications people may respond best to, it can save both the doctor and patient time and money.
“This can help us pick the drug that has the best chance of working for you as an individual,” Bates says. “We don’t do things that way right now.”
Doctors often don’t have the time or resources to handle every task expected of them for each patient. One 2022 study found that primary care physicians would need 26.7 hours a day to provide all guideline-recommended care to each patient.
Human error is inevitable, especially when health professionals are working past capacity. Think of AI as the med student that never sleeps. Combine the power of human reasoning with the power of AI and together, they form a duo that creates fewer errors, Hill says.
Another boon to AI-supported healthcare? Those living in underserved areas can have access to best-in-class care. According to a Pew Research Center survey, nearly a quarter of respondents living in rural areas said they didn’t have easy access to good doctors and hospitals.
At CHAI, Hill says patient advocacy communities are reporting that they’re interested in collaborating with AI developers to use tools like large language models to improve access to health care, especially in rural areas. In the future, AI tools could help by answering medical questions, offering credible health information and helping patients decide when it’s necessary to seek care. For example, AI may eventually be able to help someone determine if it’s imperative to make the trek to an emergency room.
When it comes to chronic disease management, AI tools can be an invaluable resource. Apps can help patients log their symptoms and make real-time dietary recommendations to help them manage their health.
“AI can also aid in how medical information is conveyed to patients, especially those with lower health literacy or speak languages that are different from the one their care team uses,” Hantel says.
In addition, wearable devices that use AI can continuously monitor data and flag health risks earlier, Hantel says, helping a patient get ahead of serious symptoms or complications.
[READ: 8 Ways to Get Care When There Are No Primary Care Doctors]
The Limitations of AI in Medicine
While AI has demonstrated immense promise for improving care, it comes with its own set of drawbacks and challenges. The biggest barrier right now is the amount of training needed to get AI tools up to speed so that they can deliver reputable and safe advice to patients.
“There’s a large gap of data that’s made available for these developers to train on,” Hill says. Patient data protection laws, such as the Health Insurance Portability and Accountability Act (HIPAA), inhibit developers from training their algorithms on patient medical records.
In addition, hospitals and clinics in rural areas are often less likely to provide data to developers, Hill says. This means AI tools may not work well for populations they haven’t been trained on, making them less accurate or useful in those settings.
“AI models require large, high-quality datasets, but medical data is inconsistent, incomplete and fragmented across systems,” Hantel says.
Plus, training and distributing AI models nationwide is costly, which can raise the risk of widening existing health disparities. “Building and deploying AI systems is expensive and time-consuming, so smaller practices and resource-limited settings won’t get AI tools as quickly or broadly,” Hantel notes.
AI, as of right now, isn’t immune to biases. Bates brought up a real-life example of an AI scheduling tool for chronic disease patients that predicted Black and Hispanic patients would have higher no-show rates. Based on this data, the algorithm recommended overbooking these groups of people, leading to patients spending an inordinate amount of time in the waiting room even though they had an appointment that day.
From the patient’s point of view, there are concerns around data privacy.
“Increased reliance on digital data and cloud-based AI systems raises concerns about data breaches and confidentiality,” Hantel says.
[See: How to Prevent Medical Errors]
The Future for AI in Medicine
AI will likely always require humans to monitor its decisions and processes. The hope for this sophisticated technology is that, one day soon, it will elevate patient care, enhance research for drug development, educate physicians and streamline administrative tasks.
“However, this needs to be done in a way that is intentionally inclusive so that AI does not worsen pre-existing biases and injustice in how medicine is practiced,” Hantel says. Right now, the cost of implementing AI systems into hospitals and doctors’ offices will likely exclude many facilities.
There may also need to be regulations on how heavily physicians rely on AI, as overdependence could diminish their clinical skills.
There must be checks and balances in play so that AI can’t dehumanize care or use patient and hospital data for profit rather than health as the ultimate goal, Hantel suggests.
Experts predict that a lot will change within the next five years, and there’s still a lot to be done to make sure AI can support physicians and patients reliably and equitably.
More from U.S. News
How to Shop for Health Insurance Covering Mental Health
Can White Noise Help You Sleep Better?
Why Are More Young Adults Getting Colon Cancer?
The Promise and Challenges of AI in Medicine originally appeared on usnews.com
