From chatbots providing customer support and facial recognition software facilitating airport security to the launch of ChatGPT and semi-autonomous driving, artificial intelligence (AI) has permeated almost every industry and segment of our lives. In particular, AI is poised to transform healthcare, specifically in effectively tailoring medicine to individual patients.
A Q&A with Eric Topol
At the Scripps Research Translational Institute, founder and director Eric Topol is using AI to reshape medicine, making it more accurate and personalized. Topol, who is also a cardiologist, executive vice president and the Gary and Mary West Chair of Innovative Medicine at Scripps Research, leads a team of clinicians and data scientists who conduct digital research trials—studies that leverage a mobile app-based research platform to gather health data shared by tens of thousands of study participants nationwide. By using AI to analyze data collected through wearable devices such as smartwatches, electronic health records, biosamples and lifestyle and environmental surveys, Topol is helping medicine move toward a future marked by accuracy, empathy and prevention.
What are some challenges in medicine that AI can help address?
Over the last several decades, we’ve witnessed a degradation of the doctor-patient relationship. Doctors are burdened with excessive amounts of clerical work, pecking at keyboards instead of connecting with their patients—and on the other side, patients find themselves struggling to communicate with their doctors during ever-shorter appointments.
Somewhat paradoxically, AI has the potential to fix this problem and bring back some of the humanity that has been lost. By automating many of the administrative tasks through the broad implementation of AI-powered synthetic notes captured through voice recordings, doctors can once again focus on what inspired them to pursue a career in medicine to begin with—listening to and caring for their patients. AI has even been shown to promote empathy, helping train and guide the physicians’ bedside manner.
Another area that AI can make a big impact in is in reducing medical diagnostic errors. In the U.S., we experience 12 million diagnostic errors each year, 800,000 of which lead to death or permanent disability.
We’re already seeing AI make a huge difference in several fields when it comes to diagnosis. In mammography, studies have shown that diagnostic accuracy improves significantly when a radiologist uses AI to interpret scans. The same applies to colonoscopies. The identification of polyps increases when gastroenterologists use machine vision. Machine eyes are trained on millions of images and can detect things human eyes will never be able to see or interpret.
In recent years, there’s been a lot of talk about “precision medicine.” I’m not a fan of that term. If you keep making the same mistake over and over, that’s still very precise. Instead, I prefer an “individualized medicine” approach where we leverage AI to tailor disease prevention, management and treatment to the individual patient based on their unique characteristics.
How will AI accelerate individualized medicine?
AI has been shown to be a remarkably powerful tool for detecting diseases early before any physical signs of illness even appear. Take the retina as an example: Studies have demonstrated how AI can interpret an image of the retina to detect a person’s risk for kidney, Alzheimer’s and Parkinson’s diseases. It can predict heart attacks and strokes in people with no history of cardiovascular disease and detect diseases related to the liver and gallbladder—all from an image of the retina.
Similarly, an AI analysis of the electrocardiogram (ECG) can predict arrhythmias and stroke risk, and detect diabetes and the risk of developing diabetes: as well as kidney disease.
In pathology, AI can identify genomic mutations driving cancer, determine whether a tumor is malignant or benign, and even predict how a person will respond to therapy and what the prognosis is.
All these examples show how we can start to leverage the power of AI to detect and predict disease much earlier than previously possible. We can identify at-risk individuals, conduct active disease surveillance and then use AI to develop individualized preventive strategies.
How is your team at the Scripps Research Translational Institute using AI and digital technologies to advance biomedical research?
We have conducted several studies that demonstrate the power of AI to improve disease detection as well as guide screening and prevention.
During the COVID-19 pandemic, we analyzed data from wrist-worn activity trackers from more than 40,000 people across the U.S. and found that sensors can detect the early onset of viral illness even before symptoms manifest. It was really a proof of concept that data from digital devices can complement virus testing and conventional screening to signal new infections.
We also used data from wearable sensors to study COVID-19’s prolonged physiological impact on the body, finding that it can take months for patients to return to their baselines with more sleep, less activity and higher resting heart rates.
In cardiology, we developed an AI tool that can recreate full 12-lead ECGs with data from only three ECG leads. This enables cardiologists to identify heart attacks with nearly the same accuracy when they review the AI-generated ECGs as compared to the original 12-lead ECGs. This simpler and more accessible electrocardiogram technology could translate to decreased costs and improve patient safety.
We’re also working in the pre-diabetes space, garnering insights into metabolic health at an individual level by analyzing data from continuous glucose monitors, the microbiome, genome, nutritional intake and much more. We hope these diverse layers of data will allow us to more accurately predict glycemic response on an individual basis.
While challenges to implementing AI in medicine more broadly continue to be addressed—from racial bias to ethical concerns about regulating AI systems—the potential of this new technology to revolutionize healthcare can start to be realized.