Artificial Intelligence (AI) is a game-changer. Not only can this rapidly advancing technology improve the speed and accuracy of disease diagnosis and treatment, it has enormous potential to predict health problems, allowing for far more effective prevention programs that target at-risk populations.
Take, for example, children born with congenital heart defects. This fate currently falls to about 40,000 babies born in the U.S. each year, and about 1.35 million newborns worldwide. What causes defective heart structures in the developing embryo is open to debate. But genetics, diet, environment, medications, and smoking are all on the list.
But what if AI could analyze vast quantities of data and learn from patterns to predict a problem pregnancy even before conception? Neonatal cardiac surgeons are studying this possibility, in hopes of putting themselves out of a job. Instead of time in operating rooms, they are designing educational programs and delivering nutritional supplements to would-be mothers most as risk.
What exactly is AI? And how does it work?
Artificial intelligence refers to computer programs, or algorithms, that use data to make decisions or predictions. To build an algorithm, scientists instruct computers to follow a set of rules in the analysis of data. In machine learning (ML), an algorithm teaches itself how to analyze data and interpret patterns. With exposure to vast amounts of data, learning and interpretation improves.
The question becomes, to what extent can the decisions being made be trusted?
Dr. Hugo Aerts, Director of the Artificial Intelligence in Medicine Program at Brigham and Women's Hospital in Boston, says, “AI can automate assessments and tasks that humans currently can do but take a lot of time.”
Scientists are developing AI tools that use screening images like mammograms to predict risk of developing cancer. To date, doctors used such images to detect if cancer is already present. Due to variation in the skill-level and experience of radiologists, results are highly subjective.
Aerts notes that relying on “a human making an interpretation of an image—say, a radiologist, a dermatologist, a pathologist —that’s where we see enormous breakthroughs being made.”
In 2018, an AI tool hit the news by outperforming 58 international dermatologists in the diagnosis of skin cancer, missing fewer melanomas and misdiagnosing fewer benign moles. AI models have shown impressive precision in identifying lung, breast, thyroid, prostate, and blood-related cancers.
With AI, medical professionals can cut costs, expedite clinical decision-making and significantly reduce wait times.
But despite these successes and benefits, there is reason to be skeptical about early computer models as stand-alone tools for screening cancers or predicting the onset of other diseases. One model, for example, was found to raise alarms not in accordance with the patients’ conditions but with the location where imaging equipment was used.
Yet, scientists are honing the instructions given AI tools by validating results against well known, trusted data. For example, the Framingham Heart Study has been collecting data from a large population cohort for over 70 years. This data provides an opportunity to assess AI findings against established records.
Will the technology become so astute that oncologists and pathologists become obsolete?
Not according to Dr. Olivier Michielin of University Hospital of Lausanne, Switzerland. “AI will enable oncologists, pathologists and other stakeholders to work more efficiently, it will not replace them,” he says.
But AI is undeniably improving the practice of medicine by having computers do what humans cannot – crunching huge amounts of data to expedite diagnosis and treatment. To what extent AI can help prevent disease remains to be seen.
See www.docgiff.com to receive our weekly e-newsletter.