By making computers "think like a doctor", researchers at Indiana University claim their contextual "artificial intelligence" framework can one-up living medical professionals in terms of both cost and accuracy. Researchers indicate their approach could cut costs by more than half and improve patient outcomes by nearly 50 percent over human doctors.

At the heart of IU's research lie two important implements: Dynamic decision networks (think "machine learning") and Markov decision processes (think "probabailistic logic"). The computer generates its diagnostic and treatment suggestions by developing simulated models which predict outcomes based on the enormous wealth of information available, like electronic health records, biomedical databases and health exchanges.

"Modeling lets us see more possibilities out to a further point, which is something that is hard for a doctor to do," one researcher said. "They just don't have all of that information available to them."

To test their project, researchers enlisted 500 randomly selected individuals from a pool of 6,700 patients clinically diagnosed with depression. About 65-70 percent of those patients were also afflicted with secondary ailments ranging from hypertension to diabetes. The team applied their computerized approach to treat those 500 patients while the remainder were handled by human doctors. Researchers then compared the outcomes of those 500 patients versus human-treated patients.

What they found was a dramatic drop in associated costs – computerized methods yielded an average treatment cost of $189 versus $497 for human doctors. Patient outcomes were also improved by a solid 30-35 percent, although researchers claim they were able to tweak certain model parameters which improved that to almost 50 percent.

Strong emphasis on data-driven decisions is just one possible reason a technology-focused approach could handle diagnosis and treatment better than human doctors. Without intimate knowledge of all possible statistics and studies, doctors sometimes have to rely on intuition and anecdotal experiences.

Even so, researchers aren't aiming to replace doctors entirely. "Even with the development of new AI techniques that can approximate or even surpass human decision-making performance, we believe that the most effective long-term path could be combining artificial intelligence with human clinicians," a member of the project added. "Let humans do what they do well, and let machines do what they do well. In the end, we may maximize the potential of both."