Hybrid Human-AI Machine Outperforms Radiologist as well as Deep-Learning Algorithms in Pneumonia Diagnoses

Hybrid Human-AI Machine Outperforms Radiologist as well as Deep-Learning Algorithms in Pneumonia Diagnoses
Share this:

Artificial intelligence (AI) can undeniably boost healthcare outcomes thanks to its impressive predictive capabilities. The technology has shown that it can outdo dermatologists, especially when it comes to diagnosing melanoma. Also, a recent study conducted by Google’s DeepMind indicated that the algorithms have equal capabilities with clinicians as far as detecting eye conditions is concerned.

Despite all that AI can achieve, Unanimous AI has raised the question of whether AI can perform better with the help of humans. The San Francisco-based startup has been pitching this idea for about four years now. Furthermore, a study carried out with the Stanford University School of Medicine revealed that Unanimous AI’s system not only diagnosed pneumonia 22 percent more accurately compared to the radiologists but also minimized errors by 33%.

“Humans have experience, knowledge, wisdom, and intuition that’s not being represented in these big datasets. It takes 12 years to become a radiologists – 12 years of talking to other doctors, observing and processing the world, and filing the databases inside their heads. There’s value in this,” Unanimous AI’s CEO and Stanford-educated scientist Louis Rosenberg said to VentureBeat.

Unanimous AI’s platform takes advantage of swarm intelligence, which is a biological phenomenon whereby groups of organisms augment their intellect through creating real-time systems in a bid to boost prediction accuracy. Rosenberg explained by saying: “animals use unique techniques they’ve evolved over millions of years with feedback loops that converge on an optimal combination of insights. [Scout] bees…vibrate their bodies, and birds detect motions propagating through flocks.”

Contrary to the swarm intelligence seen in nature, Unanimous AI includes itself to the mix through a process dubbed artificial swarm intelligence. It works by radiologists at Stanford and other institutions spread out across the nation logging into Unanimous AI’s platform through a networked computer. Together, they then attempt to shift the cursor’s position towards a prediction [diagnosis] using a touchscreen, touchpad or mouse. During this process, algorithms process the human participant’s behavior in real time.

The two-step process involves participants converging on a coarse range of numerous possibilities and on a refined value. However, this takes place a lot faster than you might think. In nearly 60 seconds on average, radiologists involved in the study settled on 50 diagnoses, with several participants exerting additional impact in comparison to the others. What’s more, Unanimous AI’s algorithms understand the conviction of swarm members based on their cursor movements and then measure their contributions.

“Over the last four years, [the system] has learned from hundreds of thousands of questions.” You can think of it as a system of feedback loops. It’s aggregating sentiment,” said Rosenberg.

The radiologists attained 82% total diagnostic accuracy in comparison to human experts-73 % accuracy. In addition, the authors of the study wrote: “[S]warm-based technologies are quite promising for use in medical diagnosis.”

Although the recent endeavor marks the first time for Unanimous AI’s technology to be employed in the medical field, it is not the first time for the startup to make headlines. In fact, back in 2016, the company correctly projected the Kentucky Derby’s results. It also predicted the Oscar winners (for the last three years) more accurately compared to any publication. Other notable forecasts include the precise score of last year’s Super Bowl, President Trump’s approval rating and the Time Person of the Year.

Rosenberg is well-known for pioneering among the world’s first augmented reality systems, particularly for the United States Air Force back in the early 90s. He is convinced that swarm intelligence is a technique of amplifying human capabilities in ways that artificial intelligence cannot.

“Most AI companies are focused on finding patterns in big data, basically, and that only works for data you can put into a database in a very consistent way. The question we ask is, how can we connect groups together and make them smarter using AI? How can we take groups of people and turn them into artificial experts?”

Currently, Unanimous AI boasts two products (Swarm AI and Swarm Insight) in its portfolio, which run on a system-as-a-service model. Swarm Insight is known for having an enterprise bent, whereby one client utilized it in predicting how customers may react to a TV ad for pancakes. In another ad, beverage entity Constellation Brands used it to measure customer reactions to its cannabis investment. Meanwhile, Swarm AI is able to tap an enterprise’s workforce for consensus.

Source VentureBeat

Share this:

Leave a Reply

avatar
  Subscribe  
Notify of