17-Year-Old Student Designs Diagnostic Test For Rare Pediatric Heart Disease, Earning Third Place And $150,000

Good News Network

In what has been called the oldest and most prestigious young adult science competition in the country, Society For Science, organized back in 1942 and receiving 1,900 kids’ contribute adventures, has a new winner, 17-year-old Ellen Xu.

She used a kind of artificial intelligence (AI) to design the first diagnostic test for a rare disease close to her heart, Kawasaki Disease. That’s because her sister was a victim of this difficult and often misdiagnosed health condition.

With such a personal story helping drive her forward, she was able to get an 85% positive diagnosis rate with just a smartphone image, giving her the $150,000 prize as the third-place finisher.

Presently, Kawasaki Disease (KD) doesn’t have a testing method, relying completely on a physician’s ability to do research and his or her years of training, and probably a bit of luck as well.

One of the main symptoms of this disease is fever, which is a general symptom of a number of health conditions. However, when it comes to KD, when left undiagnosed, children normally develop long-term heart complications, which thankfully Ellen’s sister was spared from due to a quick diagnosis.

This pushed Ms. Xu to see if she could find a way to design a diagnostic test using deep learning for her Regeneron Science Talent Search medicine and health project. What she created is now known as the convolutional neural network, a form of deep-learning algorithm that mimics the way our eyes work, programmed to analyze smartphone images for possible Kawasaki Disease identification.

Similar to our human eyes, a convolutional neural network requires a vast quantity of data in order to efficiently and rapidly analyze images in comparison to reference points.

For this particular reason, Ms. Xu chose to crowdsource images of Kawasaki’s Disease, searching for lookalike conditions from other medical databases from all around the world. The hope was to gather enough images to give the neural network a better success rate.

As a result, Ms. Xu managed to demonstrate an 85% specificity in identifying between Kawasaki Disease and non-Kawasaki Disease symptoms in children using just smartphone images. And this incredible demonstration of her testing method gave her third place and $150,000 at the Science Talent Search.

See Ms. Xu explain her invention in the video below: