MelaKnow-ing your risk
In a Congressional App Challenge video, Wilson High junior Henry Pigg says melanoma accounts for nearly 75 percent of all skin cancer deaths — many of which could have been prevented if a doctor or dermatologist had identified the cancer at an early stage.
That's why Pigg and fellow Wilson High junior Eli Winkleman sought to create an app that would allow for more immediate identification of the deadly disease and thus prevent future deaths.
They're still in the development stage, but the students have already received recognition that gives validity to their ambitious project.
U.S. Rep. Earl Blumenauer named Pigg and Winkelman the winners of the Oregon District 3 Congressional App Challenge for their application called MelaKnow, which is designed to detect and diagnose skin cancer. Pigg and Winkleman were awarded $250 in Amazon Web Service credits and were invited to attend a #HouseofCode reception in Washington D.C., in April.
Their app is also eligible to be displayed in the U.S. Capitol Building and on the House.gov website for one year.
Once it's officially released, the application will allow users to snap a photo of a worrisome skin spot with their iPhone; using artificial intelligence, it would then estimate the liklihood of the spot being malignant. The application also will identify the nearest dermatologist.
"This is the first project that any of my students have done that's been in the medical space," says Wilson High computer science teacher Chris Bartlo. "That really stands out to me. We have a big medical community in Portland with OHSU, and there's a lot of research around that, and so I thought it's cool that the kids saw that and got interested in it.
"I think there's more paths forward down that route."
Pigg gleaned the idea for the project through his parents — one of whom is a dermatologist and the other a computer science teacher. Pigg had worked on game apps in previous years, but he says this was his most ambitious project yet. He knew Winkleman had been working on machine learning, so he asked for his help on the project.
Winkleman was intrigued.
"It sounded like an interesting idea. There's good validation behind it because there's a number of other groups working on the same problem that have had good results," Winkleman said.
The duo planned a bit over the summer, worked on the project in the fall and submitted the project two months after the school year began.
Pigg used Apple Xcode — which is a program used to create iPhone applications —to develop the application. He spent most of his time coding the user interface. Using a data set of about 4,000 images, Winkleman trained an algorithm to determine the probability of a skin spot being malignant.
Pigg says the Congressional App Challenge win gave the team positive reinforcement and made them even more determined to get the application up and running.
"I think it just shows us we're on the right track," he said.
Pigg and Winkleman are currently working to expand their data set to make the algorithm more accurate and are examining how to best nativage regulatory restrictions.
"We're working with 1/20th of the data set that the Stanford team used (for a similar project). We're very restricted in terms of the data we have to train the algorithm. At this point it's not coming up with a better algorithm. It's about getting more data to make the algorithm more accurate," Winkleman said.
Pigg believes that once a fully formed version of the technology is released to the public, it likely wouldn't replace dermatologists but could be a useful tool for general practitioners to determine whether to refer patients to a dermatologist.
"I think someone, one of the research teams, could get it to a point where it's more accurate than a dermatologist, but even then I think people need that in-person (communication). Even if there's proof that the algorithm is more accurate, everyone is going to choose in-person," he said.
Winkleman is interested in theoretical studies related to machine learning, but he says he would also like to apply research to real-world problems. Pigg plans to study computer science in college.
"There's not very many areas where you can do research that's very abstract and cutting edge and take that research and apply it to a real-world problem. Machine learning's one of those areas," Winkleman said.
Pigg and Winkleman are not the first Wilson students to win the Congressional App Challenge. In fact, Marley Bennett and Juliette Coia won last year for creating the application foodDonate, which connects providers of goods and services to nonprofits. This year, Wilson students submitted projects such as a water conservation resource and a hike-scheduling application.
"It's kind of a testament to how strong the students we get are. I think if you have strong students and you give them opportunities, they will do really cool stuff," Bartlo said. "I really like that the projects over the last couple years have really focused on public good. That's something that's important to me."
And Bartlo says projects like this help bright students rise above stacks of college applications.
"It's easy to have a kid with a 4-point (4.0 grade point average) and good SAT scores. But how do you tell that kid from some other kid," he said. "Having these cool capstone-style projects, you can really tell a story of, 'Oh yeah I'm a talented kid. Look what I did with that talent.'"