Teen-created software IDs skin conditions, risky drivers and more

The programs ID skin conditions, warn of dangerous driving and translate sign language

a woman with pale skin and brown hair is grimacing as she scratches her shoulder

What’s behind that persistent skin itch? We normally go to a doctor for a diagnosis. But what if a doctor won’t be available for months? A teen’s new phone app may help. Trained on AI, it already reliably recognizes 18 different skin conditions, including eczema and psoriasis.

AntonioGuillem/iStock/Getty Images Plus

DALLAS, Texas — When looking for a science fair project, many kids scratch their heads — then ask for help. Mesk Abdalsalam found her inspiration while watching her brother scratch at the itchy dry skin in the creases of his elbows. A young computer scientist, Mesk realized software might help her brother and others find medical relief when a doctor is not readily available.

a composite of two images. TOP: a photo of Mesk Abdalsalam, a young woman with long black hair, brown skin and a big smile BOTTOM: a screenshot of the software Mesk Abdalsalam developed identifying psoriasis
Top: Mesk Abdalsalam, 16, developed a smartphone app that uses photos to help diagnose skin conditions. Bottom: Derma X correctly identifies psoriasis.M. Abdalsalam

A rising 11th grader, Mesk was among several high-school students who unveiled software solutions to vexing problems in May. All were finalists at the Regeneron International Science and Engineering Fair (ISEF). This teen competition is a program of Society for Science (which also publishes this magazine).

Wait times for medical appointments can be long in Palestine, where Mesk lives. It took nearly six months for her brother to see a doctor. Those dry patches on his arms turned out to be eczema. A chronic disease, it can cause itchy, scaly rashes.

Frustrated by the long delay in her brother’s diagnosis, Mesk decided to use artificial intelligence, or AI, to speed up the process. The 16-year-old at Talae’ Al-Amal Secondary School in Nablus trained her novel software to quickly determine skin conditions based on photos.

Called Derma X, her smartphone app can identify 18 skin conditions. Besides eczema, these also include psoriasis and acne. Training the software required using more than 50,000 publicly available photos. Derma X now correctly diagnoses skin conditions 83 percent of the time.

“It’s really easy, simple, not costly,” she notes. She adds that it’s also “really fast.” The app spits out an answer in seconds.

Mesk has shown her app to many people, including a doctor who “wanted to use it more and more.” Everyone has been shocked, she says. They’re grateful that there could soon be a way to diagnose skin conditions in the blink of an eye.

Derma X earned Mesk fourth place in the systems-software division at ISEF and a $500 prize. She also won a Mawhiba Universal Enrichment Program award and $200 from the King Abdulaziz & his Companions Foundation for Giftedness and Creativity. That last group is an educational foundation in Saudi Arabia that cultivates creativity.

Spotting dangerous driving

Dangerous drivers can be anywhere. Drunk, sleepy and distracted driving can sometimes end tragically, notes Jeesung Lee, 17. This rising high-school junior in South Korea attends St. Johnsbury Academy Jeju. He wondered whether AI could help alert people to such drivers. A system he’s designed can now warn drivers when others near them are acting erratically.

a photo of Jeesung Lee, a young Asian man with black hair smiling at the camera. He's wearing a long-sleeved navy blue Adidas shirt.
Jeesung Lee, 17, created a computer program that can not only detect dangerous driving but also alert drivers nearby. J. Lee

Jeesung’s algorithm scans video feeds for cars traveling at dangerous speeds or suddenly switching lanes. His computer program then sends a message to cellphones in the area. It can text drivers to watch out.

Live video surveys many major roads around the globe, often for traffic reports. This footage will usually be scanned after an accident occurs to assess what happened, Jeesung notes. His AI-based program instead aims to head off accidents.

It monitors streaming video in real time. Jeesung trained his algorithm with publicly available videos showing car accidents. He then tested his system using video it hadn’t seen. It correctly identified dangerous driving some 80 percent of the time, the teen now reports. It also sends out warnings less than two seconds after spotting risky driving.

You can’t get those warnings just yet. To do so, Jeesung wants to use GPS to pinpoint the locations of phone users. That would require permission from South Korea’s government, he notes. But the system should prove quick and reliable, Jeesung says. And it should cost little, he adds. It needs no additional hardware. And his algorithm can be used with existing cameras.

Car accidents are a leading cause of death and injuries. Jeesung hopes a system like his can limit these. He aims to continue tweaking and testing the system in Jeju and eventually go global.

an animated image taken from video showing how software can detect dangerous driving
Jeesung’s new software can identify the dangerous driver seen in this video and then send warning alerts to the cars in its general vicinity.J. Lee

Translating American Sign Language

While working at his part-time job in a hardware store, Liam Sweeney met a deaf customer with a question. But “he had a very complex problem,” recalls this 18-year-old. Liam just graduated from Centaurus High School in Lafayette, Colo. That encounter prompted him to think about how to create a program that can translate American Sign Language, or ASL, into English on the spot.

A photo of Liam Sweeney, a young white man with blond hair, smiling at the camera. Behind him are trees and a stream.
Liam Sweeney, 18, designed a program that translates American Sign Language into English. L. Sweeney

Liam’s translator works by having a computer program match the finger and hand motions of ASL against a list of the letters, words or phrases they represent. He’s used a camera to record signs. His system then reports back the equivalent English text.

Right now, it recognizes only 15 signs — such as the one for “I love you.” But it’s about 98 percent accurate for those signs when viewed by a camera under ideal conditions, Liam says. Good conditions include bright lighting and strong color contrast between the user’s hands and the background.

The teen aims to continue refining this translator by working with a friend who knows ASL and then switching over to AI. While the current software only requires 20 examples per sign, training an AI program will need way more. But it also will result in a system that’s much better at recognizing signs in images and videos, Liam says.

an animated image showing Liam Sweeney signing 'Thank you', there are pink lines on his hand and fingers showing how the software is tracking his movements
Liam’s computer program tracks hand movements to translate American Sign Language, or ASL, into English. Here, Liam signs, “Thanks.” L. Sweeney

Mesk, Jeesung and Liam were among more than 1,600 high school finalists from 64 countries, regions and territories. Regeneron ISEF doled out nearly $9 million in prizes this year. This program has been run by Society for Science since it launched the annual event in 1950.

McKenzie Prillaman is the Spring 2023 science writing intern at Science News. She holds a bachelor’s degree in neuroscience with a minor in bioethics from the University of Virginia and a master’s degree in science communication from the University of California, Santa Cruz.

More Stories from Science News Explores on Computing