Yookta Pandit
8th Grade, The Quarry Lane School
Dublin, CA
Reading the Brain: Using MRIs for Early Detection of Dyslexia
View Poster
Project Background
Yookta has always loved to read. One of her favorite book series, Percy Jackson & the Olympians, taught her that many kids struggle with reading because of dyslexia. “When I did my research on dyslexia, I found that one in ten children in America struggle with it and do not get timely help, because detection is hard. I wanted to help them enjoy reading as much as I did,” Yookta says. After taking a neuroscience class and one about AI, she decided to combine the two to try and detect dyslexia.
Tactics and Results
Yookta looked for publicly available data sets of Magnetic Resonance Images (MRI), which show clear anatomy of the brain. She found scans from 40 children and adults with dyslexia and 42 controls. She divided the brain into 132 anatomical parts. Yookta used the different features to train her AI model, which compared the sizes of the regions between controls and people with dyslexia. Her first effort achieved only 56.23 percent accuracy, but she then developed more models and trained the AI on specific brain areas, like the right entorhinal area, the left posterior orbital gyrus, and other brain areas linked to memory, navigation, reward and emotion. By trial three, her AI was trained to find differences between dyslexic and control brains at 87.5 percent accuracy.
Beyond the Project
Yookta loves to read, and she especially loves fantasy series like Percy Jackson & the Olympians. She is also on a swim team, and swims around two miles every day. Yookta is also a part of the AI Ambassadors program, learning about applications of AI. She would like to become a computer scientist. “I am excited to continue researching in computer science and believe that this career is the best way for me to have an impact on the world.”
