The Capstone Project, the crowning fulfillment of the Master of Science in Analytics (MScA) program, unites students with industry partners to solve real-world data science problems. At the Spring Showcase, where projects spanned industries, data types, and methodological approaches, two teams stood out for special commendation.
“Hard work and perseverance were clearly on display in the exceptional quality of all the projects,” said Greg Green, executive director for the MScA program. “We are looking forward to staying connected with the graduates and hearing about their next achievements.”
Read on for more about the winning projects.
UChicago Medicine: Advancing the Sociome for Social and Health Equity
Sandra Tilmon, James Urdiales, Megan McDermott, Megan Hagenauer
Tilmon, Urdiales, McDermott, and Hagenauer used predictive modeling to understand the influence of social determinants on pediatric hypertension, an increasingly prevalent but under-researched area. Advised by Ming-Long Lam and sponsored by UChicago Medicine, they merged clinical patient data with public data at the census-tract level and were able to identify neighborhood factors in at-risk populations associated with increased risk.
“We documented our end-to-end approach into a framework enabling its expansion for use with other diseases,” the team wrote. “Since the client’s future community-based dataset will follow the structure of the one used in this paper, we have high expectations of their being able to use most, if not all, of the structure we have created.”
Inference Analytics: Extraction of Implant Brand from Dental Images
Macy Bao, Bryce Chamberlain, Rupa Ghosh, Nick Lesh
Bao, Chamberlain, Ghosh, and Lesh aimed to improve on current methods for identifying dental implant brands. Implants are used widely today as dental root replacements. Advised by Utku Pamuksuz and sponsored by Inference Analytics, they proposed a solution that uses artificial neural networks to identify the brand of dental implant from a panoramic X-ray.
“While only 5-10% of implants require dental services, dentists need to determine the brand of dental implant to plan effective treatment,” the team wrote. “Current methods are manual and hence time-consuming and error-prone. Our proposed solution using a YOLOv5 cropping model yielded 97.4% accuracy, while the mean average precision of the implant identification model ranged from 77.5% to 99.5% for different implant types.”