Body Composition Analysis: Insights Through Regression and Machine Learning

Data Science Seminar

Nadia Drenska (Louisiana State University)

Abstract:

In this talk we discuss mathematical alternatives to dual energy x-ray absorptiometry (DXA) scans. DXA scans are a costly method of measuring body composition variables such as appendicular lean mass, bone density, and body fat percentage. These variables are fundamental benchmarks in researching osteoporosis, obesity, nutrition, and healthy aging. We employ various regression, supervised learning, and semi-supervised learning techniques, comparing and contrasting their performance on a data set encompassing information on 846 patients. The information is obtained through cheap 3D visual scans of each patient.

This is joint work with a number of students at LSU.