Abstract
Obesity is an important concern in public health, and Body Mass Index is one of the useful, common and convenient measures. However, Body Mass Index requires access to accurate scales and a stadiometer for measurements, and could be made more convenient through analysis of photographs. It could be applied to photographs comprising more than one individual, leading to population screening. We use Convolutional Neural Networks to determine Body Mass Index from photographs in a study with 161 participants. The relatively low number of participants in the data, a common problem in medicine, is addressed by reducing the information in the photographs by generating silhouette images. We successfully determine Body Mass Index for unseen test data with high correlation between prediction and actual values, with correlation measurements of greater than 0.93 and a mean absolute error of 1.20.
Benjamin Rosman
Lab Director
I am a Professor in the School of Computer Science and Applied Mathematics at the University of the Witwatersrand in Johannesburg. I work in robotics, artificial intelligence, decision theory and machine learning.