Predicting Nonalcoholic Fatty Liver Disease in Obese Populations with 3D Body Scans.

Abstract

Nonalcoholic Fatty Liver Disease (NAFDL), characterized by the accumulation of fat in the liver, is considered a significant threat to public health, particularly in conjunction with obesity and the presence of Nonalcoholic Steatohepatitis (NASH). Traditional clinical methods for evaluating NAFDL, such as liver biopsy and MRI, are either invasive or costly. The utilization of 3D body scans offers a noninvasive and efficient approach to capture precise body shape information rapidly. This study explores the correlation between NAFLD and body shape information in the obese population applying various machine learning models over diverse 3D shape features. The results of the experiments indicate that NAFDL exhibits a stronger association with features containing more intricate shape details. Point cloud features extracted from the 3D trunk region outperform other shape descriptors, such as girth measurements, achieving the highest accuracy at 72% and the F1 score exceeding 0.8 in the classification. These findings suggest that 3D body scans present a promising and cost-effective alternative for the diagnosis of hepatic steatosis. 3D body scans could be valuable in identifying NAFDL and NASH at an early stage, offering a more accessible option for individuals at risk for fatty liver.