Gender-, age-, and BMI-specific threshold values of sagittal abdominal diameter obtained by artificial neural networks

Abstract

© Taiwanese Society of Biomedical Engineering 2015. Sagittal abdominal diameter (SAD) is a valuable predictor of cardiometabolic outcomes in obese patients, but its use in clinical practice is limited due to a lack of specific threshold values. Some authors have proposed SAD thresholds using various methodologies that vary from 19.3 to 27.6 cm. All these values are static, which means that a given threshold is used for all age and body mass index (BMI) groups. The main goal of this paper is to show that SAD thresholds have gender-, age-, and BMI-dependent dynamics. We obtained SAD thresholds using feed-forward artificial neural networks (ANNs) with backpropagation as a training algorithm. SAD thresholds are derived from an evaluation of the relationship between SAD and cardiometabolic risk factors. They vary from 16.36 to 29.32 cm. ANN estimates SAD from gender, age, BMI, systolic and diastolic blood pressures, high-density lipoprotein (HDL), low-density lipoprotein (LDL) and total cholesterol, triglycerides, glycemia, fibrinogen, and uric acid. ANN training, validation, and testing were conducted in the MATLAB environment using a dataset that included 1475 persons. The accuracy of our solution is above 88%.