Artificial Neural Network Models for Accurate Predictions of Fat-Free and Fat Masses, Using Easy-to-Measure...

Area of Science:

• Biomedical Engineering
• Human Physiology
• Data Science in Health

Background:

• Abdominal fat and fat-free mass are closely linked to cardiometabolic risks.
• Assessing these specific body compartments is crucial for understanding health risks.
• Existing methods for body mass assessment can be complex or inaccessible.

Purpose of the Study:

• To develop accurate predictive algorithms for abdominal fat and fat-free masses.
• To utilize easily measurable parameters for accessible body composition analysis.
• To compare the efficacy of artificial neural network (ANN) and multiple linear regression (MLR) models.

Main Methods:

• 104 healthy subjects with elevated adiposity or waist circumference were included.
• Multiple linear regression (MLR) and artificial neural network (ANN) models were constructed.
• Data were rigorously divided into training, validation, and test sets, with 20 repetitions to minimize bias.

Main Results:

• ANN models demonstrated superior performance over MLR models, achieving higher R² values (0.96-0.98 vs. 0.80-0.94).
• ANN models exhibited significantly lower root mean square error (RMSE), indicating greater prediction accuracy.
• ANN models accurately predicted trunk fat mass (±1.84 kg error) and fat-free mass (±1.48 kg error).

Conclusions:

• Developed algorithms provide cost-effective tools for predicting key adipose and lean tissues.
• These algorithms can aid in assessing cardiometabolic risk factors.
• ANN models offer a highly accurate and accessible method for body composition analysis.