Related Experiment Video
Updated: Mar 29, 2026
07:38
Milk Collection in the Rat Using Capillary Tubes and Estimation of Milk Fat Content by Creamatocrit
Published on: December 16, 2015
13.3K
Machine Learning-Driven Computer Vision System for Automated Fat and Energy Quantification in Human Milk
Lujan E Huamanga-Chumbes1, Erwin J Sacoto-Cabrera2, Jaime Lloret3
1TESLA Laboratory, Universidad Nacional de San Antonio Abad del Cusco (UNSAAC), Cusco 08003, Peru.
Sensors (Basel, Switzerland)
|March 28, 2026
Summary
This study introduces a low-cost computer vision system for accurate human milk fat quantification. It significantly reduces measurement error compared to traditional methods, aiding neonatal nutrition assessment.
Area of Science:
- Biomedical Engineering
- Nutritional Science
- Machine Learning
Background:
- Accurate lipid quantification in human milk is crucial for neonatal health.
- Traditional methods like creamatocrit have limitations including bias and uncertainty.
- Developing automated, precise, and cost-effective quantification methods is needed.
Purpose of the Study:
- To develop and validate a low-cost Computer Vision System (CVS) for estimating cream fraction (c) in human milk.
- To utilize machine learning regression for deriving fat and energy content from the estimated cream fraction.
- To provide a reagent-free and operationally feasible alternative for neonatal nutritional assessment.
Main Methods:
- A Computer Vision System optimized for Gold-LED spectrum was developed.
- Machine Learning (ML) regression models were evaluated across Gray Scale, RGB, and Combined feature spaces.
- Rational Quadratic Gaussian Process Regression (GPR) was identified as the optimal model.
- SHAP analysis was used to interpret model feature importance.
Main Results:
- The Rational Quadratic GPR model achieved a high predictive stability with R2=0.867.
- The CVS demonstrated a 57.5% reduction in relative error compared to manual benchmarks.
- SHAP analysis highlighted the importance of Red channel intensities and Blue contrast gradients for lipid globule optical scattering.
Conclusions:
- The developed Computer Vision System is a stable, non-invasive sensing modality for human milk lipid quantification.
- This cost-effective computational framework offers a precise alternative for nutritional assessment in neonatal intensive care units and milk banks.
