Infants Sucking Pattern Identification Using Machine-Learned Computational Modeling.

Area of Science:

• Biomedical Engineering
• Pediatrics
• Machine Learning

Background:

• Breastfeeding requires coordinated sucking, swallowing, and breathing for infant nutrition.
• Assessing infant milk intake and sucking proficiency in real-time is challenging.
• Current clinical assessments rely on subjective professional judgment.

Purpose of the Study:

• To develop and validate a novel, objective method for identifying infant sucking patterns during breastfeeding.
• To leverage sensor data and machine learning for early detection of suboptimal sucking behaviors.
• To provide a decision-support tool for clinicians to improve infant feeding assessments.

Main Methods:

• Collected time-series data from infants using artificial nipple-based sensors measuring tongue forces.
• Applied machine-learned computational modeling (MLCM) algorithms to analyze sensor data.
• Developed a classification system to identify distinctive infant sucking patterns.

Main Results:

• The best MLCM model achieved 90% accuracy, 80% recall, 100% precision, 0.90 f1-score, and 0.80 AUC.
• The system effectively extracted features and identified distinct sucking patterns.
• Demonstrated the feasibility of using sensor technology and ML for infant sucking analysis.

Conclusions:

• The proposed sensor-based MLCM system offers a reliable, objective method for assessing infant sucking patterns.
• This technology can serve as a valuable decision-support tool for clinicians.
• Early identification of unhealthy sucking patterns can lead to timely interventions and improved infant health outcomes.