Application of Statistical Analysis and Machine Learning to Identify Infants' Abnormal Suckling Behavior

Area of Science:

• Biomedical Engineering
• Infant Health
• Computational Biology

Background:

• Breastfeeding offers significant health benefits but often ceases prematurely.
• Current infant screening lacks objective measures for suckling abnormalities.
• Early identification of suckling issues is critical for successful breastfeeding.

Purpose of the Study:

• To develop and validate a computational method for identifying abnormal infant suckling behavior.
• To establish normative data for non-nutritive suckling parameters.
• To assess the utility of machine learning in early screening for breastfeeding complications.

Main Methods:

• Utilized a non-nutritive suckling vacuum measurement system on 91 healthy infants.
• Recorded non-nutritive suckling for 60 seconds to gather data.
• Applied Mahalanobis distance and K-nearest neighbor (KNN) algorithms to detect anomalies.

Main Results:

• Established normative data for key suckling parameters (vacuum, frequency, duration, etc.).
• Demonstrated the ability to detect abnormal suckling behavior using statistical analysis and machine learning.
• Case studies showed the impact of ankyloglossia on suckling patterns.

Conclusions:

• Statistical analysis and machine learning offer viable, rapid interpretation of infant suckling measurements.
• This digital suck assessment provides an objective, early screening method for abnormal infant suckling.
• The approach is crucial for identifying infants at risk of breastfeeding complications.