Automatic Nonnutritive Suck Waveform Discrimination and Feature Extraction in Preterm Infants.

Area of Science:

• Neonatal Physiology
• Computational Neuroscience
• Medical Software Development

Background:

• Nonnutritive suck (NNS) patterns in preterm infants indicate brain integrity and are crucial for assessing feeding readiness and oromotor development in the neonatal intensive care unit (NICU).
• Existing methods lack integrated software for comprehensive NNS signal preprocessing, waveform discrimination, feature detection, and batch processing of large datasets across multiple NICU sites.

Purpose of the Study:

• To develop and describe NeoNNS, a cross-platform graphical user interface (GUI) and terminal application for analyzing nonnutritive suck (NNS) compression pressure waveforms.
• To enable time-series and frequency-domain analyses of NNS data for improved understanding of infant feeding readiness and oromotor development.

Main Methods:

• NeoNNS was developed using Python and the Tkinter GUI package, incorporating a signal-processing pipeline with low-pass filtering, baseline correction, peak detection, and burst classification.
• The software provides data visualizations and parametric analyses, including time- and frequency-domain views, spatiotemporal index, and hierarchical cluster analysis for modeling feeding readiness.

Main Results:

• NeoNNS successfully processed 568 NNS assessment files from 30 extremely preterm infants in batch mode, generating time- and frequency-domain analyses of NNS pressure waveforms.
• NNS cycle discrimination and burst classification quantified NNS waveform features relative to postmenstrual age, with hierarchical cluster analysis effectively labeling NNS records for feeding readiness.

Conclusions:

• NeoNNS offers a versatile platform for quantifying NNS development dynamics in time and frequency domains at the bedside, supporting individual patient monitoring and multi-site big data analytics.
• The hierarchical cluster feature analysis within NeoNNS facilitates the modeling of oral feeding readiness based on quantitative NNS compression pressure waveform features.