AGMA-PESS: a deep learning-based infant pose estimator and sequence selector software for general movement...

Area of Science:

• Neonatal Neuroscience
• Developmental Pediatrics
• Medical Imaging Analysis

Background:

• The General Movement Assessment (GMA) is crucial for evaluating brain maturation in preterm infants.
• Current GMA requires lengthy video recordings and manual selection of movement sequences, hindering clinical use.
• Accurate infant pose estimation is vital for automating GMA.

Purpose of the Study:

• To introduce the AGMA Pose Estimator and Sequence Selector (AGMA-PESS) software.
• To automate the selection of relevant video sequences for GMA.
• To enable 2D pose estimation of preterm infants for GMA.

Main Methods:

• Utilized a state-of-the-art deep learning infant pose estimation network.
• Developed software (AGMA-PESS) for automatic sequence selection and 2D pose estimation.
• Tested on preterm infants at preterm and writhing ages.

Main Results:

• AGMA-PESS successfully automates the selection of video sequences for GMA.
• The software provides accurate 2D pose estimation for infants.
• Demonstrated simplicity and efficiency in processing video data.

Conclusions:

• AGMA-PESS significantly reduces the time and effort required for GMA analysis.
• The software facilitates wider implementation of GMA in Neonatal Units.
• AGMA-PESS supports both clinical practice and research in infant development.