AGMA-PESS: a deep learning-based infant pose estimator and sequence selector software for general movement assessment
2024

AGMA-PESS: Software for Infant Movement Assessment

Sample size: 6 publication 10 minutes Evidence: high

Author Information

Author(s): Soualmi Ameur, Alata Olivier, Ducottet Christophe, Petitjean-Robert Anne, Plat Aurélie, Patural Hugues, Giraud Antoine

Primary Institution: Institut d’Optique Graduate School Université Jean Monnet Saint-Etienne

Hypothesis

Can deep learning improve the efficiency of general movement assessments in preterm infants?

Conclusion

The AGMA-PESS software effectively automates the selection of video sequences for general movement assessment, achieving high precision comparable to human experts.

Supporting Evidence

  • The AGMA-PESS software achieved a precision of 86% in selecting video sequences.
  • In 28 out of 30 cases, the software's selections intersected with those of human experts.
  • The software was developed using a dataset of over 88,000 images of infants.
  • Manual selection of sequences took an average of over 23 minutes, while the software automates this process.

Takeaway

This software helps doctors quickly find important videos of baby movements, making it easier to check if they are developing well.

Methodology

The software uses deep learning to estimate infant poses and automatically select video sequences for general movement assessment.

Limitations

The software may not generalize well outside the specific conditions of the study and cannot differentiate between movements caused by crying or fussing.

Participant Demographics

Six preterm infants born before 33 weeks of gestational age were included.

Statistical Information

P-Value

p<0.05

Statistical Significance

p<0.05

Digital Object Identifier (DOI)

10.3389/fped.2024.1465632

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication