迈向可扩展的新生儿筛查：非受控环境下的自动化整体运动评估
机器学习
2025-08-08 v4 计算机视觉与模式识别
摘要
整体运动（GMs）是婴儿自发的、协调的身体运动，能为发育中的神经系统提供宝贵见解。通过 Prechtl 整体运动评估（GMA）进行评价，GMs 是神经发育障碍的可靠预测指标。然而，GMA 需要专门训练的临床医生，而这类人员数量有限。为扩大新生儿筛查规模，需要一种算法能从婴儿视频记录中自动分类 GMs。此类数据面临挑战，包括记录时长、设备类型和环境的差异，每个视频仅有整体运动质量的粗略标注。本文介绍了一种从这些记录中提取特征的工具，并探索了多种机器学习技术用于自动化 GM 分类。
引用
@article{arxiv.2411.09821,
title = {Towards Scalable Newborn Screening: Automated General Movement Assessment in Uncontrolled Settings},
author = {Daphné Chopard and Sonia Laguna and Kieran Chin-Cheong and Annika Dietz and Anna Badura and Sven Wellmann and Julia E. Vogt},
journal= {arXiv preprint arXiv:2411.09821},
year = {2025}
}
备注
Paper at Proceedings of Machine Learning Research 298 1 22, 2025 Machine Learning for Healthcare. Iterations of previous versions accepted as oral and best paper award at ICLR 2025 Workshop on AI for Children and at the Findings track of the Machine Learning for Health (ML4H) symposium 2024