Utilising Cot-Side Cameras in Neonatal Intensive Care Unit for Deep Learning-Assisted General Movement Assessment.

Abstract

Aim: Neonatal units are increasingly utilising cot-side cameras to connect parents with their infants. Combined with deep learning, video obtained through cot-side cameras could assist clinicians in conducting seamless general movement assessment (GMA) of the writhing age.

Method: A literature search was conducted using PubMed, Embase and SCOPUS with the following keywords: cot-side cameras, deep learning, artificial intelligence, general movement assessment and writhing age.

Results: Methods for acquiring and classifying human movement are categorised into contact, non-contact and hybrid approaches. Contact modalities typically include wearable sensors placed on the body to represent human posture, while hybrid modalities combine wearable sensors or markers with non-contact sensors. Non-contact approaches include radar-based and vision-based methods, which are the most common and accessible for motion capture, employing standard or specialised cameras to capture video data. Cot-side cameras used in neonatal clinics are primarily standard red-green-blue (RGB) devices and are the leading candidates for automated GMA. Advances in deep learning can enhance motion assessment with video data through appearance- and pose-based methods, supporting computer-aided GMA.

Conclusion: Advances in deep learning can enhance the motion assessment of RGB video data, offering a scalable and non-invasive solution for computer-aided GMA that could reshape early neurodevelopmental screening.