Development of Dysphagia Evaluation Via Video Analysis Based on Deep Learning Method in Neonates and Infants and Correlation Between the Evaluation and the Development

Development of Dysphagia Evaluation Via Video Analysis Based on Deep Learning Method in Neonates and Infants and Correlation Between the Evaluation and the Development

Development of Dysphagia Evaluation Via Video Analysis Based on Deep Learning Method in Neonates and Infants and Correlation Between the Evaluation and the Development

The purpose of this study is to develop an evaluation of dysphagia through deep learning-based video analysis in newborns and infants, and to report the correlation with future development.

Although the exact frequency of dysphagia in newborns is not known, according to a paper published by Motion et al. in 2001, the prevalence of eating problems in premature infants under 37 weeks of age was 10.5%, and in 2001, Mercado-Deane et al. reported that about 26% had dysphagia. In 1996, Reilly et al. reported that more than 90% of children with polio had oral movement disorders and 38% had dysphagia. It is known that the frequency of dysphagia in newborns and infants is not low. The risk factors that cause these dysphagia are very diverse, and dysphagia in children can be induced by causes that can affect the whole process of swallowing. The evaluation of dysphagia can be divided into an evaluation method that does not use an instrument such as SOMA, SDS, and quality of life measurement, and an evaluation method that uses an instrument such as VFSS and FEES. However, it is difficult to conduct tests such as VFSS in newborns and infants due to poor coordination, and there is also a risk of radiation exposure. In addition, there are practical difficulties in applying the evaluation in all medical institutions because special facilities are required to implement VFSS and specialized clinical personnel are required. The Neonatal Oral-Motor Assessment Scale (NOMAS), developed by Marjorie Meyer Palmer in 1983, is an evaluation method for dysphagia applicable to infants under 48 weeks of PMA that evaluates whether there are abnormal findings by observing sucking for 5 minutes. However, NOMAS has the disadvantage that a person has to directly observe the sucking and may show differences in results between raters. Therefore, in this study, inspired by the evaluation method of NOMAS, investigators try to develop an evaluation method to evaluate swallowing disorders by videotaping bottle feeding in newborns and infants, then calculating and analyzing the features of the baby's face related to swallowing with artificial intelligence. investigators would like to analyze the relationship between this evaluation result and future development.

For newborns and infants who satisfy the inclusion and exclusion criteria, after taking a video of a bottle feeding, we try to develop an evaluation of swallowing disorder through artificial intelligence-based analysis. The developed evaluation will be verified for validity by comparing it with NOMAS (and VFSS if possible), and the correlation with future development will be analyzed through the relationship with the 1st and 2nd year correctional Bailey Developmental Evaluation.

Bottle feeding filming - With a camera that can measure depth, take a video of feeding milk from a bottle for 5-10 minutes. NOMAS (Neonatal Oral-Motor Assessment Scale) evaluation Evaluate NOMAS with bottle feeding video by certified evaluators. NOMAS, developed by Marjorie Meyer Palmer in 1983, is an evaluation method for dysphagia applicable to infants under 48 weeks of PMA. Implementation of video analysis using deep learning (A) Analysis method - Implement a deep learning network that sets keypoints of faces and environments related to swallowing and detects specified keypoints in units of frames. (B) Analyze the movement of key points, and figure out the correlation between swallowing states. (C) Create an input/output process that can automatically classify swallowing states with video as input.

Inclusion Criteria: Newborns and infants under 6 months of correctional age Clinically suspected to have dysphagia In case parental consent is obtained Exclusion Criteria: In case of cardiopulmonary instability (ECMO, Ventilator, etc. are being applied) When it is not possible to secure sufficient fields for artificial intelligence analysis of images due to factors such as anatomical abnormalities of the head and neck