Skip to main content
AAPM&R Events Main banner
JFK Johnson
Icon Legend
This session is not in your schedule.
This session is in your schedule. Click again to remove it.
Presentation Icons
In-Person
In-Person
On Demand
On Demand
Live Stream
Live Stream
Invite Only
Invite Only
Ticketed
Ticketed
Coffee Break
Coffee Break
Meal
Meal
Reception
Reception
Back

Pediatric Rehabilitation

(Scientific Session) Development of Novel Dysphagia Evaluation via Video Analysis Based on Deep Learning Method in Neonates and Infants: A Preliminary Study

Saturday, November 18, 2023
9:15 AM – 10:30 AM CT
Location: CC 216

    Primary Author(s)

  • Yong Jae Na, MD

    Fellowship doctor
    Asan Medical Center
    Song Pa, Seoul-t'ukpyolsi, Republic of Korea

    • Co-Author(s)

  • Hyun Ah Lee, MD

    Fellowship doctor
    n/a
    Songpa-gu, Seoul-t'ukpyolsi, Republic of Korea

  • Eun Jae Ko, MD, PhD

    assistant professor
    Asan Medical Center
    Songpa-gu, Seoul-t'ukpyolsi, Republic of Korea

  • GH

    Garam Hong, MD

    resident
    n/a
    Songpagu, Seoul-t'ukpyolsi, Republic of Korea

  • JK

    Jihoon Kweon, PhD

    Associate professor
    Asan Medical Center
    Seoul, Seoul-t'ukpyolsi, Republic of Korea

Background and/or Objectives: Dysphagia is common in newborns with varying risk factors, but video fluoroscopic swallowing study (VFSS) is not easily applicable due to radiation and cooperation issues. Neonatal Oral-Motor Assessment Scale (NOMAS) is a suitable alternative for infants under 48 weeks post menstrual age. This study proposes a new evaluation method for dysphagia using deep learning-based AI by analyzing facial landmarks related to swallowing while videotaping bottle feeding in newborns and infants, based on the NOMAS evaluation methodology.

Design: After taking a video of a bottle feeding, a NOMAS evaluation was performed based on the footage.

Setting: The section of footage where the abnormality on NOMAS evaluation was most prominent was selected, then anatomical structures related to swallowing were mapped in each frame, and deep learning-based video analysis was performed to analyze the correlation between this result and the abnormality in NOMAS evaluation.

Participants: Infants under 2 months of correction age, clinically presumed to have dysphagia, who visited the Department of Pediatric Rehabilitation Medicine at single center from December 2021 to February 2022.

Interventions: not applicable.

Main Outcome Measures: The first case had excessive and wide jaw excursion, correlated by deep learning analysis of lip vertices. The second case showed minimal jaw excursion and tongue retraction during bottle sucking, also correlated by deep learning analysis of chin movement.

Results: By deep learning-based AI analysis through recording and mapping of bottle-feeding video, it was possible to obtain correlating data to some evaluation items of NOMAS.

Conclusions: This finding may serve as the basis for the development of a novel dysphagia evaluation method using AI-based video analysis, which can overcome the limitations of traditional methods.