Video Dataset for Respiratory Rate Counting in Children
Description
This dataset contains video recordings of chest movements of children under five years of age, captured for the purpose of respiratory rate (RR) assessment. The children presented at three different level of hospitals in Bangladesh with respiratory symptoms such as cough or breathing difficulty. Each child is assigned a unique child ID and has one associated video file in MP4 format, which visually documents their chest movements for RR counting. In addition to the video files, the dataset includes a Stata data file (.dta) containing complementary metadata. This file provides information on each child's demographic characteristics, reported clinical symptoms (e.g., cough, difficulty breathing), and respiratory rate as assessed by a trained video expert panel. This dataset supports research on automated or manual respiratory rate counting, clinical diagnostics, and pediatric respiratory health.
Files
Steps to reproduce
Study Design and Setting: This was a cross-sectional study conducted at the Institute of Child and Mother Health in Dhaka, Zakiganj sub-district hospital, and three selected community clinics (CCs) in Zakiganj sub-district, Sylhet, Bangladesh. Study Population: The study enrolled: -Infants aged 0–2 months presenting with any illness. -Children aged 2–59 months presenting with cough and/or difficulty breathing. -Children exhibiting any danger signs were excluded. Data Collection – Video Recording: Chest movement videos were recorded using a Canon EOS M50 digital camera. Recording began once the child was calm and breathing naturally. During the recording, a physician/health worker simultaneously counted the child’s respiratory rate (RR) for one minute. Each child had one video recorded. Data Management: Recorded videos were transferred to a password-protected computer. Videos were edited to remove unnecessary footage while preserving RR-relevant segments. Each video was anonymized and tagged with a unique child ID. Respiratory Rate Interpretation Process: Videos were randomly assigned to two expert paediatricians or Video Expert Panel (VEP) members for RR interpretation. If both reviewers agreed a video was uninterpretable, it was excluded. If the RR difference between reviewers was ≤2 breaths per minute (bpm), the average RR was recorded as final. If the RR difference was >2 bpm or there was disagreement on interpretability, the video was reviewed by a third panel member. If no two panel members agreed within ≤2 bpm, the video was reviewed by a fourth panel member (where applicable). If the fourth member agreed with any previous member, the average RR of the agreeing pair was taken as the final value. Videos with no agreement among any two reviewers were excluded from analysis. Data Structure: Each video corresponds to a single child. A separate Stata (.dta) file includes: -Child demographics -Clinical symptoms (e.g., cough, difficulty breathing) -Final RR value as determined by expert/VEP consensus
Institutions
- The University of Edinburgh