You are here

NORC News

Research conducted by UAB and UA wins award in eHealth/mHealth Section of TOS Abstract Competition

November 20, 2015

Quantifying Infant Feeding Behavior: A Sensor Based Approach"—a joint venture between the University of Alabama and the University of Alabama at Birmingham’s Nutrition Obesity Research Center (NORC)—has been selected for an award in the eHealth/mHealth Section of The Obesity Society (TOS) Abstract Competition. UAB’s Paula C. Chandler-Laney, PhD, assistant professor in the Department of Nutrition Sciences and the NORC, collaborated with UA’s Edward Sazonov, PhD, associate professor, and Muhammad Farooq, PhD student, in the Department of Electrical and Computer Engineering, and Maria Hernandez-Reif, PhD, professor in the Department of Human Development & Family Studies and director of the Pediatric Development Research Lab, to conduct the research.

UAB’s NORC funded the study to examine whether a food intake sensor originally developed to detect and characterize eating episodes in adults can be used to measure sucking behavior in infants. The prizewinning abstract shows that regardless of whether breast- or bottle-fed, the proposed sensor and algorithm may be used to estimate the feeding rate of infants. These findings suggest that the jaw sensor has potential to provide an efficient, objective method to assess eating behaviors like sucking rate, meal duration, and meal frequency. This would be an improvement over currently used methods such as parental surveys which may be subject to bias, and weighed and timed test meals which are burdensome to the caregiver.

The researchers state, “This work presents the use of a piezoelectric film sensor for monitoring of jaw movements during sucking episodes to quantify infants’ sucking behaviors in terms of sucking counts. Meals for a cohort consisting of 6 breast-fed and bottle-fed infants were videotaped and synchronized with sensor signals. Videos and sensor signals were divided into 10 second epochs, which were annotated by two human raters for sucking counts. Sensor signals were normalized to account for amplitude variation among infants. A peak detection algorithm was used to compute sucking count where peaks were only considered above a certain threshold based on 80th percentile of the signal amplitude. A leave one out cross validation approach gave a mean absolute error of 7.11 percent between the average sucking counts of human raters and algorithm estimated sucking count. For a two-way mixed model the intra-class correlation coefficient [ICC] between the two human raters was 0.98, whereas between human and algorithm ICC was 0.92.”

Study findings indicate that the sensor and algorithm can reliably estimate infant sucking counts.