Machine Learning to Support Visual Inspection of Data: A Clinical Application
Tessa Taylor1,2, Marc J Lanovaz3
1University of Canterbury, Christchurch, New Zealand.
Behavior Modification
|August 12, 2021
Summary
Machine learning offers a new way to analyze pediatric feeding treatment effectiveness, supporting decisions made by visual analysts. This approach enhances the reliability of single-case experimental designs in behavioral interventions.
Area of Science:
- Behavioral Science
- Machine Learning Applications
- Pediatric Feeding Interventions
Background:
- Pediatric feeding programs often use single-case experimental designs (SCEDs) with visual inspection for treatment decisions.
- Current visual inspection methods for SCEDs lack consensus and can be subjective.
- There is a need for objective methods to support treatment evaluation in pediatric feeding.
Observation:
- A 5-year-old male with autism spectrum disorder (ASD) participated in a 2-week behavior-analytic feeding treatment.
- A modified reversal design was used to evaluate treatment effects.
- Machine learning (ML) was applied to analyze treatment effects and compared with expert visual analysis.
Findings:
- High interrater agreement was observed between the ML model and expert visual analysts regarding treatment effectiveness.
- The ML model generally agreed with visual analysts' conclusions on treatment efficacy.
- ML analysis provided consistent results with traditional visual inspection methods.
Implications:
- Machine learning can serve as a valuable tool to objectively support the analysis of SCEDs in pediatric feeding.
- This technology may improve the reliability and consistency of treatment decisions in applied behavior analysis.
- Integrating ML into feeding interventions could enhance evidence-based practices for children with feeding difficulties.
