Reducing Brain Volume Measurement Errors by Performing Intra-rater and Inter-rater Reliability Assessments

Abstract

By Julia Nail, Nursing

Advisor: Julia Kline

Abstract: The goal of the Early Prediction Study is to enable the early identification of neurodevelopmental impairments (such as cerebral palsy) in very premature infants. Early identification of these impairments can result in earlier diagnosis and targeted interventions, which will potentially improve the long-term quality of life for very preterm infants. At the Parikh Lab, I measured regional anatomical brain volumes on neonatal MRI scans using ONIS, a computer software. The purpose of this was to investigate a potential correlation between the size of specific brain structures in preterm infants and their risk of developing neurodevelopmental impairments. Since the measurement of these MRI scans is somewhat subjective, I calculated my intra-rater and inter-rater reliability (with another rater) through a statistical program called Stata. Intra-rater reliability is a comparison of my previous measurements against my new measurements of the same set of scans, measured several weeks apart. Inter-rater reliability is the comparison of my measurements for a set of scans against another person's measurements of the same set of scans. Ideally, both intra-rater and inter-rater reliability should be high (>0.95), so that the results from of these scans have minimal measurement error and can be replicated. My intra-rater reliability values ranged from 0.92-0.99, excluding the corpus callosum measurements.