Computer Vision to Automatically Assess Infant Neuromotor Risk.

Claire Chambers,Helen Loeb,Rachit Saluja,Nidhi Seethapathi,Konrad P Kording,Samuel R Pierce,Daniel K Bogen,Laura Prosser,Michelle J Johnson

doi:10.1109/tnsre.2020.3029121

Abstract

An infant's risk of developing neuromotor impairment is primarily assessed through visual examination by specialized clinicians. Therefore, many infants at risk for impairment go undetected, particularly in under-resourced environments. There is thus a need to develop automated, clinical assessments based on quantitative measures from widely-available sources, such as videos recorded on a mobile device. Here, we automatically extract body poses and movement kinematics from the videos of at-risk infants (N = 19). For each infant, we calculate how much they deviate from a group of healthy infants (N = 85 online videos) using a Naïve Gaussian Bayesian Surprise metric. After pre-registering our Bayesian Surprise calculations, we find that infants who are at high risk for impairments deviate considerably from the healthy group. Our simple method, provided as an open-source toolkit, thus shows promise as the basis for an automated and low-cost assessment of risk based on video recordings.

Highlights

Developmental disorders, including those caused by neuromotor disease, are the most common source of childhood disability, affecting 5-10% of children and 3.7 to 7.4 million American children (Rydz, Shevell, Majnemer, & Oskoui, 2005) and are often the cause of lifelong disability
Compare Bayes Surprise to clinical judgment infant deviates from the typical movements of healthy infants using a single score, the Naïve Gaussian Bayesian Surprise (Lonini et al, 2016). When we tested this system on a clinical population (N=19) where the level of neuromotor risk was assessed by a clinician, we found that Bayesian Surprise varied across participant groups
Our approach was to compare infants assessed in the laboratory at different levels of neuromotor risk with a normative dataset of infant movement extracted from online videos

Summary

Introduction

Developmental disorders, including those caused by neuromotor disease, are the most common source of childhood disability, affecting 5-10% of children and 3.7 to 7.4 million American children (Rydz, Shevell, Majnemer, & Oskoui, 2005) and are often the cause of lifelong disability. It is desirable that tests are based on quantitative measurements and a series of well-defined steps to reach its conclusion, i.e., an algorithm. This motivates the development of quantitative tests to supplement clinical judgment. Clinical assessments often involve expert judgment and expensive equipment, which are only available in highly-resourced environments. This makes assessment inaccessible for families of limited means and in lowresource countries, where the burden of disability is higher (World Health Organization, 2011). It is important to assess diagnostic tests by their accuracy, and by how quantitative and cost effective they may be

Objectives

Methods

Results

Conclusion