A Predictive Model for Paediatric Autism Screening

Benjamin Wingfield, Shane Miller, Y Pratheepan, Dermot Kerr, Bryan Gardiner, Sudarshi Seneviratne, Pradeepa Samarasinghe, Sonya Coleman

Research output: Contribution to journalArticle

59 Downloads (Pure)

Abstract

Autism Spectrum Disorder (ASD) is an umbrella term for a group of neurodevelopmental disorders that is associated with impairments to social interaction, communication, and behaviour. Typically, ASD is first detected with a screening tool (e.g. M-CHAT). However, the interpretation of ASD behavioural symptoms varies across cultures: the sensitivity of M-CHAT is as low as 25% in Sri Lanka. A culturally-sensitive screening tool called Pictorial Autism Assessment Schedule (PAAS) has overcome this problem. Low and Middle-Income Countries (LMIC) have a shortage of mental health specialists, which is a key barrier for obtaining an early ASD diagnosis. Early identification of ASD enables intervention before atypical patterns of behaviour and brain function become established. This article proposes a culturally sensitive ASD screening mobile application. The proposed application embeds an intelligent machine learning model and uses a clinically validated symptom checklist to monitor and detect ASD in LMIC for the first time. Machine learning models were trained on clinical PAAS data and their predictive performance evaluated, which demonstrated that the Random Forest was the optimal classifier (AUROC 0.98) for embedding into the mobile screening tool. Additionally, feature selection demonstrated that many PAAS questions are redundant, and can be removed to optimise the screening process.
Original languageEnglish
JournalHealth Informatics Journal
Publication statusAccepted/In press - 3 Oct 2019

Keywords

  • decision-support systems
  • machine learning
  • Autism Spectrum Disorder (ASD)
  • Data Mining
  • Computer aided screening tool

Fingerprint Dive into the research topics of 'A Predictive Model for Paediatric Autism Screening'. Together they form a unique fingerprint.

  • Cite this