DIGITAL LIBRARY
SPECTRUM-AGENTS: DECISION SUPPORT FOR AUTISM SPECTRUM DISORDER
Whitworth University (UNITED STATES)
About this paper:
Appears in: INTED2010 Proceedings
Publication year: 2010
Pages: 1556-1566
ISBN: 978-84-613-5538-9
ISSN: 2340-1079
Conference name: 4th International Technology, Education and Development Conference
Dates: 8-10 March, 2010
Location: Valencia, Spain
Abstract:
SPECTRUM-Agents is a suite of decision support tools providing comprehensive informatics for professional caregivers, educators and researchers of Autism Spectrum Disorder (ASD). The number of children diagnosed with ASD is escalating internationally at an alarming pace particularly evidenced in the United States, Europe and Japan. This debilitating neurological disorder impedes communication, social interaction and behavior. The disorder has far reaching impact for individuals, families, education and society.

Early identification and effective special education intervention is critical for best outcomes, yet challenges abound. Complexities include a variance of disorders with often overlapping diagnoses, diverse levels of severity, and dramatic variations of exhibited symptoms. Confusion surrounds the cause of the affliction and effectiveness of available treatments. Also problematic is that most current evaluative instruments are based on non-quantitative, functional observations. Emerging data repositories make meaningful analysis possible. Automated software poses only the viable means for analyzing varied dimensions, treatments and causes of ASD. SPECTRUM-Agents introduces analytical software designed specifically for ASD support and introduces distinctive processing approaches in so doing.

Spectrum-Agents presently includes three software modules in a highly scalable architecture. The Screener component identifies the possibility of ASD in young children. The second component provides comprehensive diagnostic evaluation and identifies patterns of the disorder. The Autism Spectrum InformaticS Tool (ASIST) tracks and correlates interventions with patient progress, thereby also identifying effective treatments. The system employs a backdrop of federated, collaborative agents and a hybrid inference model. Agents decompose complexity into dynamic processes acting as domain specialists and analysts. Layered fuzzy logic, belief networks and trend analysis perform reasoning in a domain characterized by uncertainty and partial belief demands.

SPECTRUM-Agents assists identification of ASD in infants and toddlers early on, encouraging advanced intervention services and thereby better outcomes as the child enters formal education. The system will eventually facilitate identification of items on diagnostic assessment instruments as the most accurate indicators of ASD. The ASIST tool further assists in providing an evidence base study of effective treatments, decreasing speculation and expense for special education. This paper introduces SPECTRUM-Agents, challenges, methodologies, features and details of the system modules.
Keywords:
Autism, decision support systems, autonomous agents, belief networks.