Comparison of Multi-class Machine Learning Methods for the Identification of Factors Most Predictive of Prognosis in Neurobehavioral assessment of Pediatric Severe Disorder of Consciousness through LOCFAS scale

Severe Disorders of Consciousness (DoC) are generally caused by brain trauma, anoxia or stroke, and result in conditions ranging from coma to the confused-agitated state. Prognostic decision is difficult to achieve during the first year after injury, especially in the pediatric cases. Nevertheless,...

Full description

Saved in:
Bibliographic Details
Main Authors: Molteni, Erika, Colombo, Katia, Beretta, Elena, Galbiati, Susanna, Santos Canas, Liane Dos, Modat, Marc, Strazzer, Sandra
Format: Conference Proceeding
Language:English
Subjects:
Brain modeling
Injuries
Machine learning
Prognostics and health management
Radio frequency
Training
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Severe Disorders of Consciousness (DoC) are generally caused by brain trauma, anoxia or stroke, and result in conditions ranging from coma to the confused-agitated state. Prognostic decision is difficult to achieve during the first year after injury, especially in the pediatric cases. Nevertheless, prognosis crucially informs rehabilitation decision and family expectations. We compared four multi-class machine learning classification approaches for the prognostic decision in pediatric DoC. We identified domains of a neurobehavioral assessment tool, Level of Cognitive Functioning Assessment Scale, mostly contributing to decision in a cohort of 124 cases. We showed the possibility to generalize to new admitted pediatric cases, thus paving the way for real employment of machine learning classifiers as an assistive tool to prognostic decision in clinics.
ISSN:1558-4615
2694-0604
DOI:10.1109/EMBC.2019.8856880