One of the major concerns of recent studies is the correct discrimination between vegetative and minimally conscious state as the distinction between these two conditions has major implications for subsequent patient rehabilitation. In particular, it would be advantageous to establish communication with these patients. This work describes a procedure using EEG to detect brain responses to imagery instruction in patients with disorders of consciousness. Five healthy subjects and five patients with different disorders of consciousness took part in the study. A support vector machine classifier applied to EEG data was used to distinguish two mental tasks (Imagery Trial) and to detect answers to simple yes or no questions (pre-Communication Trial). The proposed procedure uses feature selection based on a nested-leave-one-out algorithm to reduce the number of electrodes required. We obtained a mean classification accuracy of 82.0% (SD 5.1%) for healthy subjects and 84.6% (SD 9.1%) for patients in the Imagery Trial, and a mean classification accuracy of 80.7% (SD 11.5%) for healthy subjects and 91.7% (SD 7.4%) for patients in the pre-Communication Trial. The subset of electrodes selected was subject and session dependent.

A Feasibility Study of an Improved Procedure for Using EEG to Detect Brain Responses to Imagery Instruction in Patients with Disorders of Consciousness

MANGIA, ANNA LISA;PIRINI, MARCO;CAPPELLO, ANGELO
2014

Abstract

One of the major concerns of recent studies is the correct discrimination between vegetative and minimally conscious state as the distinction between these two conditions has major implications for subsequent patient rehabilitation. In particular, it would be advantageous to establish communication with these patients. This work describes a procedure using EEG to detect brain responses to imagery instruction in patients with disorders of consciousness. Five healthy subjects and five patients with different disorders of consciousness took part in the study. A support vector machine classifier applied to EEG data was used to distinguish two mental tasks (Imagery Trial) and to detect answers to simple yes or no questions (pre-Communication Trial). The proposed procedure uses feature selection based on a nested-leave-one-out algorithm to reduce the number of electrodes required. We obtained a mean classification accuracy of 82.0% (SD 5.1%) for healthy subjects and 84.6% (SD 9.1%) for patients in the Imagery Trial, and a mean classification accuracy of 80.7% (SD 11.5%) for healthy subjects and 91.7% (SD 7.4%) for patients in the pre-Communication Trial. The subset of electrodes selected was subject and session dependent.
PLOS ONE
Anna Lisa Mangia; Marco Pirini; Laura Simoncini; Angelo Cappello
File in questo prodotto:
Eventuali allegati, non sono esposti

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: http://hdl.handle.net/11585/353833
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? 4
  • Scopus 6
  • ???jsp.display-item.citation.isi??? 3
social impact