OBJECTIVE: To evaluate, using receiver operation characteristic (ROC) curves, the performance of several statistical indicators in the objective detection of 80 Hz steady state auditory evoked responses. DESIGN: Steady state auditory evoked responses elicited by amplitude modulated tones of 500 and 1000 Hz, were obtained in 16 normal adults. Recordings were made at intensities ranging from 80 to 30 dB SPL and without stimulation. Four statistics: coherence synchrony measure, circular T2, a new variant of Hotelling T2 (labeled HT2) and a test for hidden periodicity (F test) were calculated. The statistics were compared using ROC curves and bootstrapping techniques. Two outcome measures were considered: behavioral threshold prediction and averaging efficiency. RESULTS: All indicators were highly accurate to detect a response (8 to 9 dB above mean behavioral threshold for the 1000 Hz and 14 to 16 dB for the 500 Hz carrier). Responses could be reliable detected after averaging about five individual epochs of long duration. No statistically significant differences were evidenced though in their capability to predict behavioral threshold or their averaging efficiency. CONCLUSIONS: Despite the more adequate statistical properties of some of these indicators no significant differences were found in their performance. Thus all of these indicators could be recommended for automatic detection of 80 Hz auditory steady state responses.
OBJECTIVE: To evaluate, using receiver operation characteristic (ROC) curves, the performance of several statistical indicators in the objective detection of 80 Hz steady state auditory evoked responses. DESIGN: Steady state auditory evoked responses elicited by amplitude modulated tones of 500 and 1000 Hz, were obtained in 16 normal adults. Recordings were made at intensities ranging from 80 to 30 dB SPL and without stimulation. Four statistics: coherence synchrony measure, circular T2, a new variant of Hotelling T2 (labeled HT2) and a test for hidden periodicity (F test) were calculated. The statistics were compared using ROC curves and bootstrapping techniques. Two outcome measures were considered: behavioral threshold prediction and averaging efficiency. RESULTS: All indicators were highly accurate to detect a response (8 to 9 dB above mean behavioral threshold for the 1000 Hz and 14 to 16 dB for the 500 Hz carrier). Responses could be reliable detected after averaging about five individual epochs of long duration. No statistically significant differences were evidenced though in their capability to predict behavioral threshold or their averaging efficiency. CONCLUSIONS: Despite the more adequate statistical properties of some of these indicators no significant differences were found in their performance. Thus all of these indicators could be recommended for automatic detection of 80 Hz auditory steady state responses.
Authors: Julian Jenkins; Ariane E Rhone; William J Idsardi; Jonathan Z Simon; David Poeppel Journal: Brain Topogr Date: 2011-03-06 Impact factor: 3.020