Objective: This article proposes a method to automatically identify and label event-related potential (ERP) components with high accuracy and precision. Methods: We present a framework, referred to as peak-picking Dynamic Time Warping (ppDTW), where a priori knowledge about the ERPs under investigation is used to define a reference signal. We developed a combination of peak-picking and Dynamic Time Warping (DTW) that makes the temporal intervals for peak-picking adaptive on the basis of the morphology of the data. We tested the procedure on experimental data recorded from a control group and from children diagnosed with developmental dyslexia. Results: We compared our results with the traditional peak-picking. We demonstrated that our method achieves better performance than peak-picking, with an overall precision, recall and F-score of 93%, 86% and 89%, respectively, versus 93%, 80% and 85% achieved by peak-picking. Conclusion: We showed that our hybrid method outperforms peak-picking, when dealing with data involving several peaks of interest. Significance: The proposed method can reliably identify and label ERP components in challenging event-related recordings, thus assisting the clinician in an objective assessment of amplitudes and latencies of peaks of clinical interest.
Automated identification of ERP peaks through Dynamic Time Warping: an application to developmental dyslexia / S. Assecondi, A.M. Bianchi, H. Hallez, S. Staelens, S. Casarotto, I. Lemahieu, G.A. Chiarenza. - In: CLINICAL NEUROPHYSIOLOGY. - ISSN 1388-2457. - 120:10(2009), pp. 1819-1827.
Automated identification of ERP peaks through Dynamic Time Warping: an application to developmental dyslexia
S. Casarotto;
2009
Abstract
Objective: This article proposes a method to automatically identify and label event-related potential (ERP) components with high accuracy and precision. Methods: We present a framework, referred to as peak-picking Dynamic Time Warping (ppDTW), where a priori knowledge about the ERPs under investigation is used to define a reference signal. We developed a combination of peak-picking and Dynamic Time Warping (DTW) that makes the temporal intervals for peak-picking adaptive on the basis of the morphology of the data. We tested the procedure on experimental data recorded from a control group and from children diagnosed with developmental dyslexia. Results: We compared our results with the traditional peak-picking. We demonstrated that our method achieves better performance than peak-picking, with an overall precision, recall and F-score of 93%, 86% and 89%, respectively, versus 93%, 80% and 85% achieved by peak-picking. Conclusion: We showed that our hybrid method outperforms peak-picking, when dealing with data involving several peaks of interest. Significance: The proposed method can reliably identify and label ERP components in challenging event-related recordings, thus assisting the clinician in an objective assessment of amplitudes and latencies of peaks of clinical interest.Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.