Robust Neonatal Eeg Seizure Detection Through Adaptive Background Modeling

Abstract

Adaptive probabilistic modelling of the EEG background is proposed for seizure detection in neonates with hypoxic ischemic encephalopathy. The decision is made based on the temporal derivative of the seizure probability with respect to the adaptively modeled level of background activity. The robustness of the system to long duration 'seizure-like' artifacts, in particular those due to respiration, is improved. The system was developed using statistical leave-one-patient-out performance assessment, on a large clinical dataset, comprising 38 patients of 1479 hours total duration. The developed technique was then validated by a single test on a separate totally unseen randomized prospective dataset of 51 neonates totaling 2540 hours of duration. By exploiting the proposed adaptation, the ROC area is increased from 93.4% to 96.1% (41% relative improvement). The number of false detections per hour is decreased from 0.42 to 0.24, while maintaining the correct detection of seizure burden at 70%. These results on the unseen data were predicted from the rigorous leave-one-patient-out validation and confirm the validity of our algorithm development process.

Toward a Personalized Real-Time Diagnosis in Neonatal Seizure Detection

“…In fact, the LOO performance estimated on this dataset was shown to closely match the performance which was independently assessed on a separate large clinical dataset, as reported in [14] and [18].…”

“…The relative improvement is calculated as (AUC90PA-FUSION – AUC90PI-FUSION)/ (1 – AUC90PI-FUSION). The statistically significant difference at \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$\alpha $ \end{document} set to 1% as in [14] is indicated with the asterisk.

FIGURE 6.Relative improvement in AUC90 between the PA-FUSION and PI-FUSION systems.

…”

A Longitudinal EEG Study of Alzheimer's Disease Progression Based on A Complex Network Approach

“…The approach proposed in this work can be applied to other neurological pathologies. 23,[45][46][47] Appendix A. MI Derivation MI 25 has been originally proposed as a quantity that measures the level of dependency existing between two random variables, say X and Y , assumed to be continuous (in this work, a discrete version of it has been used). MI can be expressed, in terms of the estimated density probability functions p(x), p(y), and p(x, y) of the random variables X and Y , as follows:…”

A new music-empirical wavelet transform methodology for time–frequency analysis of noisy nonlinear and non-stationary signals

“…Many physical signals include significant noise and closely-spaced frequencies that cannot be effectively analyzed using these methods [12]. Further, real-life time series signals such as biomedical signals [38,43,30,48,8,2,[9][10][11]6,7], or vibration signals obtained from civil structure subjected to dynamic excitations [33,14,26,13,32,27] include nonlinear and non-stationary properties that cannot be adequately analyzed using these methods. In order to overcome these limitations, time-frequency representation (TFR) has become a good alternative to analyze nonlinear and non-stationary signals since a TFR can provide information about the frequencies contained in signal over time [40].…”