System Level Framework for Assessing the Accuracy of Neonatal EEG Acquisition

Significant research has been conducted in recent years to design low-cost alternatives to the current EEG monitoring systems used in healthcare facilities. Testing such systems on a vulnerable population such as newborns is complicated due to ethical and regulatory considerations that slow down the technical development. This paper presents and validates a method for quantifying the accuracy of neonatal EEG acquisition systems and electrode technologies via clinical data simulations that do not require neonatal participants. The proposed method uses an extensive neonatal EEG database to simulate analogue signals, which are subsequently passed through electrical models of the skin-electrode interface, which are developed using wet and dry EEG electrode designs. The signal losses in the system are quantified at each stage of the acquisition process for electrode and acquisition board losses. SNR, correlation and noise values were calculated. The results verify that low-cost EEG acquisition systems are capable of obtaining clinical grade EEG. Although dry electrodes result in a significant increase in the skin-electrode impedance, accurate EEG recordings are still achievable.

I. Introduction

Clinical evaluation of electroencephalography (EEG) is vital across neonatal, paediatric and adult patients presenting abnormal neurological behaviour. Previous publications have shown that using EEG monitoring to aid the diagnosis of neonatal seizures drastically improves the percentage of correct seizure diagnoses compared to diagnosing seizures based on clinical signs alone [1]. However, the equipment and expertise required to conduct EEG monitoring is expensive and scarcely available. Factors such as size, cost, preparation time and interpretation skills inhibit the use of EEG monitoring in a range of cases and settings. Several attempts have been made to rectify this [2], however such devices have yet to be adopted into common clinical practice.