Harmonising neonatal EEG preprocessing through an EEG101 Virtual Mobility

Portrait of Jovana Kljajic — Jovana Kljajic, EEG101 Virtual Mobility grantee.

Virtual Mobility

Towards harmonised neonatal EEG preprocessing across clinical settings

Jovana Kljajic’s EEG101 Virtual Mobility brought researchers in Novi Sad and Cork together to compare neonatal EEG acquisition and preprocessing practices and develop an initial harmonised framework for heterogeneous clinical settings.

When: May–August 2026
Where: Online collaboration between Novi Sad, Serbia and Cork, Ireland
Host: INFANT Research Centre/CergenX, University College Cork
Working Groups: WG1, WG2

Why this mattered

Making neonatal EEG comparisons more robust

Neonatal EEG is collected in diverse clinical environments, using different acquisition systems, channel configurations, sampling frequencies, filtering practices, and approaches to signal-quality assessment. This Virtual Mobility examined how those differences shape preprocessing decisions and downstream comparisons, particularly where machine-learning approaches and cross-dataset analysis are envisaged.

What the collaboration did

From clinical observation to a shared framework

Regular online meetings brought together Jovana Kljajic and Prof. Tamara Škorić at the University of Novi Sad with Dr. John O’Toole and Dr. Robert Hogan at the INFANT Research Centre/CergenX, University College Cork. In parallel, clinical visits to the Neonatal Intensive Care Unit at the Institute for Child and Youth Health Care of Vojvodina informed the work with direct insight into artefacts encountered during routine aEEG monitoring.

Key outcomes

A comparative review of neonatal EEG acquisition hardware, channel configurations, sampling frequencies, filtering, and signal-quality assessment across participating settings
An initial harmonised preprocessing framework applicable to both single-channel and multichannel neonatal EEG recordings
Initial testing across a local aEEG dataset and the open-access Cork neonatal EEG dataset
Identification of clinically relevant artefacts, including movement, electrode displacement, and interference from clinical equipment
Methodological input to EEG101 Deliverables D1.1, D1.2, D2.1, and D2.2

People and places

Harmonised preprocessing workflow

Four-panel workflow diagram showing (1) heterogeneous EEG acquisition from two clinical settings in Serbia and Ireland, (2) acquisition characteristics and signal variability comparison, (3) the preliminary harmonised preprocessing framework covering subsampling, filtering, amplitude normalisation, artefact handling, and standardised output, and (4) initial evaluation results and future work. — Overview of the harmonised neonatal EEG preprocessing workflow developed during this Virtual Mobility, from heterogeneous clinical acquisition through to initial cross-site evaluation.

People and places

A continuing neonatal EEG collaboration

Jovana Kljajic, Faculty of Technical Sciences, University of Novi Sad
Prof. Tamara Škorić, Faculty of Technical Sciences, University of Novi Sad
Dr. John O’Toole, INFANT Research Centre/CergenX, University College Cork
Dr. Robert Hogan, INFANT Research Centre/CergenX, University College Cork
Institute for Child and Youth Health Care of Vojvodina, Neonatal Intensive Care Unit

What comes next

Further validation across datasets

The initial framework provides a methodological foundation rather than a finalised pipeline. Further validation will focus on signal-quality variability, artefact handling, and the effects of preprocessing choices on downstream neonatal EEG analyses.

Continue regular online collaboration and refine the framework with additional neonatal EEG datasets
Assess the effect of harmonised preprocessing on AI-based seizure-detection and outcome-prediction analyses
Prepare joint scientific publications arising from the collaboration

Explore further

Working Group 2 Cork neonatal EEG dataset

This activity was supported through an EEG101 Virtual Mobility grant under COST Action CA24148.