Marina Diachenko

Marina Diachenko

Auto-generated profile, based on publicly available information

Marina Diachenko

Department of Integrative Neurophysiology, Center for Neurogenomics and Cognitive Research, Amsterdam Neuroscience, Vrije Universiteit Amsterdam , Amsterdam 1081 HV

Research Focus & Metrics

Academic Career Length

5Years (2022-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
27 September 2022 in eneuro
Verified
Journal Article
Open access
PDF file
MDMarina Diachenko
Marina Diachenko
SHSimon J. Houtman
Simon J. Houtman
Erika L Juarez MartinezErika L Juarez Martinez
Erika L Juarez Martinez
JRJennifer R. Ramautar
Jennifer R. Ramautar
RWRobin Weiler
Robin Weiler
KLKlaus Linkenkaer-Hansen
Klaus Linkenkaer-Hansen
The development of validated algorithms for automated handling of artifacts is essential for reliable and fast processing of EEG signals. Recently, there have been methodological advances in designing machine-learning algorithms to improve artifact detection of trained professionals who usually meticulously inspect and manually annotate EEG signals. However, validation of these methods is hindered by the lack of a gold standard as data are mostly private and data annotation is time consuming and error prone. In the effort to circumvent these issues, we propose an iterative learning model to speed up and reduce errors of manual annotation of EEG. We use a convolutional neural network (CNN) to train on expert-annotated eyes-open and eyes-closed resting-state EEG data from typically developing children (n= 30) and children with neurodevelopmental disorders (n= 141). To overcome the circular reasoning of aiming to develop a new algorithm and benchmarking to a manually-annotated gold standard, we instead aim to improve the gold standard by revising the portion of the data that was incorrectly learned by the network. When blindly presented with the selected signals for re-assessment (23% of the data), the two independent expert-annotators changed the annotation in 25% of the cases. Subsequently, the network was trained on the expert-revised gold standard, which resulted in improved separation between artifacts and nonartifacts as well as an increase in balanced accuracy from 74% to 80% and precision from 59% to 76%. These results show that CNNs are promising to enhance manual annotation of EEG artifacts and can be improved further with better gold-standard data.

30 October 2025 in Brain Communications
Verified
Journal Article
Open access
MDMarina Diachenko
Marina Diachenko
GKGeorgii Krivoshein
Georgii Krivoshein
Arn M.J.M. van den MaagdenbergArn M.J.M. van den Maagdenberg
Arn M.J.M. van den Maagdenberg
HMHuibert D Mansvelder
Huibert D Mansvelder
RKRonald E Van Kesteren
Ronald E Van Kesteren
KLKlaus Linkenkaer-Hansen
Klaus Linkenkaer-Hansen
Early stages of Alzheimer's disease are marked by brain hyperexcitability, evidenced by subclinical epileptiform features suggesting an excitation–inhibition imbalance. Clinically translatable biomarkers for early detection of excitation–inhibition changes at the network level, however, are lacking. We investigated the functional excitation–inhibition ratio, theta–gamma phase–amplitude coupling and epileptiform features in hippocampal and cortical local field potentials recorded weekly in freely behaving male APPswe/PS1dE9 (APP/PS1) mice (n = 10) and wild-type controls (n = 10) between 3 and up to and including 11 months of age. APP/PS1 mice exhibited a shift towards increased excitation, reflected in the elevated functional excitation–inhibition ratio emerging most prominently in the hippocampus at 6 months. Additionally, elevated population spiking activity and age-related impairments in theta–gamma phase–amplitude coupling were observed in the local field potentials of APP/PS1 mice in both the hippocampus and the cortex. Importantly, the functional excitation–inhibition ratio correlated positively with elevated population spiking activity in both brain regions in APP/PS1 mice. Our findings highlight the functional excitation–inhibition ratio as a promising biomarker of hippocampal and cortical network disinhibition and hyperexcitability in APP/PS1 mice, with potential value as an early disease marker in Alzheimer's disease.

7 August 2022 in bioRxiv
Verified
Preprint
Open access
PDF file
RWRobin Weiler
Robin Weiler
MDMarina Diachenko
Marina Diachenko
Erika L Juarez MartinezErika L Juarez Martinez
Erika L Juarez Martinez
AAArthur-Ervin Avramiea
Arthur-Ervin Avramiea
PBPeter Bloem
Peter Bloem
KLKlaus Linkenkaer-Hansen
Klaus Linkenkaer-Hansen
Machine learning techniques such as deep learning have been increasingly used to assist EEG annotation, by automating artifact recognition, sleep staging, and seizure detection. In lack of automation, the annotation process is prone to bias, even for trained annotators. On the other hand, completely automated processes do not offer the users the opportunity to inspect the models’ output and re-evaluate potential false predictions. As a first step towards addressing these challenges, we developed Robin’s Viewer (RV), a Python-based EEG viewer for annotating time-series EEG data. The key feature distinguishing RV from existing EEG viewers is the visualization of output predictions of deep-learning models trained to recognize patterns in EEG data. RV was developed on top of the plotting library Plotly, the app-building framework Dash, and the popular M/EEG analysis toolbox MNE. It is an open-source, platform-independent, interactive web application, which supports common EEG-file formats to facilitate easy integration with other EEG toolboxes. RV includes common features of other EEG viewers, e.g., a view-slider, tools for marking bad channels and transient artifacts, and customizable preprocessing. Altogether, RV is an EEG viewer that combines the predictive power of deep-learning models and the knowledge of scientists and clinicians to optimize EEG annotation. With the training of new deep-learning models, RV could be developed to detect clinical patterns other than artifacts, for example sleep stages and EEG abnormalities.

5 November 2026 in bioRxiv
Verified
Preprint
Open access
PDF file
MBMarit van Den Berg
Marit van Den Berg
TGTina Ganjipour
Tina Ganjipour
MDMarina Diachenko
Marina Diachenko
MCMaria Candellero
Maria Candellero
HMHuibert D. Mansvelder
Huibert D. Mansvelder
KLKlaus Linkenkaer-Hansen
Klaus Linkenkaer-Hansen
EEG source reconstruction requires an anatomical head model, ideally constructed from individual magnetic resonance imaging (MRI) scans. In developmental studies of children and adolescents, however, MRIs are rarely available. Researchers therefore routinely rely on adult templates, such as the FreeSurfer average (fsaverage), despite substantial anatomical changes throughout development. How much this reliance on adult models biases downstream EEG biomarkers remains unclear. Here, we investigated whether age-specific pediatric templates offer a better alternative to fsaverage for EEG-derived biomarkers in children and adolescents, using individual MRI-based reconstruction as the reference. We analyzed eyes-closed resting-state EEGs from 384 Healthy Brain Network participants aged 5.5 to 17.5 years. For each participant, we tested three anatomical models: their individual MRI, an age-specific pediatric template from a series spanning ages 6–17 at one-year intervals, and the adult fsaverage template. We compared four EEG biomarkers in childhood and adolescence: absolute power, relative power, long-range temporal correlations quantified with detrended fluctuation analysis (DFA), and functional excitation-inhibition (fE/I) ratio. Pediatric templates agreed more closely with individual MRI-based biomarkers in both developmental groups and had fewer parcels with significantly different biomarker values than the adult template, except for absolute power in older adolescents. We conclude that when individual MRI is not available, age-specific pediatric templates match individual MRI-based EEG biomarker values more closely than the adult template across all four biomarkers up to approximately 12 years of age and should be the preferred choice at least up to this age.

11 November 2026
Verified
Thesis
Open access
MDMarina Diachenko
Marina Diachenko
Electroencephalography (EEG) is one of the most accessible and cost-effective methods for measuring human brain activity. While EEG is routinely used in clinical practice for epilepsy and sleep disorders, its broader application in neuropsychiatry and neurodegenerative disease remains limited. This likely reflects not only overlapping pathophysiological mechanisms and symptom profiles across disorders, but also methodological challenges in EEG research, including heterogeneous preprocessing procedures, limited reproducibility, and the lack of standardized frameworks for validating biomarkers across cohorts and research sites. This thesis explores computational approaches to facilitate the development and translation of EEG-based biomarkers into clinical applications. It focuses on three interconnected challenges, including model-informed biomarker development and validation, EEG preprocessing and quality control, and the integration of these components within a shared, standardized infrastructure that supports reproducible analyses across datasets, cohorts, and clinical settings. Using the theory of critical brain dynamics, in which the brain operates in or near a critical state, balancing its excitation and inhibition, we develop a method for quantifying the excitation/inhibition ratio from EEG across narrow frequency bands (Chapter 1). We show that this ratio reflects deviations from near-critical network dynamics associated with changes in excitation/inhibition balance. To investigate its biological relevance in disease, we apply the method in a preclinical mouse model of Alzheimer's disease, where it detects shifts toward excitatory dominance and supercritical dynamics in hippocampal and cortical recordings (Chapter 2). Because reliable estimation of EEG-based biomarkers depends on high-quality signal preprocessing, we develop a deep-learning model for artifact detection (Chapter 3) and integrate it into an AI-assisted signal viewer for guided annotation of resting-state EEG (Chapter 4). The results suggest that AI is not yet sufficiently mature to replace expert review but can support and guide manual annotation while facilitating the creating of large, well-annotated datasets for future model development. Finally, we present a proof-of-concept cloud platform designed to support reproducible, collaborative EEG biomarker research and clinical translation. The platform creates a shared environment for end-to-end EEG analysis, integrating EEG visualization, preprocessing, quality control, phenotypic profiling, biomarker extraction, and interactive clinician-oriented reporting. The platform is evaluated in both academic and clinical settings, demonstrating its ability to scale EEG preprocessing and analysis while supporting transparent and reproducible workflows. By supporting data reuse, version-controlled processing pipelines, and model-validated algorithms, such infrastructure may help accelerate the translation of quantitative EEG into clinial practice. Taken together, this thesis contributes methodological and technological advances that support the development, validation, and clinical implementation of EEG-based biomarkers.