- Homepage
- Programme
- Registration
- Practical Guide
- About Neuronus
Joanna Duda-Golawska, Jaroslaw Zygierewicz
Biomedical Physics Division, Faculty of Physics, University of Warsaw, Poland
Abstract: Currently, the most popular approach to the analysis of the event-related components observed in EEG signals in psychological experiments is one based on detailed hypotheses regarding the effects of a priori assumed locations and latencies. We propose a new tool in ERP studies based on multi-channel Matching Pursuit combined with clustering. The algorithm aims to find patterns in EEG signals that are similar across different experimental conditions, but it allows for slight variability in topography and variations in amplitude. The method has yielded the expected results in numerical simulations. Using signals from the emotional categorisation task experiment we showed that the algorithm could be used in two ways. First, the method can be utilised as a specific filter reducing the variability of components, as defined classically, within each experimental condition. Second, equivalent dipoles fitted to items of the activity clusters identified by the algorithm localise in compact brain areas related to the task performed by the subjects across experimental conditions. This suggests that detected activities can be studied as hypothetical hidden components.
Piotr Biegański
University of Warsaw, Faculty of Physics, Warsaw, Poland
Abstract: Actigraphy, which stands for measuring acceleration of various body parts, is considered a robust method of assessing circadian rhythms. At the same time, one of hypotheses regarding diagnosis of DOC links the state of the patient to the rhythmicity of his movements—as they may reflect restoration of circadian rhythmicity. However, simple methods of circadian rhythms detection failed us in the past due to how complex and diverse are the data gathered from patients. Recurrence plots, together with quantitative analysis (RQA) are nonlinear methods of assessing dependencies in the data at different time scales and seem promising as a method more sophisticated than standard algorithms used in the actigraphic field. The analysis is conducted on data gathered in the Alarm Clock Clinic in Warsaw. Our exploratory approach focuses mostly on qualitative differences between various stages of DOC which can be observed on generated recurrence plots. Those differences seem very promising, even though the dataset is small and extremely divergent. Standard measures used in RQA allow to cluster one of the stages (Unresponsive Wakefulness Syndrome)—acting like a necessary condition. Altogether preliminary observations made during this study give a glimpse of hope regarding construction of DOC stages classifier based on actigraphic data.
Funding: This research was supported by the Polish National Science Centre (UMO-2018/31/B/ST7/01888)
Anna Grabowska1, Filip Sondej2, Magdalena Senderecka2
1 Doctoral School in the Social Sciences, Jagiellonian University, Krakow, Poland
2 Institute of Philosophy, Jagiellonian University, Krakow, Poland
Abstract: Alterations in error processing are observable in a range of anxiety-related psychiatric disorders. For instance, enhanced electrophysiological responses to errors (i.e., error-related negativity; ERN) characterise generalised anxiety disorder while schizophrenia shows attenuated ERN. Diagnostic categories in psychiatry are, however, heterogeneous and numerous studies reported contradictory and non-replicating findings. Thus, precise mapping of ERN to psychiatric symptoms remains unclear. To reveal symptoms central for elevated ERN and error-related positivity (Pe), we recorded electroencephalograms from 171 volunteers (120 F; 41 excluded), aged 18–40, while performing speeded Go/No–Go task and collected scores on 7 questionnaires assessing subclinical symptoms. We applied machine learning methods with cross-validation (N = 96) and tested models on a hold-out set (N = 34) to identify generalised brain-behaviour associations. Our findings indicate that enhanced ERN is associated with rumination (R2 = 0.068); overestimation of threat (R2 = 0.055); inhibitory intolerance of uncertainty (R2 = 0.036). Found associations, however, are significantly less robust than usually assumed. Pe is associated with behavioural inhibition (R2 = 0.023); rumination (R2 = 0.063); prospective intolerance of uncertainty (R2 = 0.044). Our results call for the change in results’ validating method to move towards robust findings that reflect stable individual differences and clinically useful biomarkers.
Funding: Sonata Bis grant 2020/38/E/HS6/00490 from the National Science Centre of Poland
Nikodem Hryniewicz1, Marcin Sińczuk1, Rafał Rola2, Ewa-Piątkowska Janko1,3, Danuta Ryglewicz2, Piotr Bogorodzki1,3
1CNS Lab, Nalecz Institute of Biocybernetics and Biomedical Engineering PAS, Warsaw, Poland
2Neurology Department of The Military Institute of Aviation Medicine, Warsaw, Poland
3The Institute of Radioelectronics and Multimedia Technology, Warsaw University of Technology, Warsaw, Poland
Abstract: Epilepsy is one of the most common neurological diseases. Using the discharge onset times derived from the EEG signal, one can compute statistical parametric maps (SPM) from fMRI data. It is necessary to prepare GLM regressors by convolving the driving neuronal activity function with the hemodynamic response function (HRF). This work aimed to prepare a MATLAB application that allows EEG-fMRI analysis with different either HRFs or driving functions. Additionally, we compared the results obtained from one patient's data. We proposed 4 different driving functions and 4 models of the HRF. Standard statistical analysis in SPM12 showed activation cluster in thalamus, the voxel showing the maximum statistical value was therefore chosen as the voxel of interest. The BOLD signal from the voxel was extracted and the beta and mean square error values (MSE) were determined for each HRF model using different driving functions. Prepared toolbox enabled efficient processing and analysis of EEG-fMRI data. The calculated beta values and MSE showed differences in the analysis with the use of various regressors. It has also been shown that changing the parameters of the HRF can improve the fit of estimation to the actual BOLD response, which can improve the result of the analysis.
Sandra Frycz, Marek Binder
Institute of Psychology, Jagiellonian University, Krakow, Poland
Abstract: Prolonged disorders of consciousness (pDoC) include unresponsive wakefulness syndrome, minimally conscious state, and emergence from minimally conscious state. The rate of misdiagnosis of pDoC amounts to 40%, mainly due to the difficulty of behavioural clinical diagnosis in that patient group. Methods based on objective evaluation of brain activity may help in establishing a more accurate diagnosis. We have used resting-state EEG evaluation to investigate its diagnostic applicability. The group of 52 pDoC patients were behaviorally diagnosed using Coma Recovery Scale-Revised (CRS-R). The 10-minute resting-state EEG data were transformed into power spectra and then fitted using FOOOF algorithm.
The resulting set of features describing spectral composition of the signal (spectral peak frequencies and aperiodic component), obtained from 16 channels comprising frontal, parietal and occipital regions, was examined with a supervised machine learning (ML) Gaussian Naive Bayes classifier. The results demonstrated that the classifier can accurately determine the diagnosis with 75% accuracy with CRS-R diagnosis as reference. Classifier performance was probably impeded by varied pDOC aetiology, lesion location and diagnostic limitations of CRS-R. Further development and refinement of pectral features and ML technique may help to develop an objective pDoC diagnostic tool based on a simple EEG measurement.
Funding: OPUS16 project financed by the Polish National Science Centre under the award number 2018/31/B/HS6/03920
