Ing. Eduard Bakštein, Ph.D.

Schizophrenia (SCHZ) notably impacts various human perceptual modalities, including vision. Prior research has identified marked abnormalities in perceptual organization in SCHZ, predominantly attributed to deficits in bottom-up processing. Our study introduces a novel paradigm to differentiate the roles of top-down and bottom-up processes in visual perception in SCHZ. We analysed eye-tracking fixation ground truth maps from 28 SCHZ patients and 25 healthy controls (HC), comparing these with two mathematical models of visual saliency: one bottom-up, based on the physical attributes of images, and the other top-down, incorporating machine learning. While the bottom-up (GBVS) model revealed no significant overall differences between groups (beta = 0.01, p = 0.281, with a marginal increase in SCHZ patients), it did show enhanced performance by SCHZ patients with highly salient images. Conversely, the top-down (EML-Net) model indicated no general group difference (beta = −0.03, p = 0.206, lower in SCHZ patients) but highlighted significantly reduced performance in SCHZ patients for images depicting social interactions (beta = −0.06, p < 0.001). Over time, the disparity between the groups diminished for both models. The previously reported bottom-up bias in SCHZ patients was apparent only during the initial stages of visual exploration and corresponded with progressively shorter fixation durations in this group. Our research proposes an innovative approach to understanding early visual information processing in SCHZ patients, shedding light on the interplay between bottom-up perception and top-down cognition.

In this work, we investigated the accuracy of chronotype (CHT) estimation from actigraphy (ACT) while evaluating the required recording length and stability over time. CHTs have an important role in chronobiological and sleep research. In outpatient studies, CHTs are typically evaluated by questionnaires. Alternatively, ACT provides potential means for measuring CHT characteristics objectively, which opens many applications in chronobiology research. However, studies providing objective, critical evaluation of agreement between questionnaire-based and ACT-based CHTs are lacking. We recorded 3-months of ACT and collected MEQ, and MCTQ results from 122 women. Regression models were applied to evaluate the questionnaire-based CHTs scores using selected ACT features. Changes in predictive strength were evaluated based on ACT recordings of different duration. The ACT was significantly associated with the questionnaire-based CHT, and the best single-feature-based models explained 37% of the variability (R2) for MEQ (p<.001), 47% for mid-sleep time MCTQ-MSFsc (p<.001), and 19% for social jetlag MCTQ-SJLrel (p<.001). Concerning stability in time, the Mid-sleep and Acrophase features showed high levels of stability (test-retest R ~ 0.8), and ACT-based MSFscacti and SJLrelacti showed high temporal variability (test-retest R ~ 0.45). Concerning required recording length, features estimated from recordings with 3-week and longer observation periods had sufficient predictive power on unseen data. Additionally, our data showed that the subjectively reported extremes of the MEQ, MCTQ-MSFsc, and MCTQ-SJLrel are commonly overestimated compared to objective activity peak and middle of sleep differences measured by ACT. Such difference may be associated with CHT time-variation. As ACT is considered accurate in sleep-wake cycle detection, we conclude that ACT-based chronotyping is appropriate for large-scale studies, especially where higher temporal variability in CHT is expected.

OBJECTIVE: Bipolar disorder (BD) is linked to circadian rhythm disruptions resulting in aberrant motor activity patterns. We aimed to explore whether motor activity alone, as assessed by longitudinal actigraphy, can be used to classify accurately BD patients and healthy controls into their respective groups. METHODS: Ninety-day actigraphy records from 25 inter-episode BD patients (i.e. Montgomery-Asberg Depression Rating Scale (MADRS) and Young Mania Rating Scale (YMRS) <15) and 25 sex- and age-matched healthy controls (HC), were used in order to identify latent actigraphic biomarkers capable of discriminating between BD patients and HC. Mean values and time-variations of a set of standard actigraphy features were analysed and further validated using the random forest classifier. RESULTS: Using all actigraphy features, this method correctly assigned 88% (sensitivity=85%, specificity=91%) of BD patients and HC to their respective group. The classification success may be confounded by differences in employment between BD patients and HC. When motor activity features resistant to the employment status were used (the strongest feature being time variation of intradaily variability, Cohen’s d=1.33), 79% of the subjects (sensitivity=76%, specificity=81%) were correctly classified. CONCLUSION: A machine learning actigraphy-based model was capable of distinguishing between inter-episode BD patients and HC solely on the basis of motor activity. The classification remained valid even when features influenced by employment status were omitted. The findings suggest that temporal variability of actigraphic parameters may provide discriminative power for differentiating between BD patients and HC while being less affected by employment status.

Bipolar Disorder (BD) is an illness with high prevalence and a huge social and economic impact. It is recurrent, with a long-term evolution in most cases. Early treatment and continuous monitoring have proven to be very effective in mitigating the causes and consequences of BD. However, no tools are currently available for a massive and semi-automatic BD patient monitoring and control. Taking advantage of recent technological developments in the field of wearables, this paper studies the feasibility of a BD episodes classification analysis while using entropy measures, an approach successfully applied in a myriad of other physiological frameworks. This is a very difficult task, since actigraphy records are highly non-stationary and corrupted with artifacts (no activity). The method devised uses a preprocessing stage to extract epochs of activity, and then applies a quantification measure, Slope Entropy, recently proposed, which outperforms the most common entropy measures used in biomedical time series. The results confirm the feasibility of the approach proposed, since the three states that are involved in BD, depression, mania, and remission, can be significantly distinguished.

Background: Schizophrenia is often characterized by a general disruption of self-processing and self-demarcation. Previous studies have shown that self-monitoring and sense of agency (SoA, i.e., the ability to recognize one's own actions correctly) are altered in schizophrenia patients. However, research findings are inconclusive in regards to how SoA alterations are linked to clinical symptoms and their severity, or cognitive factors. Methods: In a longitudinal study, we examined 161 first-episode schizophrenia patients and 154 controls with a continuous-report SoA task and a control task testing general cognitive/sensorimotor processes. Clinical symptoms were assessed with the Positive and Negative Syndrome Scale (PANSS). Results: In comparison to controls, patients performed worse in terms of recognition of self-produced movements even when controlling for confounding factors. Patients' SoA score correlated with the severity of PANSS-derived “Disorganized” symptoms and with a priori defined symptoms related to self-disturbances. In the follow-up, the changes in the two subscales were significantly associated with the change in SoA performance. Conclusion: We corroborated previous findings of altered SoA already in the early stage of schizophrenia. Decreased ability to recognize self-produced actions was associated with the severity of symptoms in two complementary domains: self-disturbances and disorganization. While the involvement of the former might indicate impairment in self-monitoring, the latter suggests the role of higher cognitive processes such as information updating or cognitive flexibility. The SoA alterations in schizophrenia are associated, at least partially, with the intensity of respective symptoms in a state-dependent manner.

Identification of subcortical structures is an essential step in surgical planning for interventions such as the deep brain stimulation (DBS), in which permanent electrode is implanted in a precisely defined location. For refinement of the target localisation and compensation of brain shift occurring during the surgery, intra-operative electrophysiological recording using microelectrodes is usually undertaken. In this paper, we present a multimodal method that consists of a) subthalamic nucleus (STN) segmentation from magnetic resonance T2 images using 3D active contour fitting and b) a subsequent brain shift compensation step, increasing the accuracy of microelectrode placement localisation by the probabilistic electrophysiology-based fitting. The method is evaluated on a data set of 39 multi-electrode trajectories from 20 patients undergoing DBS surgery for Parkinson’s disease in a leave-one-subject-out scenario. The performance comparison shows increased sensitivity and slightly decreased specificity of STN identification using the individually-segmented 3D contours, compared to electrophysiology-based refinement of a standard 3D atlas. To achieve accurate segmentation from the low-resolution clinical T2 images, a more sophisticated approach, including shape priors and intensity model, needs to be implemented. However, the presented approach is a step towards automatic identification of microelectrode recording sites and possibly also an assistive system for the DBS surgery.

The deep brain stimulation (DBS) is a treatment technique for late-stage Parkinson’s disease (PD), based on chronic electrical stimulation of neural tissue through implanted electrodes. To achieve high level of symptom suppression with low side effects, precise electrode placement is necessary, although difficult due to small size of the target nucleus and various sources of inaccuracy, especially brain shift and electrode bending. To increase accuracy of electrode placement, electrophysiological recording using several parallel microelectrodes (MER) is used intraoperatively in most centers. Location of the target nucleus is identified from manual expert evaluation of characteristic neuronal activity. Existing studies have presented several models to classify individual recordings or trajectories automatically. In this study, we extend this approach by fitting a 3D anatomical atlas to the recorded electrophysiological activity, thus adding topological information. Methods: We developed a probabilistic model of neuronal activity in the vicinity the subthalamic nucleus (STN), based on normalized signal energy. The model is used to find a maximum-likelihood transformation of an anatomical surface-based atlas to the recorded activity. The resulting atlas fit is compared to atlas position estimated from pre-operative MRI scans. Accuracy of STN classification is then evaluated in a leave-one-subject-out scenario using expert MER annotation. Results: In an evaluation on a set of 27 multi-electrode trajectories from 15 PD patients, the proposed method showed higher accuracy in STN-nonSTN classification (88.1%) compared to the reference methods (78.7%) with an even more pronounced advantage in sensitivity (69.0% vs 44.6%). Conclusion: The proposed method allows electrophysiology-based refinement of atlas position of the STN and represents a promising direction in refining accuracy of MER localization in clinical DBS setting, as well as in research of DBS mechanisms.

Deep brain stimulation (DBS) is an internationally accepted form of treatment option for selected patients with Parkinson's disease and dystonia. Intraoperative extracellular microelectrode recordings (MER) are considered as the standard electrophysiological method for the precise positioning of the DBS electrode into the target brain structure. Pre-processing of MERs is a key phase in clinical analysis, with intraoperative microelectrode recordings being prone to several artifact groups (up to 25 %). The aim of this methodological article is to provide a convolutional neural network (CNN) processing pipeline for the detection of artifacts in an MER. We applied continuous wavelet transform (CWT) to generate an over-complete time-frequency representation. We demonstrated that when attempting to find artifacts in an MER, the new CNN + CWT provides a high level of accuracy (ACC = 88.1 %), identifies individual classes of artifacts (ACC = 75.3 %) and also offers artifact time onset detail, which can lead to a reduction in false positives/negatives. In summary, the presented methodology is capable of identifying and removing various artifacts in a comprehensive database of MER and represents a substantial improvement over the existing methodology. We believe that this approach will assist in the proposal of interesting clinical hypotheses and will have neurologically relevant effects.

Clinical motor and non-motor effects of deep brain stimulation (DBS) of the subthalamic nucleus (STN) in Parkinson's disease (PD) seem to depend on the stimulation site within the STN. We analysed the effects of the position of the stimulation electrode within the motor STN on subjective emotional experience, expressed as emotional valence and arousal ratings to pictures representing primary rewards and aversive fearful stimuli in 20 PD patients. Patients’ ratings from both aversive and erotic stimuli matched the mean ratings from a group of 20 control subjects at similar position within the STN. Patients with electrodes located more posteriorly reported both valence and arousal ratings from both the rewarding and aversive pictures as more extreme. Moreover, posterior electrode positions were associated with a higher occurrence of depression at a long-term follow-up. This brain–behavior relationship suggests a complex emotion topography in the motor part of the STN. Both valence and arousal representations overlapped and were uniformly arranged anterior-posteriorly in a gradient-like manner, suggesting a specific spatial organization needed for the coding of the motivational salience of the stimuli. This finding is relevant for our understanding of neuropsychiatric side effects in STN DBS and potentially for optimal electrode placement.

Introduction The study investigates the association between circadian phenotype (CP), its stability (interdaily stability - IS) and physical activity (PA) in a weight loss (WL) programme. Methods Seventy-five women in WL conservative treatment (BMI >= 25 kg/m2) were measured (for about 3 months in between 2016 and 2018) by actigraphy. Results We observed a difference in time of acrophase (p = 0.049), but no difference in IS (p = 0.533) between women who lost and did not lose weight. There was a difference in PA (mesor) between groups of women who lost weight compared to those who gained weight (p = 0.007). There was a relationship between IS and PA parametres mesor: p0.001; and the most active 10 h of a day (M10): p < 0.001 - the more stable were women in their rhythm, the more PA they have. Besides confirming a relationship between PA and WL, we also found a relation between WL and CP based on acrophase. Although no direct relationship was found for the indicators of rhythm stability (IS), they can be considered very important variables because of their close connection to PA - a main factor that contributes to the success of the WL programme. Discussion According to the results of the study, screening of the CP and its stability may be beneficial in the creation of an individualized WL plan.

The deep brain stimulation (DBS) is a symptomatic treatment technique used mainly for movement disorders, consisting of chronic electrical stimulation of subcortical structures. To achieve very precise electrode implantation, which is necessary for a good clinical outcome, many surgical teams use electrophysiological recording around the assumed target, planned in pre-operative MRI images. In our previous work, we developed a probabilistic model to fit a 3D anatomical atlas of the subthalamic nucleus to the recorded microelectrode activity in Parkinson’s disease (PD) patients. In this paper, we extend the model to incorporate characteristic landmarks of the target nucleus, manually annotated in pre-operative MRI data. We validate the approach on a set of 27 exploration five-electrode trajectories from 15 PD patients. The results show that such combined approach may lead to a vast improvement in optimization reliability, while maintaining good fit to the electrophysiology data. The combination of electrophysiology and MRI-based data thus provides a promising approach for compensating brain shift, occuring during the surgery and achieving accurate localization of recording sites in DBS surgery.

Background: Detailed study of the period before schizophrenic relapse when early warning signs (EWS) are present is crucial to effective pre-emptive strategies. Aims: To investigate the temporal properties of EWS self-reported weekly via a telemedicine system. Method: EWS history was obtained for 61 relapses resulting in hospitalization involving 51 patients with schizophrenia. Up to 20 weeks of EWS history per case were evaluated using a non-parametric bootstrap test and generalized mixed-effects model to test the significance and homogeneity of the findings. Results: A statistically significant increase in EWS sum score was detectable 5 weeks before hospitalization. However, analysis of EWS dynamics revealed a gradual, monotonic increase in EWS score across during the 8 weeks before a relapse. Conclusions: The findings-in contrast to earlier studies-suggest that relapse is preceded by a lengthy period during which pathophysiological processes unfold; these changes are reflected in subjective EWS.

BACKGROUND: Extracellular microelectrode recording (MER) is a prominent technique for studies of extracellular single-unit neuronal activity. In order to achieve robust results in more complex analysis pipelines, it is necessary to have high quality input data with a low amount of artifacts. We show that noise (mainly electromagnetic interference and motion artifacts) may affect more than 25% of the recording length in a clinical MER database. NEW METHOD: We present several methods for automatic detection of noise in MER signals, based on (i) unsupervised detection of stationary segments, (ii) large peaks in the power spectral density, and (iii) a classifier based on multiple time- and frequency-domain features. We evaluate the proposed methods on a manually annotated database of 5735 ten-second MER signals from 58 Parkinson's disease patients. COMPARISON WITH EXISTING METHODS: The existing methods for artifact detection in single-channel MER that have been rigorously tested, are based on unsupervised change-point detection. We show on an extensive real MER database that the presented techniques are better suited for the task of artifact identification and achieve much better results. RESULTS: The best-performing classifiers (bagging and decision tree) achieved artifact classification accuracy of up to 89% on an unseen test set and outperformed the unsupervised techniques by 5–10%. This was close to the level of agreement among raters using manual annotation (93.5%). CONCLUSION: We conclude that the proposed methods are suitable for automatic MER denoising and may help in the efficient elimination of undesirable signal artifacts.

We present a probabilistic model for classification of micro- EEG signals, recorded during deep brain stimulation surgery for Parkinson’s disease. The model uses parametric representation of neuronal background activity, estimated using normalized root-mean-square of the signal. Contrary to existing solutions using Bayes classifiers or Hidden Markov Models, our model uses smooth state-transitions represented by sigmoid functions, which ensures flexible model structure in combination with general optimizers for parameter estimation and model fitting. The presented model can easily be extended with additional parameters and constraints and is intended for fitting of a 3D anatomical model to micro-EEG data in further perspective. In an evaluation on 260 trajectories from 61 patients, the model showed classification accuracy 90.0%, which was comparable to existing solutions. The evaluation proved the model successful in target identification and we conclude that its use for more complex tasks in the area of DBS planning and modeling is feasible.

Local field potentials (LFP) were recorded in the subthalamic nucleus of 7 Parkinsonian patients through the implanted electrodes of a deep brain stimulation (DBS) device prior to its internalization. Measured LFP signals were preprocessed by means of splinting, down sampling, filtering, normalization and rectification. Then, feature extraction was conducted through a multi-level decomposition via a wavelet transform. Finally, artificial intelligence techniques were applied to feature selection, clustering of tremor types, and tremor detection.

Appropriate detection of clean signal segments in extracellular microelectrode recordings (MER) is vital for maintaining high signal-to-noise ratio in MER studies. Existing alternatives to manual signal inspection are based on unsupervised change-point detection. We present a method of supervised MER artifact classification, based on power spectral density (PSD) and evaluate its performance on a database of 95 labelled MER signals. The proposed method yielded test-set accuracy of 90%, which was close to the accuracy of annotation (94%). The unsupervised methods achieved accuracy of about 77% on both training and testing data.

Promoting the students motivation is surely an integral part of the university education. It enhances their interest in the study branch and increases the probability of finishing the class not to say the whole university education. In our contribution, we discuss different methods which were used in innovated classes at our university. The methods include working on a common project, using real data, increasing course interdisciplinarity. making software assignments more attractive by implementation into a robotic device and applicability of solved tasks in the commercial sphere. We focus in more detail on results of one of our projects, which aimed at development of a toolbox for sample psychological experiments, synchronized with an EEG device. Thanks to the fact that the whole project is using modular architecture, students can implement individual modules during their semestral, bachelor or diploma theses or can further extend functionality of the toolbox itself.

Photoplethysmography (PPG) is a non-invasive low-cost investigation method, suitable for long-term monitoring of various physiological properties, such as heart rate or blood oxygen saturation. This paper describes method of estimation of respiratory parameters from the PPG signal. In our preliminary study, PPG signal was recorded along with air intake on 8 healthy subjects. Two algorithms, using signal filtering and envelope detection for respiratory parameter estimation are described and evaluated. Relative breath depth was estimated from the recorded PPG signal and compared to actual parameters of the subject's breath, measured by a spirometer device.

This paper explores the development of multi-feature classification techniques used to identify tremor-related characteristics in the Parkinsonian patient. Local field potentials were recorded from the subthalamic nucleus and the globus pallidus internus of eight Parkinsonian patients through the implanted electrodes of a Deep brain stimulation (DBS) device prior to device internalization.A range of signal processing techniques were evaluated with respect to their tremor detection capability and used as inputs in a multi-feature neural network classifier to identify the activity of Parkinsonian tremor. The results of this study show that a trained multi-feature neural network is able, under certain conditions, to achieve excellent detection accuracy on patients unseen during training. Overall the tremor detection accuracy was mixed, although an accuracy of over 86% was achieved in four out of the eight patients.

This paper focuses on wrapper-based feature selection for a 1-nearest neighbor classifier. We consider in particular the case of a small sample size with a few hundred instances, which is common in biomedical applications. We propose a technique for calculating the complete bootstrap for a 1-nearest-neighbor classifier (i.e., averaging over all desired test/train partitions of the data). The complete bootstrap and the complete cross-validation error estimate with lower variance are applied as novel selection criteria and are compared with the standard bootstrap and cross-validation in combination with three optimization techniques - sequential forward selection (SFS), binary particle swarm optimization (BPSO) and simplified social impact theory based optimization (SSITO). The experimental comparison based on ten datasets draws the following conclusions: for all three search methods examined here, the complete criteria are a significantly better choice than standard 2-fold cross-validation, 10-fold cross-validation and bootstrap with 50 trials irrespective of the selected output number of iterations. All the complete criterion-based 1NN wrappers with SFS search performed better than the widely-used FILTER and SIMBA methods. We also demonstrate the benefits and properties of our approaches on an important and novel real-world application of automatic detection of the subthalamic nucleus.

In this paper we describe design of a visualization tool for evaluation of microelectrode recordings, captured during surgery related to placement of electrodes of a deep brain stimulation device. The systém allows easy inspection of variation in the activity of different brain structures and can help searching for strong features for the description of this activity. Hence, the system can be used in the process of designing an automatic classification system. We compared output of the system with the annotation provided by an experienced surgeon, using traditional evaluation techniques, and found that the system output correlated well in most cases. The visualization system was especially successful in determination of the subthalamic nucleus, which is the target structure in the deep brain stimulation surgery. The system also helped us find data problematic to annotateand also revealed several errors in our database.

Although single-cell microelectrode recordings (MER) are used to confirm stereotactic targets during surgery for movement disorders, there is a lack of automatic exploration methods which are designed for guiding surgeon during identifying appropriate target for deep-brain stimulation (DBS) implant. We propose automatic visualization method for MER to determine corresponding deep brain nuclei. The underground hypothesis is that nuclei automatic identification can help to determine the subthalamic nucleus (STN) in Parkinson's disease patients. This approach aims at improving patient outcome by helping neurosurgeons objectively identify target structures.

Deep Brain Stimulation (DBS) is a treatment routinely used to alleviate the symptoms of Parkinson's disease (PD). In this type of treatment, electrical pulses are applied through electrodes implanted into the basal ganglia of the patient. As the symptoms are not permanent in most patients, it is desirable to develop an on-demand stimulator, applying pulses only when onset of the symptoms is detected. This study evaluates a feature set created for the detection of tremor - a cardinal symptom of PD. The designed feature set was based on standard signal features and researched properties of the electrical signals recorded from subthalamic nucleus (STN) within the basal ganglia, which together included temporal, spectral, statistical, autocorrelation and fractal properties. The most characterized tremor related features were selected using statistical testing and backward algorithms then used for classification on unseen patient signals.