Осадчий Алексей Евгеньевич
Институт когнитивных нейронаук
Профессиональные интересы
Должности
- Директор центра — Институт когнитивных нейронаук, Центр биоэлектрических интерфейсов
- Профессор — Факультет компьютерных наук, Департамент анализа данных и искусственного интеллекта
Био
- · Начал работать в НИУ ВШЭ в 2013 году.
- · Научно-педагогический стаж: 21 год.
Образование
- 2023 · Доктор наук: Национальный исследовательский университет "Высшая школа экономики"
- 2003 · PhD: Университет Южной Калифорнии, специальность 01.00.00 «Физико-математические науки» и 03.03.06 «Нейробиология», тема диссертации: Автоматическое неинвазивное обнаружение и анализ взаимодействия эпилептогенных зон на основании МЭГ и ЭЭГ измерений
- 1997 · Специалитет: Московский государственный технический университет им. Н.Э. Баумана, специальность «Автономные информационные и управляющие системы», квалификация «Инженер-радиотехник»
Опыт работы
- · Директор Центра биоэлектрических интерфейсов НИУ ВШЭ Ведущий научный сотрудник Центра познания и принятия решений НИУ ВШЭ Профессор кафедры анализа данных и искусственного интеллекта НИУ ВШЭ Старший научный сотрудник Центра познания и принятия решений НИУ ВШЭ
- · 2007-2013: гг. Доцент див. для высшей нервной деятельности, Биолого-почвенный факультет Санкт-Петербургского государственного университета
- · 2005-2015: гг. Независимый консультант по визуализации сигналов источника, Сан-Диего, Калифорния
- · 2003-2005: гг. Старший ученый. Source Signal Imaging Inc., Сан-Диего, Калифорния
- · 1999 — 2003: 09/ Научный сотрудник. Лаборатория нейровизуализации при USC, адв. Р. Лихи
- · 2002 — 2003: 09/ Научный сотрудник. Отделение MEG в Huntington Medical Res. Inst
- · 2001: 06/ 01/
- · 2002: Консультационный отдел химии, USC
- · 2000: 05/ 08/
- · 2000: Research Intern. Исследовательские лаборатории Хьюза, Малибу, Калифорния
- · 1998: 09/ 08/
- · 1999: Научный сотрудник. Центр интегрированных медиа-систем (IMSC, USC)
- · 1995: 03/ 07/
- · 1998: Научный сотрудник. Исследовательский центр «Модуль», Москва
- · 01.1993 — 03.1995: Научный сотрудник. Кафедра автономных систем управления МГТУ им. Н. Э. Баумана
Награды и поощрения
- · Медаль "Признание - 10 лет успешной работы" НИУ ВШЭ (июль 2025)
- · Благодарность Высшей школы экономики (сентябрь 2021)
- · Благодарность Факультета компьютерных наук НИУ ВШЭ (август 2018)
- · Надбавка за защиту докторской диссертации (2023–2026)
- · Надбавка за публикацию в международном рецензируемом научном издании (2019–2021, 2018–2020, 2017–2018)
- · Надбавка за регулярные публикации в международных рецензируемых научных изданиях (2024–2029, 2023–2028, 2021–2026)
- · Надбавка за статью в зарубежном рецензируемом журнале (2015–2017)
Гранты и проекты
- 1017 · Система регистрации и декодирования биоэлектрической активности мозга человека, госконтракт, Министерство Образования и Науки РФ, совместно с ННГУ. 2014-1017 г.
- — · Новая неинвазивная экспериментально-математическая парадигма предоперационного магнитоэнцефалографического картирования речевой коры головного мозга, Грант РФФИ 14-02-00917
- — · РФФИ 16-04-01863 Эндогенное повышение эффективности работы интерфейсов мозг-компьютер
Конференции (10)
Показать все
- · 2024: 27th European Conference on Artificial Intelligence (ECAI 2024) (Сантьяго-де-Компостела). Доклад: EEG-Based fMRI Digital Twin: Towards a Cheap and Ecological Approach to Measure Subcortical Brain Activity
- · 2023: The Fifth International Conference «Neurotechnologies and Neurointerfaces» (CNN 2023) (Kaliningrad). Доклад: Interpretable neural networks in neurointerfaces and neuroimaging methods
- · 2023: Volga Neuroscience Meeting 2023 (Нижний Новгород). Доклад: Diagnostic approaches for precision medicine in epilepsy
- · 2016: IEEE International Symposium «Video and Audio Signal Processing in the Context of Neurotechnologies» (Санкт-Петербург). Доклад: MEG correlates of internalization of social influence
- · 2016: Biomag 2016 (Сеул). Доклад: Power and shift invariant imaging of coherent sources from MEG data (PSIICoS)
- · 2015: V Международная конференция по биотехнологиям и фармацевтике ФизтехБио — 2015 (Москва). Доклад: MEG and EEG based neuroimaging of transient networks
- · 2015: Методические проблемы оценки функциональной синхронизации зон коры мозга на основании ЭЭГ-/МЭГ данных» (Москва). Доклад: МЭГ как результат активности и взаимодействия динамических сетей: метод порождающей модели
- · 2014: International conference on biomagnetism, Biomag 2014 (Галифакс). Доклад: Interaction Space RAP-MUSIC for estimation of transient networks from MEG data
- · 2014: 9th FENS Forum of Neuroscience (Милан). Доклад: MPFC activity varies with differences in social conformity: MEG study
- · 2014: Научная сессия "Проблемы мозга" Российской Академии Наук (Москва). Доклад: Эффективное нейробиоуправление на основе пространственно-временных динамических моделей
Идентификаторы исследователя
- ORCID:
0000-0001-8827-9429 - ResearcherID:
M-9067-2013 - SPIN РИНЦ:
5631-4743 - Google Scholar: https://scholar.google.ru/citations?user=uVunrzkAAAAJ&hl=en
- Scopus AuthorID:
6603011121
Публикации (98)
fMRI from EEG is only Deep Learning away: the use of interpretable DL to unravel EEG-fMRI relationships
2022 · ARTICLE · en
The access to activity of subcortical structures offers unique opportunity for building intention dependent brain-computer interfaces, renders abundant options for exploring a broad range of cognitive phenomena in the realm of affective neuroscience including complex decision making processes and the eternal free-will dilemma and facilitates diagnostics of a range of neurological deceases. So far this was possible only using bulky, expensive and immobile fMRI equipment. Here we present an interpretable domain grounded solution to recover the activity of several subcortical regions from the multichannel EEG data and demonstrate up to 60 % correlation between the actual subcortical blood oxygenation level dependent (sBOLD) signal and its EEG-derived twin. Then, using the novel and theoretically justified weight interpretation methodology we recover individual spatial and time-frequency patterns of scalp EEG predictive of the hemodynamic signal in the subcortical nuclei. The described results not only pave the road towards wearable subcortical activity scanners but also showcase an automatic knowledge discovery process facilitated by deep learning technology in combination with an interpretable domain constrained architecture and the appropriate downstream task.
ECoG Based Classification of Hand Movement Direction in The Stylus Center-Out Paradigm
2022 · CHAPTER · en
Brain-computer interfaces (BCIs) based on electrocorticographic (ECoG) activity has become relatively popular due to sensitivity of ECoG to specific details of actual and imagined actions. It has been shown that significant differences in the brain activity while performing hand movements in different directions can be found in relation to cosine tuning theory. This research seeks to investigate brain activity properties during performance hand motor tasks and distinguish its special patterns for the future ECoG BCIs.
Prestimulus beta rhythm influence reaction time during real-time brain-dependent stimuli presentation
2022 · CHAPTER · en
In the present study, a fast and adaptive technique for the presentation of stimuli based on ongoing brain rhythm is described. Sensorimotor cortical mu rhythm (divided by two components: alpha (mu) and beta) was used as target for assessment of prestimulus rhythm’s power influence on the consequent reaction time. The final sample consisted of 15 participants who was instructed to response immediately after change of stimuli color. As a result of the method application, a longer reaction time in the case of highly synchronized beta oscillations compared to desynchronization was achieved in the simple reaction time task. It indicates, firstly, a crucial role of baseline, prestimulus beta in motor action initiation and, secondly, the possibility to change reaction using adaptive processing and timing of presentation in real-time.
MEG signatures of remote effects of agreement and disagreement with the majority
2021 · ARTICLE · en
People often change their beliefs by succumbing to an opinion of the majority. Such changes are often referred to as majority influence or conformity. While some previous studies have focused on the reinforcement learning mechanisms of conformity or on its internalization, others have reported evidence of changes in sensory processing evoked by majority opinion. In this study, we used magnetoencephalographic (MEG) source imaging to further investigate the remote effects of agreement and disagreement with the majority. During the first session, participants rated the trustworthiness of faces and subsequently learned how the majority of their peers had previously rated each face. To identify the neural correlates of the post-effect of agreeing or disagreeing with the group, we recorded MEG activity while participants rated faces during the next session. We found MEG traces of past disagreement or agreement with the peer group at the parietal cortices as early as approximately 230 ms after the face onset. The neural activity of the superior parietal lobule, intraparietal sulcus, and precuneus was significantly stronger if the participant’s rating had previously differed from the ratings of his or her peers. The early MEG correlates of disagreement with the majority were followed by activity in the orbitofrontal cortex starting at about 320 ms after the face onset. Altogether, the results reveal the temporal dynamics of the neural mechanism of remote effects of disagreement with the peer group: early signatures of modified face processing were followed by later markers of long-term social influence on the valuation process at the ventromedial prefrontal cortex
Modified covariance beamformer for solving MEG inverse problem in the environment with correlated sources
2021 · ARTICLE · en
Magnetoencephalography (MEG) is a neuroimaging method ideally suited for non-invasive studies of brain dynamics. MEG’s spatial resolution critically depends on the approach used to solve the ill-posed inverse problem in order to transform sensor signals into cortical activation maps. Over recent years non-globally optimized solutions based on the use of adaptive beamformers (BF) gained popularity. When operating in the environment with a small number of uncorrelated sources the BFs perform optimally and yield high spatial resolution. However, the BFs are known to fail when dealing with correlated sources acting like poorly tuned spatial filters with low signal-to-noise ratio (SNR) of the output timeseries and often meaningless cortical maps of power distribution. This fact poses a serious limitation on the broader use of this promising technique especially since fundamental mechanisms of brain functioning, its inherent symmetry and task-based experimental paradigms result into a great deal of correlation in the activity of cortical sources. To cope with this problem, we developed a novel data covariance modification approach that allows for building beamformers that maintain high spatial resolution when operating in the environments with correlated sources. At the core of our method is a projection operation applied to the vectorized sensor-space covariance matrix. This projection does not remove the activity of the correlated sources from the sensor-space covariance matrix but rather selectively handles their contributions to the covariance matrix and creates a sufficiently accurate approximation of an ideal data covariance that could hypothetically be observed should these sources be uncorrelated. Since the projection operation is reciprocal to the PSIICOS method developed by us earlier (Ossadtchi et al., 2018) we refer to the family of algorithms presented here as ReciPSIICOS. We assess the performance of the novel approach using realistically simulated MEG data and show its superior performance in comparison to the classical BF approaches and well established MNE as a method immune to source synchrony by design. We have also applied our approach to the MEG datasets from the two experiments involving two different auditory tasks. The analysis of experimental MEG datasets showed that beamformers from ReciPSIICOS family, but not the classical BF, discovered the expected bilateral focal sources in the primary auditory cortex and detected motor cortex activity associated with the audio-motor task. In most cases MNE managed well but as expected produced more spatially diffuse source distributions. Notably, ReciPSIICOS beamformers yielded cortical activity estimates with SNR several times higher than that obtained with the classical BF, which may indirectly indicate the severeness of the signal cancellation problem when applying classical beamformers to MEG signals generated by synchronous sources.
Linear Systems Theoretic Approach to Interpretation of Spatial and Temporal Weights in Compact CNNs: Monte-Carlo Study
2021 · CHAPTER · en
Interpretation of the neural networks architectures for decoding the signals of the brain usually reduced to the analysis of spatial and temporal weights. We propose a theoretically justified method of their interpretation within the simple architecture based on a priori knowledge of the subject area. This architecture is comparable in decoding quality to the winner of the BCI IV competition and allows for automatic engineering of physiologically meaningful features. To demonstrate the operation of the algorithm, we performed Monte Carlo simulations and received a significant improvement in the restoration of patterns for different noise levels and also investigated the relation between the decoding quality and patterns reconstruction fidelity.
Decoding Neural Signals with a Compact and Interpretable Convolutional Neural Network
2021 · CHAPTER · en
In this work, we motivate and present a novel compact CNN. For the architectures that combine the adaptation in both space and time, we describen a theoretically justified approach to interpreting the temporal and spatial weights. We apply the proposed architecture to Berlin BCI IV competition and our own datasets to decode electrocorticogram into finger kinematics. Without feature engineering our architecture delivers similar or better decoding accuracy as compared to the BCI competition winner. After training the network, we interpret the solution (spatial and temporal convolution weights) and extract physiologically meaningful patterns.
Decoding and interpreting cortical signals with a compact convolutional neural network
2021 · ARTICLE · en
Objective: Brain-computer interfaces (BCIs) decode information from neural activity and send it to external devices. The use of Deep Learning approaches for decoding allows for automatic feature engineering within the specific decoding task. Physiologically plausible interpretation of the network parameters ensures the robustness of the learned decision rules and opens the exciting opportunity for automatic knowledge discovery. Approach: We describe a compact convolutional network-based architecture for adaptive decoding of electrocorticographic (ECoG) data into finger kinematics. We also propose a novel theoretically justified approach to interpreting the spatial and temporal weights in the architectures that combine adaptation in both space and time. The obtained spatial and frequency patterns characterizing the neuronal populations pivotal to the specific decoding task can then be interpreted by fitting appropriate spatial and dynamical models. Main results: We first tested our solution using realistic Monte-Carlo simulations. Then, when applied to the ECoG data from Berlin BCI competition IV dataset, our architecture performed comparably to the competition winners without requiring explicit feature engineering. Using the proposed approach to the network weights interpretation we could unravel the spatial and the spectral patterns of the neuronal processes underlying the successful decoding of finger kinematics from an ECoG dataset. Finally we have also applied the entire pipeline to the analysis of a 32-channel EEG motor-imagery dataset and observed physiologically plausible patterns specific to the task. Significance: We described a compact and interpretable CNN architecture derived from the basic principles and encompassing the knowledge in the field of neural electrophysiology. For the first time in the context of such multibranch architectures with factorized spatial and temporal processing we presented theoretically justified weights interpretation rules. We verified our recipes using simulations and real data and demonstrated that the proposed solution offers a good decoder and a tool for investigating motor control neural mechanisms.
Data-Driven Parametric Statistical Testing of Functional Connectivity Between Brain Sources Characterized by Activity with Close-to-Zero Phase Lags
2021 · CHAPTER · en
One of the main methodological problems in evaluation of functional connectivity is the spatial leakage (SL) effect which occurs due to volume conduction and leads to false positives in coherence or phase-locking estimates. Several solutions have been already suggested, including the use of the imaginary part of coherency or cross-spectrum. Because these standard metrics are insensitive to zero-phase interactions, they prevent detection of false coupling, resulting from SL, but may underestimate true physiological interactions, characterized by close-to-zero phase lags. Due to the broad neurophysiological evidence, such interactions should not be excluded from consideration. The recently proposed method, referred as Phase Shift Invariant Imaging of Coherent Sources (PSIICOS), became the first implementation of the algorithm which reliably detects interactions for all the range of phase-lags by suppressing the power of SL subspace components of cross-spectrum. However, connectivity values obtained via PSIICOS are non-normalized by construction and depend on source power, so that uncoupled sources with high power profiles may become false positives. This limitation motivated us to develop a statistical test based on randomization of original time series or cross-spectrum in such a way that power distribution in source space is preserved, but phase interactions are eliminated. The generation of covariance matrices from Wishart distribution appeared to be the most reliable method, when applied to data from simulations. Thus, together with the proposed statistical test PSIICOS can be used as an effective instrument applicable to real EEG- or MEG-data in fundamental research or for clinical purposes.
Compact and interpretable architecture for speech decoding from stereotactic EEG
2021 · CHAPTER · en
Background: Brain-computer interfaces (BCIs) decode neural activity and extract from it information that can be meaningfully interpreted. One of the most intriguing opportunities is to employ BCIs for decoding speech, a uniquely human trait, which opens up plentiful applications from rehabilitation of patients to a direct and seamless communication between human species. To decipher neuronal code complex deep neural networks furnish only limited success. In such solutions an iffy performance gain is achieved with uniterpretable decision rules characterised by thousands of parameters to be identified from a limited amount of training data. Our recent experience shows that when applied to neural activity data compact neural networks with trainable and physiologically meaningful feature extraction layers [1] deliver comparable performance, ensure robustness of the learned decision rules and offer the exciting opportunity of automatic knowledge discovery. Methods: We collected approximately one hour of data (from two sessions) where we recorded stereotactic EEG (sEEG) activity during overt speech (6 different randomly shuffled phrases and rest). We have also recorded synchronized audio speech signal. The sEEG recording was carried out in an epilepsy patient implanted for medical reasons with an sEEG electrode passing through Broca area with 6 contacts spaced at 5 mm. We then used a compact convolutional network-based architecture to recover speech mel-cepstrum coefficients followed by a 2D convolutional network to classify individual words. We then interpreted the former network weights using the theoretically justified approach devised by us earlier [1]. Results: We achieved on average 44% accuracy in classifying 26+ 1 words (3.7% chance level) using only 6 channels of data recorded with a single minimally invasive sEEG electrode. We compared the performance of our compact convolutional network to that of the DenseNet-like architecture that has recently been featured in neural speech decoding literature and did not find statistically significant performance differences. Moreover, our architecture appeared to be able to learn faster and resulted in a stable, interpretable and physiologically meaningful decision rule successfully operating over a contiguous data segment no-overlapping with the training data interval. Spatial characteristics of neuronal population pivotal to the task corroborate the results of active speech mapping procedure and frequency domain patterns show primary involvement of the high frequency activity. Conclusions: Most of the speech decoding solutions availabel to date either use potentially harmful intracortical electrodes or rely on the data recorded with impractically massive multielectrode grids covering large cortical area. Here we for the first time achieved practically usable decoding accuracy for the vocabulary of 26 words + 1 silence class backed by only 6 channels of cortical activity sampled with a single sEEG shaft. The decoding was implemented using a compact and interpretable architecture which ensures robustness of the solution and requires small amount of training data. The proposed approach is the first step towards minimally invasive implantable BCI solution for restoring speech function.
Курсы (1)
-
Mathematical Aspects of EEG and MEG Based Neuroimaging · 5 раза
2025/2026, 2024/2025, 2023/2024, 2022/2023, 2021/2022 · Магистратура / Маго-лего · Анг