1,721,013 research outputs found
Explaining Network Decision Provides Insights on the Causal Interaction Between Brain Regions in a Motor Imagery Task
No full textNeural decoding widely exploits machine learning for classifying electroencephalographic (EEG) signals for brain-computer interface applications. Recent advancements in neural decoding regards the use of brain functional connectivity estimates as input features and the adoption of convolutional neural networks (CNNs) to realize decoders. Moreover, explainable artificial intelligence (XAI) approaches based on CNNs are growing interest in the neuroscience community, for validating the knowledge learned by networks and for using the decoder not only to classify the EEG but also to analyze it in a data-driven way, without a priori assumptions. However, the adoption of connectivity estimates for neural decoding is still in its infancy, as adopts non-directed connectivity measures, limits the analysis of few interactions/frequency ranges, and exploits classic machine learning approaches without exploring CNNs. Moreover, XAI approaches have never been applied to analyze EEG-based functional connectivity. To overcome these limitations, we design and apply a CNN for processing directed connectivity measures estimated via spectral Granger causality. The CNN automatically learns features in the frequency and spatial domains, and it is coupled with an explanation technique (DeepLIFT) for highlighting the most relevant connectivity inflow and outflow associated to each decoded brain state. Our approach is applied to motor imagery decoding, and achieves state-of-the-art performance compared to existing networks. DeepLIFT relevance representations match the directional interactions known occurring when imagining movements, validating the features related to the brain network, as learned by the CNN
Deep learning-based EEG analysis: investigating P3 ERP components
Full text linkThe neural processing of incoming stimuli can be analysed from the electroencephalogram (EEG) through event-related potentials (ERPs). The P3 component is largely investigated as it represents an important psychophysiological marker of psychiatric disorders. This is composed by several subcomponents, such as P3a and P3b, reflecting distinct but interrelated sensory and cognitive processes of incoming stimuli. Due to the low EEG signal-to-noise-ratio, ERPs emerge only after an averaging procedure across trials and subjects. Thus, this canonical ERP analysis lacks in the ability to highlight EEG neural signatures at the level of single-subject and single-trial. In this study, a deep learning-based workflow is investigated to enhance EEG neural signatures related to P3 subcomponents already at single-subject and at single-trial level. This was based on the combination of a convolutional neural network (CNN) with an explanation technique (ET). The CNN was trained using two different strategies to produce saliency representations enhancing signatures shared across subjects or more specific for each subject and trial. Cross-subject saliency representations matched the signatures already emerging from ERPs, i.e., P3a and P3b-related activity within 350–400 ms (frontal sites) and 400–650 ms (parietal sites) post-stimulus, validating the CNN+ET respect to canonical ERP analysis. Single-subject and single-trial saliency representations enhanced P3 signatures already at the single-trial scale, while EEG-derived representations at single-subject and single-trial level provided no or only mildly evident signatures. Empowering the analysis of P3 modulations at single-subject and at single-trial level, CNN+ET could be useful to provide insights about neural processes linking sensory stimulation, cognition and behaviour
A Lightweight Multi-Scale Convolutional Neural Network for P300 Decoding: Analysis of Training Strategies and Uncovering of Network Decision
Full text linkConvolutional neural networks (CNNs), which automatically learn features from raw data to approximate functions, are being increasingly applied to the end-to-end analysis of electroencephalographic (EEG) signals, especially for decoding brain states in brain-computer interfaces (BCIs). Nevertheless, CNNs introduce a large number of trainable parameters, may require long training times, and lack in interpretability of learned features. The aim of this study is to propose a CNN design for P300 decoding with emphasis on its lightweight design while guaranteeing high performance, on the effects of different training strategies, and on the use of post-hoc techniques to explain network decisions. The proposed design, named MS-EEGNet, learned temporal features in two different timescales (i.e., multi-scale, MS) in an efficient and optimized (in terms of trainable parameters) way, and was validated on three P300 datasets. The CNN was trained using different strategies (within-participant and within-session, within-participant and cross-session, leave-one-subject-out, transfer learning) and was compared with several state-of-the-art (SOA) algorithms. Furthermore, variants of the baseline MS-EEGNet were analyzed to evaluate the impact of different hyper-parameters on performance. Lastly, saliency maps were used to derive representations of the relevant spatio-temporal features that drove CNN decisions. MS-EEGNet was the lightest CNN compared with the tested SOA CNNs, despite its multiple timescales, and significantly outperformed the SOA algorithms. Post-hoc hyper-parameter analysis confirmed the benefits of the innovative aspects of MS-EEGNet. Furthermore, MS-EEGNet did benefit from transfer learning, especially using a low number of training examples, suggesting that the proposed approach could be used in BCIs to accurately decode the P300 event while reducing calibration times. Representations derived from the saliency maps matched the P300 spatio-temporal distribution, further validating the proposed decoding approach. This study, by specifically addressing the aspects of lightweight design, transfer learning, and interpretability, can contribute to advance the development of deep learning algorithms for P300-based BCIs
Interpretable and lightweight convolutional neural network for EEG decoding: Application to movement execution and imagination
Full text linkConvolutional neural networks (CNNs) are emerging as powerful tools for EEG decoding: these techniques, by automatically learning relevant features for class discrimination, improve EEG decoding performances without relying on handcrafted features. Nevertheless, the learned features are difficult to interpret and most of the existing CNNs introduce many trainable parameters. Here, we propose a lightweight and interpretable shallow CNN (Sinc-ShallowNet), by stacking a temporal sinc-convolutional layer (designed to learn band-pass filters, each having only the two cut-off frequencies as trainable parameters), a spatial depthwise convolutional layer (reducing channel connectivity and learning spatial filters tied to each band-pass filter), and a fully-connected layer finalizing the classification. This convolutional module limits the number of trainable parameters and allows direct interpretation of the learned spectral–spatial features via simple kernel visualizations. Furthermore, we designed a post-hoc gradient-based technique to enhance interpretation by identifying the more relevant and more class-specific features. Sinc-ShallowNet was evaluated on benchmark motor-execution and motor-imagery datasets and against different design choices and training strategies. Results show that (i) Sinc-ShallowNet outperformed a traditional machine learning algorithm and other CNNs for EEG decoding; (ii) The learned spectral–spatial features matched well-known EEG motor-related activity; (iii) The proposed architecture performed better with a larger number of temporal kernels still maintaining a good compromise between accuracy and parsimony, and with a trialwise rather than a cropped training strategy. In perspective, the proposed approach, with its interpretative capacity, can be exploited to investigate cognitive/motor aspects whose EEG correlates are yet scarcely known, potentially characterizing their relevant features
EEG Motor Execution Decoding via Interpretable Sinc-Convolutional Neural Networks
No full textThe decoding of brain signals is a fundamental component of a brain-computer interface. Despite the success of deep convolutional neural networks (CNNs) in other fields, only recently these techniques have been applied to electroencephalographic (EEG) signals. One drawback of CNNs is the lack of interpretation of the learned features. In this study we introduce for the first time a sinc-convolutional layer into a CNN for EEG motor execution decoding, allowing a straightforward interpretation of the learned kernels. Furthermore, we apply a gradient-based analysis to assess the most relevant EEG bands for each movement and the most relevant EEG electrodes exploited in these bands. In addition to a slight accuracy improvement from 91.9 to 92.4%, our results suggest that the band is the most relevant EEG band, with gradient-based scalp distributions well localized at specific subsets of electrodes
A Dual-Branched Convolutional Neural Network for P300 detection and feature interpretation
No full textConvolutional Neural Network for a P300 Brain-Computer Interface to Improve Social Attention in Autistic Spectrum Disorder
No full textA Brain-Computer Interface (BCI) relies on machine learning algorithms to decode the brain signals. An accurate detection of P300 response in electroencephalography (EEG) data can be used to design P300-based BCIs to improve social attention in Autistic Spectrum Disorder (ASD). Recently, there was a growing interest in the application of Convolutional Neural Networks (CNNs) to decode P300 in an end-to-end fashion. However, the complexity of these models needs to be carefully taken into account. In this study, a lightweight CNN previously validated for P300 detection (EEGNet) was used to decode which object ASD participants were paying attention to in a virtual environment. Two learning strategies were deepened: within-session and cross-session trainings. Cross-session training resulted in a higher target object accuracy scoring 92.27% on average across sessions and subjects, and in a lower decoding variability across sessions
Augmented reality for restoration/reconstruction of artefacts with artistic or historical value
Full text linkThe artistic or historical value of a structure, such as a monument, a mosaic, a painting or, generally speaking, an artefact, arises from the novelty and the development it represents in a certain field and in a certain time of the human activity. The more faithfully the structure preserves its original status, the greater its artistic and historical value is. For this reason it is fundamental to preserve its original condition, maintaining it as genuine as possible over the time.
Nevertheless the preservation of a structure cannot be always possible (for traumatic events as wars can occur), or has not always been realized, simply for negligence, incompetence, or even guilty unwillingness. So, unfortunately, nowadays the status of a not irrelevant number of such structures can range from bad to even catastrophic.
In such a frame the current technology furnishes a fundamental help for reconstruction/restoration purposes, so to bring back a structure to its original historical value and condition. Among the modern facilities, new possibilities arise from the Augmented Reality (AR) tools, which combine the virtual reality (VR) settings with real physical materials and instruments.
The idea is to realize a virtual reconstruction/restoration before materially acting on the structure itself. In this way main advantages are obtained among which: the manpower and machine power are utilized only in the last phase of the reconstruction; potential damages/abrasions of some parts of the structure are avoided during the cataloguing phase; it is possible to precisely define the forms and dimensions of the eventually missing pieces, etc.
Actually the virtual reconstruction/restoration can be even improved taking advantages of the AR, which furnish lots of added informative parameters, which can be even fundamental under specific circumstances. So we want here detail the AR application to restore and reconstruct the structures with artistic and/or historical valu
Consumare latte in maniera sostenibile: Il ruolo congiunto di variabili socio-psicologiche e della valutazione degli attributi intrinseci ed estrinseci del prodotto
No full textKinematic characteristics of adults walking in shallow water using wearable inertial magnetic sensors
No full text- …
