Multi-Domain Convolutional Neural Networks for Lower-Limb Motor Imagery Using Dry vs. Wet Electrodes

Motor imagery (MI) brain–computer interfaces (BCIs) have been used for a wide variety of applications due to their intuitive matching between the user’s intentions and the performance of tasks. Applying dry electroencephalography (EEG) electrodes to MI BCI applications can resolve many constraints and achieve practicality. In this study, we propose a multi-domain convolutional neural networks (MD-CNN) model that learns subject-specific and electrode-dependent EEG features using a multi-domain structure to improve the classification accuracy of dry electrode MI BCIs. The proposed MD-CNN model is composed of learning layers for three domain representations (time, spatial, and phase). We first evaluated the proposed MD-CNN model using a public dataset to confirm 78.96% classification accuracy for multi-class classification (chance level accuracy: 30%). After that, 10 healthy subjects participated and performed three classes of MI tasks related to lower-limb movement (gait, sitting down, and resting) over two sessions (dry and wet electrodes). Consequently, the proposed MD-CNN model achieved the highest classification accuracy (dry: 58.44%; wet: 58.66%; chance level accuracy: 43.33%) with a three-class classifier and the lowest difference in accuracy between the two electrode types (0.22%, d = 0.0292) compared with the conventional classifiers (FBCSP, EEGNet, ShallowConvNet, and DeepConvNet) that used only a single domain. We expect that the proposed MD-CNN model could be applied for developing robust MI BCI systems with dry electrodes.

Towards Decoding Selective Attention Through Cochlear Implant Electrodes as Sensors in Subjects with Contralateral Acoustic Hearing

Objectives: Focusing attention on one speaker in a situation with multiple background speakers or noise is referred to as auditory selective attention. Decoding selective attention is an interesting line of research with respect to future brain-guided hearing aids or cochlear implants (CIs) that are designed to adaptively adjust sound processing through cortical feedback loops. This study investigates the feasibility of using the electrodes and backward telemetry of a CI to record electroencephalography (EEG). Approach: The study population included 6 normal-hearing (NH) listeners and 5 CI users with contralateral acoustic hearing. Cortical auditory evoked potentials (CAEP) and selective attention were recorded using a state-of-the-art high-density scalp EEG and, in the case of CI users, also using two CI electrodes as sensors in combination with the backward telemetry system of these devices (iEEG). Main results: The peak amplitudes of the CAEPs recorded with iEEG were lower and the latencies were higher than those recorded with scalp EEG. In the selective attention paradigm with multi-channel scalp EEG the mean decoding accuracy across subjects was 92.0 and 92.5% for NH listeners and CI users, respectively. With single-channel scalp EEG the accuracy decreased to 65.6 and to 75.8% for NH listeners and CI users, respectively, and was above chance level in 9 out of 11 subjects. With the single-channel iEEG, the accuracy for CI users decreased to 70% and was above chance level in 3 out of 5 subjects. Significance: This study shows that single-channel EEG is suitable for auditory selective attention decoding, even though it reduces the decoding quality compared to a multi-channel approach. CI-based iEEG can be used for the purpose of recording CAEPs and decoding selective attention. However, the study also points out the need for further technical development for the CI backward telemetry regarding long-term recordings and the optimal sensor positions.

The problem of universals in cross-cultural studies: Insights from Sámi animal melodies (yoik)

This article offers a critique of the notion of “universals” in cross-cultural studies on music and emotions based on empirical observations and philosophical arguments. The empirical material comes from experiments with songs evoking animals and belonging to the Indigenous Sámi “yoik” tradition. Participants from the Belgian Ardenne untrained to the yoik ( N = 114, age 4–79) listened to recordings and tried to guess which animal was evoked. While their scores were significantly above chance level, additional data about their own environment and relationships to animals illustrate that interpretations in terms of “universals” would obscure the interrelational processes and (productive or unproductive) “misrecognitions” at work during the experiments. By analogy, this illustrates the need for a down-to-earth approach in cross-cultural studies on music that acknowledges the creative role of experimental designs and laboratory conditions in the production of universals. This approach may imply a move away from the nature/culture divide and a renewed attention to experimental subjects in a postcolonial context, with the aim of informing us on the entanglement of human musicality in “relational places” and the productive biases these offer to relate across different environments.

Discovering Digital Tumor Signatures—Using Latent Code Representations to Manipulate and Classify Liver Lesions

Modern generative deep learning (DL) architectures allow for unsupervised learning of latent representations that can be exploited in several downstream tasks. Within the field of oncological medical imaging, we term these latent representations “digital tumor signatures” and hypothesize that they can be used, in analogy to radiomics features, to differentiate between lesions and normal liver tissue. Moreover, we conjecture that they can be used for the generation of synthetic data, specifically for the artificial insertion and removal of liver tumor lesions at user-defined spatial locations in CT images. Our approach utilizes an implicit autoencoder, an unsupervised model architecture that combines an autoencoder and two generative adversarial network (GAN)-like components. The model was trained on liver patches from 25 or 57 inhouse abdominal CT scans, depending on the experiment, demonstrating that only minimal data is required for synthetic image generation. The model was evaluated on a publicly available data set of 131 scans. We show that a PCA embedding of the latent representation captures the structure of the data, providing the foundation for the targeted insertion and removal of tumor lesions. To assess the quality of the synthetic images, we conducted two experiments with five radiologists. For experiment 1, only one rater and the ensemble-rater were marginally above the chance level in distinguishing real from synthetic data. For the second experiment, no rater was above the chance level. To illustrate that the “digital signatures” can also be used to differentiate lesion from normal tissue, we employed several machine learning methods. The best performing method, a LinearSVM, obtained 95% (97%) accuracy, 94% (95%) sensitivity, and 97% (99%) specificity, depending on if all data or only normal appearing patches were used for training of the implicit autoencoder. Overall, we demonstrate that the proposed unsupervised learning paradigm can be utilized for the removal and insertion of liver lesions at user defined spatial locations and that the digital signatures can be used to discriminate between lesions and normal liver tissue in abdominal CT scans.

Computational challenges in explaining communication: How deep the rabbit hole goes

When people are unsure of the intended meaning of a word, they often ask for clarification. One way of doing so—often assumed in models of communication—is to point at a potential target: “Do you mean [points at the rabbit]?” However, what if the target is unavailable? Then the only recourse is language itself, which seems equivalent to pulling oneself up from a swamp by one’s hair. We created two computational models of communication, one able to point and one not. The latter incorporates inference to resolve the meaning of non-pointing signals. Simulations show agents in both models reach perceived understanding equally quickly. While this means agents think they are successfully communicating, non-pointing agents understand each other only at chance level. This shows that state- of-the-art computational explanations have difficulty explaining how people solve the puzzle of underdetermination, and that doing so will require a fundamental leap forward.

Universality and Cultural Specificity in the Emotion Perception from Touch

Can people communicate distinct emotions by touch? Previous studies in Western cultures have indicated that certain emotions could be perceived above the chance level when an encoder conveys emotions by touching a decoder's arm. However, the perception of emotions from touch has not been investigated in Japan, where it is uncommon to use touch as a method of daily communication. Therefore, we conducted an experiment with Japanese participants, which was nearly identical to previous studies with non-Japanese people. Results indicated that anger, love, and gratitude were categorized above chance, and fear, disgust, surprise, envy, and sympathy could also be accurately recognized above chance at a less detailed level such as pleasant or unpleasant, and aroused or non-aroused. These findings suggest universality and differences between Japanese and Westerners regarding the perception of emotions by touch. Note: The original preprint had been uploaded on 17-10-2020 (https://doi.org/10.31234/osf.io/pg8fy). This manuscript is the same as the original preprint.

The ventral striatum harbours patient specific intracranial neural signatures of obsessions and compulsions

Deep brain stimulation is a treatment option for patients with refractory obsessive-compulsive disorder. A new generation of stimulators hold promise for closed loop stimulation, with adaptive stimulation in response to biological signals Here we aimed to discover a suitable biomarker in the ventral striatum in patients with obsessive compulsive disorder using local field potentials. We induced obsessions and compulsions and trained a deep learning model on the recorded time series. Average classification sensitivities were 47% for obsessions and 66% for compulsions for patient specific models at 25% chance level. Sensitivity for obsessions reached over 90% in one patient, whereas performance was near chance level when the model was trained across patients. Optimal sensitivity for obsessions and compulsions was obtained at different recording sites. This study shows that closed loop stimulation is a viable option for OCD, but that intracranial biomarkers for obsessive-compulsive disorder are patient and not disorder specific.

Screen Fast, Screen Faster: A Pilot Study to Screen for Depressive Symptoms Using the Beck Depression Inventory Fast Screen in Parkinson's Disease With Mild Cognitive Impairment

Objective: Depressive symptoms have a high prevalence in patients with Parkinson's disease (PD) and are associated with cognitive dysfunction. Especially in PD with mild cognitive impairment (MCI), a time-efficient and valid instrument for the assessment of depression primarily focusing on psychological symptoms and disregarding confounding somatic symptoms is needed. We performed an examination of the psychometric properties of the Beck Depression Inventory II (BDI-II) and the Beck Depression Inventory Fast Screen (BDI-FS).Methods: The sample consisted of 64 patients [22 females and 42 males, mean age: 67.27 years (SD = 7.32)]. Depressive symptoms were measured in a cohort of PD patients with MCI. For the BDI-II and BDI-FS the psychometric concepts of internal consistency, convergent validity and diagnostic agreement were assessed.Results: Patients gave higher ratings on test items addressing somatic symptoms than those addressing non-somatic ones. The correlation between the absolute total scores of the BDI-II and the BDI-FS was significant (r = 0.91, p < 0.001), which indicated convergent validity. The Cronbach's alpha values indicated adequate internal consistencies for both measures (BDI-II: 0.84; BDI-FS: 0.78). There was a higher than chance level agreement of diagnoses of the two questionnaires, measured by Cohen's kappa (0.58, p < 0.001). The agreements between previous diagnosis of depression and the diagnoses of the BDI-II/BDI-FS were also significantly higher than chance level (BDI-II: 0.34, p = 0.007, BDI-FS: 0.39, p = 0.002). Additional AUC analysis across different cutoffs showed that performance of BDI-FS was better than BDI-II, supporting the observation of an equivalent or better performance of BDI-FS than BDI-II. Importantly, AUC analysis confirmed that a cutoff = 4 for BDI-FS was suitable in the considered sample of patients with PD-MCI.Discussion: In a cohort of PD-MCI, the BDI-FS demonstrates adequate psychometric properties in comparison to the BDI-II and can be used as a screening measure for assessing depression in cognitively impaired PD patients, focusing solely on psychological symptoms. Still, further research is needed to validate this instrument.