Contribution of low-level image statistics to EEG decoding of semantic content in multivariate and univariate models with feature optimization.

Spatio-temporal patterns of evoked brain activity contain information that can be used to decode and categorize the semantic content of visual stimuli. However, this procedure can be biased by low-level image features independently of the semantic content present in the stimuli, prompting the need to understand the robustness of different models regarding these confounding factors. In this study, we trained machine learning models to distinguish between concepts included in the publicly available THINGS-EEG dataset using electroencephalography (EEG) data acquired during a rapid serial visual presentation paradigm. We investigated the contribution of low-level image features to decoding accuracy in a multivariate model, utilizing broadband data from all EEG channels. Additionally, we explored a univariate model obtained through data-driven feature selection applied to the spatial and frequency domains. While the univariate models exhibited better decoding accuracy, their predictions were less robust to the confounding effect of low-level image statistics. Notably, some of the models maintained their accuracy even after random replacement of the training dataset with semantically unrelated samples that presented similar low-level content. In conclusion, our findings suggest that model optimization impacts sensitivity to confounding factors, regardless of the resulting classification performance. Therefore, the choice of EEG features for semantic decoding should ideally be informed by criteria beyond classifier performance, such as the neurobiological mechanisms under study.

The Long and Winding Road to Real-Life Experiments: Remote Assessment of Executive Functions with Computerized Games-Results from 8 Years of Naturalistic Interventions.

Mate Marote is an open-access cognitive training software aimed at children between 4 and 8 years old. It consists of a set of computerized games specifically tailored to train and evaluate Executive Functions (EF), a class of processes critical for purposeful, goal-directed behavior, including working memory, planning, flexibility, and inhibitory control. Since 2008, several studies were performed with this software at children's own schools in interventions supervised in-person by cognitive scientists. After 2015, we incorporated naturalistic, yet controlled, interventions with children's own teachers' help. The platform includes a battery of standardized tests, disguised as games, to assess children's EF. The main question that emerges is whether the results, obtained with these traditional tasks but conducted without the presence of researchers, are comparable to those widely reported in the literature, that were obtained in more supervised settings. In this study, we were able to replicate the expected difficulty and age effects in at least one of the analyzed dependent variables of each employed test. We also report important discrepancies between the expected and the observed response time patterns, specifically for time-constrained tasks. We hereby discuss the benefits and setbacks of a new possible strategy for this type of assessment in naturalistic settings. We conclude that this battery of established EF tasks adapted for its remote usage is appropriate to measure the expected mental processes in naturalistic settings, enriching opportunities to upscale cognitive training interventions at schools. These types of tools can constitute a concerted strategy to bring together educational neuroscience research and real-life practice.

Assessing robustness and generalization of a deep neural network for brain MS lesion segmentation on real-world data.

OBJECTIVES: Evaluate the performance of a deep learning (DL)-based model for multiple sclerosis (MS) lesion segmentation and compare it to other DL and non-DL algorithms. METHODS: This ambispective, multicenter study assessed the performance of a DL-based model for MS lesion segmentation and compared it to alternative DL- and non-DL-based methods. Models were tested on internal (n = 20) and external (n = 18) datasets from Latin America, and on an external dataset from Europe (n = 49). We also examined robustness by rescanning six patients (n = 6) from our MS clinical cohort. Moreover, we studied inter-human annotator agreement and discussed our findings in light of these results. Performance and robustness were assessed using intraclass correlation coefficient (ICC), Dice coefficient (DC), and coefficient of variation (CV). RESULTS: Inter-human ICC ranged from 0.89 to 0.95, while spatial agreement among annotators showed a median DC of 0.63. Using expert manual segmentations as ground truth, our DL model achieved a median DC of 0.73 on the internal, 0.66 on the external, and 0.70 on the challenge datasets. The performance of our DL model exceeded that of the alternative algorithms on all datasets. In the robustness experiment, our DL model also achieved higher DC (ranging from 0.82 to 0.90) and lower CV (ranging from 0.7 to 7.9%) when compared to the alternative methods. CONCLUSION: Our DL-based model outperformed alternative methods for brain MS lesion segmentation. The model also proved to generalize well on unseen data and has a robust performance and low processing times both on real-world and challenge-based data. CLINICAL RELEVANCE STATEMENT: Our DL-based model demonstrated superior performance in accurately segmenting brain MS lesions compared to alternative methods, indicating its potential for clinical application with improved accuracy, robustness, and efficiency. KEY POINTS: • Automated lesion load quantification in MS patients is valuable; however, more accurate methods are still necessary. • A novel deep learning model outperformed alternative MS lesion segmentation methods on multisite datasets. • Deep learning models are particularly suitable for MS lesion segmentation in clinical scenarios.

A Machine Learning Approach to Personalize Computerized Cognitive Training Interventions.

Executive functions are a class of cognitive processes critical for purposeful goal-directed behavior. Cognitive training is the adequate stimulation of executive functions and has been extensively studied and applied for more than 20 years. However, there is still a lack of solid consensus in the scientific community about its potential to elicit consistent improvements in untrained domains. Individual differences are considered one of the most important factors of inconsistent reports on cognitive training benefits, as differences in cognitive functioning are both genetic and context-dependent, and might be affected by age and socioeconomic status. We here present a proof of concept based on the hypothesis that baseline individual differences among subjects would provide valuable information to predict the individual effectiveness of a cognitive training intervention. With a dataset from an investigation in which 73 6-year-olds trained their executive functions using an online software with a fixed protocol, freely available at www.matemarote.org.ar, we trained a support vector classifier that successfully predicted (average accuracy = 0.67, AUC = 0.707) whether a child would improve, or not, after the cognitive stimulation, using baseline individual differences as features. We also performed a permutation feature importance analysis that suggested that all features contribute equally to the model's performance. In the long term, this results might allow us to design better training strategies for those players who are less likely to benefit from the current training protocols in order to maximize the stimulation for each child.

The History of Writing Reflects the Effects of Education on Discourse Structure: Implications for Literacy, Orality, Psychosis and the Axial Age.

BACKGROUND: Graph analysis detects psychosis and literacy acquisition. Bronze Age literature has been proposed to contain childish or psychotic features, which would only have matured during the Axial Age (∼800-200 BC), a putative boundary for contemporary mentality. METHOD: Graph analysis of literary texts spanning ∼4,500 years shows remarkable asymptotic changes over time. RESULTS: While lexical diversity, long-range recurrence and graph length increase away from randomness, short-range recurrence declines towards random levels. Bronze Age texts are structurally similar to oral reports from literate typical children and literate psychotic adults, but distinct from poetry, and from narratives by preliterate preschoolers or Amerindians. Text structure reconstitutes the "arrow-of-time", converging to educated adult levels at the Axial Age onset. CONCLUSION: The educational pathways of oral and literate traditions are structurally divergent, with a decreasing range of recurrence in the former, and an increasing range of recurrence in the latter. Education is seemingly the driving force underlying discourse maturation.

From beat tracking to beat expectation: Cognitive-based beat tracking for capturing pulse clarity through time.

Pulse is the base timing to which western music is commonly notated, generally expressed by a listener by performing periodic taps with their hand or foot. This cognitive construction helps organize the perception of timed events in music and is the most basic expectation in rhythms. The analysis of expectations, and more specifically the strength with which the beat is felt-the pulse clarity-has been used to analyze affect in music. Most computational models of pulse clarity, and rhythmic expectation in general, analyze the input as a whole, without exhibiting changes through a rhythmic passage. We present Tactus Hypothesis Tracker (THT), a model of pulse clarity over time intended for symbolic rhythmic stimuli. The model was developed based on ideas of beat tracking models that extract beat times from musical stimuli. Our model also produces possible beat interpretations for the rhythm, a fitness score for each interpretation and how these evolve in time. We evaluated the model's pulse clarity by contrasting against tapping variability of human annotators achieving results comparable to a state-of-the-art pulse clarity model. We also analyzed the clarity metric dynamics on synthetic data that introduced changes in the beat, showing that our model presented doubt in the pulse estimation process and adapted accordingly to beat changes. Finally, we assessed if the beat tracking generated by the model was correct regarding listeners tapping data. We compared our beat tracking results with previous beat tracking models. The THT model beat tracking output showed generally correct estimations in phase but exhibits a bias towards a musically correct subdivision of the beat.

COVID-19 pneumonia accurately detected on chest radiographs with artificial intelligence.

PURPOSE: To investigate the diagnostic performance of an Artificial Intelligence (AI) system for detection of COVID-19 in chest radiographs (CXR), and compare results to those of physicians working alone, or with AI support. MATERIALS AND METHODS: An AI system was fine-tuned to discriminate confirmed COVID-19 pneumonia, from other viral and bacterial pneumonia and non-pneumonia patients and used to review 302 CXR images from adult patients retrospectively sourced from nine different databases. Fifty-four physicians blind to diagnosis, were invited to interpret images under identical conditions in a test set, and randomly assigned either to receive or not receive support from the AI system. Comparisons were then made between diagnostic performance of physicians working with and without AI support. AI system performance was evaluated using the area under the receiver operating characteristic (AUROC), and sensitivity and specificity of physician performance compared to that of the AI system. RESULTS: Discrimination by the AI system of COVID-19 pneumonia showed an AUROC curve of 0.96 in the validation and 0.83 in the external test set, respectively. The AI system outperformed physicians in the AUROC overall (70% increase in sensitivity and 1% increase in specificity, p < 0.0001). When working with AI support, physicians increased their diagnostic sensitivity from 47% to 61% (p < 0.001), although specificity decreased from 79% to 75% (p = 0.007). CONCLUSIONS: Our results suggest interpreting chest radiographs (CXR) supported by AI, increases physician diagnostic sensitivity for COVID-19 detection. This approach involving a human-machine partnership may help expedite triaging efforts and improve resource allocation in the current crisis.

Human and computer estimations of Predictability of words in written language.

When we read printed text, we are continuously predicting upcoming words to integrate information and guide future eye movements. Thus, the Predictability of a given word has become one of the most important variables when explaining human behaviour and information processing during reading. In parallel, the Natural Language Processing (NLP) field evolved by developing a wide variety of applications. Here, we show that using different word embeddings techniques (like Latent Semantic Analysis, Word2Vec, and FastText) and N-gram-based language models we were able to estimate how humans predict words (cloze-task Predictability) and how to better understand eye movements in long Spanish texts. Both types of models partially captured aspects of predictability. On the one hand, our N-gram model performed well when added as a replacement for the cloze-task Predictability of the fixated word. On the other hand, word embeddings were useful to mimic Predictability of the following word. Our study joins efforts from neurolinguistic and NLP fields to understand human information processing during reading to potentially improve NLP algorithms.

Faithfulness-boost effect: Loyal teammate selection correlates with skill acquisition improvement in online games.

The problem of skill acquisition is ubiquitous and fundamental to life. Most tasks in modern society involve the cooperation with other subjects. Notwithstanding its fundamental importance, teammate selection is commonly overlooked when studying learning. We exploit the virtually infinite repository of human behavior available in Internet to study a relevant topic in anthropological science: how grouping strategies may affect learning. We analyze the impact of team play strategies in skill acquisition using a turn-based game where players can participate individually or in teams. We unveil a subtle but strong effect in skill acquisition based on the way teams are formed and maintained during time. "Faithfulness-boost effect" provides a skill boost during the first games that would only be acquired after thousands of games. The tendency to play games in teams is associated with a long-run skill improvement while playing loyally with the same teammate significantly accelerates short-run skill acquisition.

Robust EEG-based cross-site and cross-protocol classification of states of consciousness.

Determining the state of consciousness in patients with disorders of consciousness is a challenging practical and theoretical problem. Recent findings suggest that multiple markers of brain activity extracted from the EEG may index the state of consciousness in the human brain. Furthermore, machine learning has been found to optimize their capacity to discriminate different states of consciousness in clinical practice. However, it is unknown how dependable these EEG markers are in the face of signal variability because of different EEG configurations, EEG protocols and subpopulations from different centres encountered in practice. In this study we analysed 327 recordings of patients with disorders of consciousness (148 unresponsive wakefulness syndrome and 179 minimally conscious state) and 66 healthy controls obtained in two independent research centres (Paris Pitié-Salpêtrière and Liège). We first show that a non-parametric classifier based on ensembles of decision trees provides robust out-of-sample performance on unseen data with a predictive area under the curve (AUC) of ~0.77 that was only marginally affected when using alternative EEG configurations (different numbers and positions of sensors, numbers of epochs, average AUC = 0.750 ± 0.014). In a second step, we observed that classifiers based on multiple as well as single EEG features generalize to recordings obtained from different patient cohorts, EEG protocols and different centres. However, the multivariate model always performed best with a predictive AUC of 0.73 for generalization from Paris 1 to Paris 2 datasets, and an AUC of 0.78 from Paris to Liège datasets. Using simulations, we subsequently demonstrate that multivariate pattern classification has a decisive performance advantage over univariate classification as the stability of EEG features decreases, as different EEG configurations are used for feature-extraction or as noise is added. Moreover, we show that the generalization performance from Paris to Liège remains stable even if up to 20% of the diagnostic labels are randomly flipped. Finally, consistent with recent literature, analysis of the learned decision rules of our classifier suggested that markers related to dynamic fluctuations in theta and alpha frequency bands carried independent information and were most influential. Our findings demonstrate that EEG markers of consciousness can be reliably, economically and automatically identified with machine learning in various clinical and acquisition contexts.

An entropic barriers diffusion theory of decision-making in multiple alternative tasks.

We present a theory of decision-making in the presence of multiple choices that departs from traditional approaches by explicitly incorporating entropic barriers in a stochastic search process. We analyze response time data from an on-line repository of 15 million blitz chess games, and show that our model fits not just the mean and variance, but the entire response time distribution (over several response-time orders of magnitude) at every stage of the game. We apply the model to show that (a) higher cognitive expertise corresponds to the exploration of more complex solution spaces, and (b) reaction times of users at an on-line buying website can be similarly explained. Our model can be seen as a synergy between diffusion models used to model simple two-choice decision-making and planning agents in complex problem solving.

Natural speech algorithm applied to baseline interview data can predict which patients will respond to psilocybin for treatment-resistant depression.

BACKGROUND: Natural speech analytics has seen some improvements over recent years, and this has opened a window for objective and quantitative diagnosis in psychiatry. Here, we used a machine learning algorithm applied to natural speech to ask whether language properties measured before psilocybin for treatment-resistant can predict for which patients it will be effective and for which it will not. METHODS: A baseline autobiographical memory interview was conducted and transcribed. Patients with treatment-resistant depression received 2 doses of psilocybin, 10 mg and 25 mg, 7 days apart. Psychological support was provided before, during and after all dosing sessions. Quantitative speech measures were applied to the interview data from 17 patients and 18 untreated age-matched healthy control subjects. A machine learning algorithm was used to classify between controls and patients and predict treatment response. RESULTS: Speech analytics and machine learning successfully differentiated depressed patients from healthy controls and identified treatment responders from non-responders with a significant level of 85% of accuracy (75% precision). CONCLUSIONS: Automatic natural language analysis was used to predict effective response to treatment with psilocybin, suggesting that these tools offer a highly cost-effective facility for screening individuals for treatment suitability and sensitivity. LIMITATIONS: The sample size was small and replication is required to strengthen inferences on these results.

Prediction of psychosis across protocols and risk cohorts using automated language analysis.

Language and speech are the primary source of data for psychiatrists to diagnose and treat mental disorders. In psychosis, the very structure of language can be disturbed, including semantic coherence (e.g., derailment and tangentiality) and syntactic complexity (e.g., concreteness). Subtle disturbances in language are evident in schizophrenia even prior to first psychosis onset, during prodromal stages. Using computer-based natural language processing analyses, we previously showed that, among English-speaking clinical (e.g., ultra) high-risk youths, baseline reduction in semantic coherence (the flow of meaning in speech) and in syntactic complexity could predict subsequent psychosis onset with high accuracy. Herein, we aimed to cross-validate these automated linguistic analytic methods in a second larger risk cohort, also English-speaking, and to discriminate speech in psychosis from normal speech. We identified an automated machine-learning speech classifier - comprising decreased semantic coherence, greater variance in that coherence, and reduced usage of possessive pronouns - that had an 83% accuracy in predicting psychosis onset (intra-protocol), a cross-validated accuracy of 79% of psychosis onset prediction in the original risk cohort (cross-protocol), and a 72% accuracy in discriminating the speech of recent-onset psychosis patients from that of healthy individuals. The classifier was highly correlated with previously identified manual linguistic predictors. Our findings support the utility and validity of automated natural language processing methods to characterize disturbances in semantics and syntax across stages of psychotic disorder. The next steps will be to apply these methods in larger risk cohorts to further test reproducibility, also in languages other than English, and identify sources of variability. This technology has the potential to improve prediction of psychosis outcome among at-risk youths and identify linguistic targets for remediation and preventive intervention. More broadly, automated linguistic analysis can be a powerful tool for diagnosis and treatment across neuropsychiatry.

The interpretation of dream meaning: Resolving ambiguity using Latent Semantic Analysis in a small corpus of text.

Computer-based dreams content analysis relies on word frequencies within predefined categories in order to identify different elements in text. As a complementary approach, we explored the capabilities and limitations of word-embedding techniques to identify word usage patterns among dream reports. These tools allow us to quantify words associations in text and to identify the meaning of target words. Word-embeddings have been extensively studied in large datasets, but only a few studies analyze semantic representations in small corpora. To fill this gap, we compared Skip-gram and Latent Semantic Analysis (LSA) capabilities to extract semantic associations from dream reports. LSA showed better performance than Skip-gram in small size corpora in two tests. Furthermore, LSA captured relevant word associations in dream collection, even in cases with low-frequency words or small numbers of dreams. Word associations in dreams reports can thus be quantified by LSA, which opens new avenues for dream interpretation and decoding.

Brain-heart interactions reveal consciousness in noncommunicating patients.

OBJECTIVE: We here aimed at characterizing heart-brain interactions in patients with disorders of consciousness. We tested how this information impacts data-driven classification between unresponsive and minimally conscious patients. METHODS: A cohort of 127 patients in vegetative state/unresponsive wakefulness syndrome (VS/UWS; n = 70) and minimally conscious state (MCS; n = 57) were presented with the local-global auditory oddball paradigm, which distinguishes 2 levels of processing: short-term deviation of local auditory regularities and global long-term rule violations. In addition to previously validated markers of consciousness extracted from electroencephalograms (EEG), we computed autonomic cardiac markers, such as heart rate (HR) and HR variability (HRV), and cardiac cycle phase shifts triggered by the processing of the auditory stimuli. RESULTS: HR and HRV were similar in patients across groups. The cardiac cycle was not sensitive to the processing of local regularities in either the VS/UWS or MCS patients. In contrast, global regularities induced a phase shift of the cardiac cycle exclusively in the MCS group. The interval between the auditory stimulation and the following R peak was significantly shortened in MCS when the auditory rule was violated. When the information for the cardiac cycle modulations and other consciousness-related EEG markers were combined, single patient classification performance was enhanced compared to classification with solely EEG markers. INTERPRETATION: Our work shows a link between residual cognitive processing and the modulation of autonomic somatic markers. These results open a new window to evaluate patients with disorders of consciousness via the embodied paradigm, according to which body-brain functions contribute to a holistic approach to conscious processing. Ann Neurol 2017;82:578-591.

Time to decide: Diurnal variations on the speed and quality of human decisions.

Human behavior and physiology exhibit diurnal fluctuations. These rhythms are entrained by light and social cues, with vast individual differences in the phase of entrainment - referred as an individual's chronotype - ranging in a continuum between early larks and late owls. Understanding whether decision-making in real-life situations depends on the relation between time of the day and an individual's diurnal preferences has both practical and theoretical implications. However, answering this question has remained elusive because of the difficulty of measuring precisely the quality of a decision in real-life scenarios. Here we investigate diurnal variations in decision-making as a function of an individual's chronotype capitalizing on a vast repository of human decisions: online chess servers. In a chess game, every player has to make around 40 decisions using a finite time budget and both the time and quality of each decision can be accurately determined. We found reliable diurnal rhythms in activity and decision-making policy. During the morning, players adopt a prevention focus policy (slower and more accurate decisions) which is later modified to a promotion focus (faster but less accurate decisions), without daily changes in performance.

How language flows when movements don't: An automated analysis of spontaneous discourse in Parkinson's disease.

To assess the impact of Parkinson's disease (PD) on spontaneous discourse, we conducted computerized analyses of brief monologues produced by 51 patients and 50 controls. We explored differences in semantic fields (via latent semantic analysis), grammatical choices (using part-of-speech tagging), and word-level repetitions (with graph embedding tools). Although overall output was quantitatively similar between groups, patients relied less heavily on action-related concepts and used more subordinate structures. Also, a classification tool operating on grammatical patterns identified monologues as pertaining to patients or controls with 75% accuracy. Finally, while the incidence of dysfluent word repetitions was similar between groups, it allowed inferring the patients' level of motor impairment with 77% accuracy. Our results highlight the relevance of studying naturalistic discourse features to tap the integrity of neural (and, particularly, motor) networks, beyond the possibilities of standard token-level instruments.

Impact of Universal Health Coverage on Child Growth and Nutrition in Argentina.

OBJECTIVES: To estimate trends of undernutrition (stunting and underweight) among children younger than 5 years covered by the universal health coverage programs Plan Nacer and Programa Sumar. METHODS: From 2005 to 2013, Plan Nacer and Programa Sumar collected high-quality information on birth and visit dates, age (in days), gender, weight (in kg), and height (in cm) for 1.4 million children in 6386 health centers (13 million records) with broad coverage of vulnerable populations in Argentina. RESULTS: The prevalence of stunting and underweight decreased 45.0% (from 20.6% to 11.3%) and 38.0% (from 4.0% to 2.5%), respectively, with differences between rural versus urban areas, gender, regions, age, and seasons. CONCLUSIONS: Undernutrition prevalence substantially decreased in 2 programs in Argentina as a result of universal health coverage.

Quantitative Pedagogy: A Digital Two Player Game to Examine Communicative Competence.

Inner concepts are much richer than the words that describe them. Our general objective is to inquire what are the best procedures to communicate conceptual knowledge. We construct a simplified and controlled setup emulating important variables of pedagogy amenable to quantitative analysis. To this aim, we designed a game inspired in Chinese Whispers, to investigate which attributes of a description affect its capacity to faithfully convey an image. This is a two player game: an emitter and a receiver. The emitter was shown a simple geometric figure and was asked to describe it in words. He was informed that this description would be passed to the receiver who had to replicate the drawing from this description. We capitalized on vast data obtained from an android app to quantify the effect of different aspects of a description on communication precision. We show that descriptions more effectively communicate an image when they are coherent and when they are procedural. Instead, the creativity, the use of metaphors and the use of mathematical concepts do not affect its fidelity.

Neglect in human communication: quantifying the cost of cell-phone interruptions in face to face dialogs.

There is a prevailing belief that interruptions using cellular phones during face to face interactions may affect severely how people relate and perceive each other. We set out to determine this cost quantitatively through an experiment performed in dyads, in a large audience in a TEDx event. One of the two participants (the speaker) narrates a story vividly. The listener is asked to deliberately ignore the speaker during part of the story (for instance, attending to their cell-phone). The speaker is not aware of this treatment. We show that total amount of attention is the major factor driving subjective beliefs about the story and the conversational partner. The effects are mostly independent on how attention is distributed in time. All social parameters of human communication are affected by attention time with a sole exception: the perceived emotion of the story. Interruptions during day-to-day communication between peers are extremely frequent. Our data should provide a note of caution, by indicating that they have a major effect on the perception people have about what they say (whether it is interesting or not . . .) and about the virtues of the people around them.