• Digital Humanities and Cultural Heritage
• 2017-2021close
• Publications close
• Research software close
• Other research products close
• English close
• EBRAINS close

Brain mapping studies the spatial organization of the brain and associationsbetween mental functions or diseases and anatomical structures.It relies on structural and functional neuroimaging techniques such asfMRI.The neuroimaging literature is growing rapidly, with thousands of publicationsevery year.Extracting general and reliable knowledge about the brain from this hugeamount of results is difficult.Indeed, individual studies lack statistical power and report many spuriousfindings. Even genuine effects are often specific to particular experimentalsettings and difficult to reproduce.Moreover, the mechanisms under study are very diverse, and terminology is notyet fully formalized. Therefore, integrating results across experiments andresearch domains is challenging.Meta-analysis aggregates studies to identify consistent trends in reportedassociations between brain structure and behavior.The standard approach to meta-analysis starts by gathering a sample of studiesthat investigate a same mental process or disease. Then, a statistical test isperformed at every voxel to delineate brain regions where there is a significantagreement among reported findings.In this thesis, we develop a different kind of meta-analysis that focuses onprediction rather than hypothesis testing. We build predictive models that maptextual descriptions of experiments, mental processes or diseases to anatomicalregions in the brain.We use multivariate models to combine terms used to describe experimentsor observations rather than studying each concept in isolation.We introduce models that learn from very few examples, thus extending the scopeof meta-analysis to seldom-studied diseases and mental processes.Our supervised learning approach comes with a natural quantitative evaluationframework, and we conduct extensive experiments to validate and comparestatistical models.To train these models, we collect and share the largest existing dataset ofneuroimaging studies and stereotactic coordinates. This dataset contains thefull text and locations of neurological observations for over 13,000publications.In the last part, we turn to decoding: inferring mental states from brainactivity.We perform this task through meta-analysis of fMRI statistical mapscollected from an online data repository. We use fMRI data to distinguish awide range of mental conditions.Standard meta-analysis is an essential tool to distinguish true discoveries fromnoise and artifacts. This thesis introduces methods for predictivemeta-analysis, which complement the standard approach and helpinterpret neuroimaging results and formulate hypotheses or formalstatistical priors.; La neuroimagerie permet d'étudier les liens entre la structure et lefonctionnement du cerveau.La littérature de neuroimagerie croît rapidement, avec des milliers depublications par an.Il est difficile d'extraire des connaissances sur le fonctionnement du cerveaude cette grande quantité de résultats.En effet, chaque étude manque de puissance statistique et peut reporter beaucoupde faux positifs. De plus, certains effets sont spécifiques à unprotocole expérimental et difficiles à reproduire.Les méta-analyses rassemblent plusieurs études pour identifier les associationsentre structures anatomiques et processus cognitifs qui sont établies de manièreconsistente dans la littérature.Les méthodes classiques de méta-analyse commencent par constituer un échantillond'études focalisées sur un même processus mental ou une même maladie. Ensuite,un test statistique est effectué pour chaque voxel, afin de délimiter lesrégions cérébrales dans lesquelles le nombre d'observations reportées estsignificatif.Dans cette thèse, nous introduisons une nouvelle forme de méta-analyse, quis'attache à construire des prédictions plutôt qu'à tester des hypothèses. Nousintroduisons des modèles statistiques qui prédisent la distribution spatiale desobservations neurologiques à partir de la description textuelle d'uneexpérience, d'un processus cognitif ou d'une maladie cérébrale.Nos modèles peuvent apprendre avec très peu d'exemples à associer une cartecérébrale à une fonction ou une maladie mentale.Notre approche est basée sur l'apprentissage statistique supervisé quifournit un cadre classique pour évaluer et comparer les modèles.Pour entraîner nos modèles, nous construisons le plus grand jeu de donnéesd'études de neuroimagerie et de coordonnées stéréotaxiques existant, quirassemble plus de 13,000 publications.Dans la dernière partie, nous nous intéressons au décodage, qui consisteà prédire des états psychologiques à partir de l'activité cérébrale.Nous apprenons à décoder des images de cerveau à travers la méta-analyse dedonnées d'IRM fonctionnelle (IRMf). Les images d'IRMf nous permettent declassifier un grand nombre d'états psychologiques.La méta-analyse standard est un outil indispensable pour distinguer les vraiesdécouvertes du bruit et des artefacts parmi les résultats publiés enneuroimagerie. Cette thèse introduit des méthodes adaptées à la méta-analyseprédictive. Cette approche est complémentaire de la méta-analyse standard, etaide à interpréter les résultats d'études de neuroimagerie ainsi qu'à formulerdes hypothèses ou des a priori statistiques.

Qu’il existe un lien entre vision et connaissance ne peut être nié. Cependant ce lien est loin d’être direct, et peut s’exercer dans les deux directions. Pour préciser cette liaison il convient d’abord d’établir une distinction entre information et connaissance. Ensuite il faut examiner ce qui, dans la perception visuelle, revient véritablement au stimulus, et ce qui résulte d’une élaboration périphérique. Enfin il faut se demander comment peut s’opérer, à partir d’une saisie visuelle globale, une transposition dans un langage chronosyntaxique.Ces préalables étant posés, on pourra analyser une série d’exemples, tirés des domaines les plus divers, et illustrant le rôle du visuel dans la découverte scientifique, technique ou mathématique, dans la pratique didactique, et enfin dans l’imagination symbolique ou poétique. That a link exists between vision and knowledge cannot be denied, but this link is by no means straightforward, and can function in either direction. In order to describe it accurately, the distinction between information and knowledge must first be clarified. We must then examine that which, in our visual perception, really stems from the stimuli, as opposed to that which results from some reworking by the brain. And finally we need to investigate how a global visual image can be transformed into a linear verbal sequence. We may then analyze a series of examples, some very famous, that illustrate the role of vision in such fields as epistemology, general science, techniques, teaching and finally in poetic and symbolic imagination.

Brain mapping studies the spatial organization of the brain and associationsbetween mental functions or diseases and anatomical structures.It relies on structural and functional neuroimaging techniques such asfMRI.The neuroimaging literature is growing rapidly, with thousands of publicationsevery year.Extracting general and reliable knowledge about the brain from this hugeamount of results is difficult.Indeed, individual studies lack statistical power and report many spuriousfindings. Even genuine effects are often specific to particular experimentalsettings and difficult to reproduce.Moreover, the mechanisms under study are very diverse, and terminology is notyet fully formalized. Therefore, integrating results across experiments andresearch domains is challenging.Meta-analysis aggregates studies to identify consistent trends in reportedassociations between brain structure and behavior.The standard approach to meta-analysis starts by gathering a sample of studiesthat investigate a same mental process or disease. Then, a statistical test isperformed at every voxel to delineate brain regions where there is a significantagreement among reported findings.In this thesis, we develop a different kind of meta-analysis that focuses onprediction rather than hypothesis testing. We build predictive models that maptextual descriptions of experiments, mental processes or diseases to anatomicalregions in the brain.We use multivariate models to combine terms used to describe experimentsor observations rather than studying each concept in isolation.We introduce models that learn from very few examples, thus extending the scopeof meta-analysis to seldom-studied diseases and mental processes.Our supervised learning approach comes with a natural quantitative evaluationframework, and we conduct extensive experiments to validate and comparestatistical models.To train these models, we collect and share the largest existing dataset ofneuroimaging studies and stereotactic coordinates. This dataset contains thefull text and locations of neurological observations for over 13,000publications.In the last part, we turn to decoding: inferring mental states from brainactivity.We perform this task through meta-analysis of fMRI statistical mapscollected from an online data repository. We use fMRI data to distinguish awide range of mental conditions.Standard meta-analysis is an essential tool to distinguish true discoveries fromnoise and artifacts. This thesis introduces methods for predictivemeta-analysis, which complement the standard approach and helpinterpret neuroimaging results and formulate hypotheses or formalstatistical priors. La neuroimagerie permet d'étudier les liens entre la structure et lefonctionnement du cerveau.La littérature de neuroimagerie croît rapidement, avec des milliers depublications par an.Il est difficile d'extraire des connaissances sur le fonctionnement du cerveaude cette grande quantité de résultats.En effet, chaque étude manque de puissance statistique et peut reporter beaucoupde faux positifs. De plus, certains effets sont spécifiques à unprotocole expérimental et difficiles à reproduire.Les méta-analyses rassemblent plusieurs études pour identifier les associationsentre structures anatomiques et processus cognitifs qui sont établies de manièreconsistente dans la littérature.Les méthodes classiques de méta-analyse commencent par constituer un échantillond'études focalisées sur un même processus mental ou une même maladie. Ensuite,un test statistique est effectué pour chaque voxel, afin de délimiter lesrégions cérébrales dans lesquelles le nombre d'observations reportées estsignificatif.Dans cette thèse, nous introduisons une nouvelle forme de méta-analyse, quis'attache à construire des prédictions plutôt qu'à tester des hypothèses. Nousintroduisons des modèles statistiques qui prédisent la distribution spatiale desobservations neurologiques à partir de la description textuelle d'uneexpérience, d'un processus cognitif ou d'une maladie cérébrale.Nos modèles peuvent apprendre avec très peu d'exemples à associer une cartecérébrale à une fonction ou une maladie mentale.Notre approche est basée sur l'apprentissage statistique supervisé quifournit un cadre classique pour évaluer et comparer les modèles.Pour entraîner nos modèles, nous construisons le plus grand jeu de donnéesd'études de neuroimagerie et de coordonnées stéréotaxiques existant, quirassemble plus de 13,000 publications.Dans la dernière partie, nous nous intéressons au décodage, qui consisteà prédire des états psychologiques à partir de l'activité cérébrale.Nous apprenons à décoder des images de cerveau à travers la méta-analyse dedonnées d'IRM fonctionnelle (IRMf). Les images d'IRMf nous permettent declassifier un grand nombre d'états psychologiques.La méta-analyse standard est un outil indispensable pour distinguer les vraiesdécouvertes du bruit et des artefacts parmi les résultats publiés enneuroimagerie. Cette thèse introduit des méthodes adaptées à la méta-analyseprédictive. Cette approche est complémentaire de la méta-analyse standard, etaide à interpréter les résultats d'études de neuroimagerie ainsi qu'à formulerdes hypothèses ou des a priori statistiques.

Music is known to evoke emotions through a range of mechanisms, but empirical investigation into the mechanisms underlying different emotions is sparse. This study investigated how affective experiences to music and pictures vary when induced by personal memories or mere stimulus features. Prior to the experiment, participants were asked to select eight types of stimuli according to distinct criteria concerning the emotion induction mechanism and valence. In the experiment, participants (N = 30) evaluated their affective experiences with the self-chosen material. EEG was recorded throughout the session. The results showed certain interaction effects of mechanism (memory vs. stimulus features), emotional valence of the stimulus (pleasant vs. unpleasant), and stimulus modality (music vs. pictures). While effects were mainly similar in music and pictures, the findings suggest that when personal memories are involved, stronger positive emotions were experienced in the context of music, even when the music was experienced as unpleasant. Memory generally enhanced social emotions specifically in pleasant conditions. As for sadness and melancholia, stimulus features did not evoke negative experiences; however, these emotions increased strongly with the involvement of memory, particularly in the condition of unpleasant music. Analysis of EEG-data corroborated the findings by relating frontomedial theta activity to memory-evoking material. peerReviewed

BACKGROUND AND OBJECTIVES:The kidney stone's structure might provide clinical information in addition to the stone composition. The Raman chemical imaging is a technology used for the production of two-dimension maps of the constituents' distribution in samples. We aimed at determining the use of Raman chemical imaging in urinary stone analysis. MATERIAL AND METHODS:Fourteen calculi were analyzed by Raman chemical imaging using a confocal Raman microspectrophotometer. They were selected according to their heterogeneous composition and morphology. Raman chemical imaging was performed on the whole section of stones. Once acquired, the data were baseline corrected and analyzed by MCR-ALS. Results were then compared to the spectra obtained by Fourier Transform Infrared spectroscopy. RESULTS:Raman chemical imaging succeeded in identifying almost all the chemical components of each sample, including monohydrate and dihydrate calcium oxalate, anhydrous and dihydrate uric acid, apatite, struvite, brushite, and rare chemicals like whitlockite, ammonium urate and drugs. However, proteins couldn't be detected because of the huge autofluorescence background and the small concentration of these poor Raman scatterers. Carbapatite and calcium oxalate were correctly detected even when they represented less than 5 percent of the whole stones. Moreover, Raman chemical imaging provided the distribution of components within the stones: nuclei were accurately identified, as well as thin layers of other components. Conversion of dihydrate to monohydrate calcium oxalate was correctly observed in the centre of one sample. The calcium oxalate monohydrate had different Raman spectra according to its localization. CONCLUSION:Raman chemical imaging showed a good accuracy in comparison with infrared spectroscopy in identifying components of kidney stones. This analysis was also useful in determining the organization of components within stones, which help locating constituents in low quantity, such as nuclei. However, this analysis is time-consuming, making it more suitable for research studies rather than routine analysis.

Successful diagnosis and management of neurological dysfunction relies on proper communication between the neurologist and the primary physician (or other specialists). Because this communication is documented within medical records, the ability to automatically infer the clinical correlations for a patient from his or her medical records would provide an important step towards enabling health care systems to automatically identify patients requiring additional follow-up as well as flagging any unexpected clinical correlations for review. In this paper, we present a Deep Section Recovery Model (DSRM) which applies deep neural learning on a large body of EEG reports in order to infer the expected clinical correlations for a patient from the information in a given EEG report by (1) automatically extracting word- and report- level features from the report and (2) inferring the most likely clinical correlations and expressing those clinical correlations in natural language. We evaluated the performance of the DSRM by removing the clinical correlation sections from EEG reports and measuring how well the model could recover that information from the remainder of the report. The DSRM obtained a 17% improvement over the top-performing baseline, highlighting not only the power of the DSRM but also the promise of automatically recognizing unexpected clinical correlations in the future.

This work consists of four projects that explore human behavior from two perspectives: language use and neural patterns.In my first and second projects, I focused on language, which can be used to categorize human behavior. In the first project, I used topic models to categorize the subjects and symptoms a patient discussed during psychotherapy treatment. The model functions by identifying topics that are representative of each subject or symptom. The model can predict the subjects and symptoms discussed in new therapy sessions with higher accuracy than discriminative techniques. Furthermore, the model can identify specific passages of text representative of a given subject or symptom.My second project developed an automated system for routing citizen requests to federal agencies within the Mexican government. The automated system functions by linking pat- terns in language and the appropriate federal agency. The automated system routes requests more efficiently than the current routing system.The third and fourth projects focused on neuroimaging, which is used to understand the underlying neural processes associated with human behavior. My neuroimaging work related blood-oxygen-level-dependent (BOLD) variability (BV) to experimental condition, behavior, and subject identity. The first phase of the neural work built on previous analyses showing that functional connectivity (FC) is predictive of the task a subject is performing and the identity of the subject performing a task. We extended these analyses to BV and compared its predictive accuracy with that of FC to assess whether some of the predictive power of FC is due to changes in BV. BV performed well compared to FC, suggesting that some of the predictive performance based on FC might be attributed to independent region-specific fluctuations.Given the predictive relationship between BV and task/subject, the second phase of my neu- ral work developed the Variance Design General Linear Model (VDGLM), a novel framework to facilitate the detection of BV effects. The framework models the mean and variance in the BOLD time course as functions of experimental design. This allows the VDGLM to i) simul- taneously make inferences about a mean or variance effect while controlling for the other and ii) test for variance effects that could be associated with multiple conditions and/or noise regressors. We demonstrated the use of the VDGLM in a working memory application and showed that engagement in a working memory task is associated with whole-brain decreases in BOLD variance.

This work consists of four projects that explore human behavior from two perspectives: language use and neural patterns.In my first and second projects, I focused on language, which can be used to categorize human behavior. In the first project, I used topic models to categorize the subjects and symptoms a patient discussed during psychotherapy treatment. The model functions by identifying topics that are representative of each subject or symptom. The model can predict the subjects and symptoms discussed in new therapy sessions with higher accuracy than discriminative techniques. Furthermore, the model can identify specific passages of text representative of a given subject or symptom.My second project developed an automated system for routing citizen requests to federal agencies within the Mexican government. The automated system functions by linking pat- terns in language and the appropriate federal agency. The automated system routes requests more efficiently than the current routing system.The third and fourth projects focused on neuroimaging, which is used to understand the underlying neural processes associated with human behavior. My neuroimaging work related blood-oxygen-level-dependent (BOLD) variability (BV) to experimental condition, behavior, and subject identity. The first phase of the neural work built on previous analyses showing that functional connectivity (FC) is predictive of the task a subject is performing and the identity of the subject performing a task. We extended these analyses to BV and compared its predictive accuracy with that of FC to assess whether some of the predictive power of FC is due to changes in BV. BV performed well compared to FC, suggesting that some of the predictive performance based on FC might be attributed to independent region-specific fluctuations.Given the predictive relationship between BV and task/subject, the second phase of my neu- ral work developed the Variance Design General Linear Model (VDGLM), a novel framework to facilitate the detection of BV effects. The framework models the mean and variance in the BOLD time course as functions of experimental design. This allows the VDGLM to i) simul- taneously make inferences about a mean or variance effect while controlling for the other and ii) test for variance effects that could be associated with multiple conditions and/or noise regressors. We demonstrated the use of the VDGLM in a working memory application and showed that engagement in a working memory task is associated with whole-brain decreases in BOLD variance.