Temporal and spatial patterns differences of basal and cognitive brain states

• Autores: Katerina Capouskova
• Directores de la Tesis: Gustavo Deco (dir. tes.)
• Lectura: En la Universitat Pompeu Fabra ( España ) en 2023
• Idioma: español
• Tribunal Calificador de la Tesis: Joana Cabral (presid.), Murat Demirtas (secret.), Albert Compte Braquets (voc.)
• Programa de doctorado: Programa de Doctorado en Biomedicina por la Universidad Pompeu Fabra
• Materias:
  • Psicología
    • Psicología experimental
      • Actividad cerebral
• Enlaces
  • Tesis en acceso abierto en: TDX
• Resumen
  • English

    Dynamic brain states, temporal functional unions of diverse brain areas, are the driving force in people’s everyday cognition. They help individuals flexibly integrate and segregate the assorted information with which the environment provides them. Previous research has focussed primarily on resting because it serves as a baseline. In our work, we included six cognitively demanding tasks and rest conditions to identify and explore spatiotemporal patterns of functional connectivity characterised in a probabilistic metastable substate space.

    In the first study presented in the thesis, we identified seven general recurrent metastable substates with our novel analysis framework (DADA) that uses an autoencoder, an artificial neural network, as a dimensionality reduction method. In the pipeline, we also used clustering model to separate between the substates. The substates are characterized through probability of occurrences and lifetimes for each of the six cognitive task and in rest, in addition with their entropy, probability of switching between substates, and the distances between probability distributions measured by Kullback- Leibler divergence. We found that the brain tends to occupy states with more connectivity in tasks and states with less connectivity in rest and that rest has significantly highest entropy.

    The second study in this thesis explored particular cases of metastable substates that are integrated and segregated. We estimated modularity and global efficiency for a dynamic functional connectivity matrix for each time point to cluster between segregated substate with high modularity and low global efficiency and integrated substate with low modularity and high global efficiency. Through latent space of an autoencoder, we were able to identify that integration serves as a data compression with lower entropy and that segregation means specialization with high entropy. We also show that modularity and global efficiency are good predictors for classification between tasks. We were able to separate between tasks and rest in the integrated and segregated substates with scores above 90% of accuracy.

  • English

    Los estados cerebrales dinámicos, unidades funcionales temporales de diversas áreas del cerebro, son el motor de nuestra cognición diaria. Nos ayudan tanto a integrar de manera flexible como a segregar la información presente en el ambiente. Investigaciones previas se han enfocado primariamente en estados cerebrales en reposo, ya que sirven como referencia de base. En nuestro trabajo, incluimos seis tareas cognitivamente demandantes y condiciones en reposo para identificar y explorar patrones espacio-temporales de conectividad funcional en un subespacio probabilístico de estados metaestables. Con el uso de una red neuronal para reducir dimensiones y un autoencoder, desarrollamos una nueva estrategia para el análisis de datos que permite la exploración de subestados metaestables. Hallamos que el cerebro tiende a ocupar estados con mayor conectividad en tareas demandantes y estados con menor conectividad en reposo. También identificamos estados integrados y segregados, donde 'integración' se refiere a la compresión de datos con menor entropía, mientras que 'segregación' significa especialización con entropía alta.