Automatic signal adaptation for real-time bidirectional interaction with the nervous system

• Autores: Manuel Reyes Sánchez
• Directores de la Tesis: Pablo Varona (dir. tes.)
• Lectura: En la Universidad Autónoma de Madrid ( España ) en 2023
• Idioma: inglés
• Número de páginas: 121
• Títulos paralelos:
  • Adaptación automática de señales para interacción bidireccional en tiempo real con el sistema nervioso
• Tribunal Calificador de la Tesis: Marcelo Reyes (presid.) , Eduardo Serrano Jerez (secret.) , Reynaldo Pinto (voc.)
• Enlaces
  • Tesis en acceso abierto en: Biblos-e Archivo
• Resumen

  • Biohybrid neural circuits connect artificial and living neurons. These circuits have proven to be useful for addressing neuronal, synaptic and network dynamics in neuroscience research and provide realistic closed-loop interactions to characterize information processing phenomena in the nervous system. However, biohybrid circuits require experiment-specific adaptations to establish effective communication between living and artificial entities, which are non-trivial and time-consuming. Artificial neuron models and connections need time and amplitude scaling to match those observed in their biological counterparts. Also, the interaction in a biohybrid circuit requires real-time management as model neurons and connections should work with precision and at the same speed as living cells. These difficulties prevent hybrid circuits from being widely employed in the neuroscience experimental community. The goal of this thesis is to expand the use of biohybrid circuits by providing effective automatic calibration and adaptation protocols. The first part of this thesis assesses the development and validation of algorithms to autocalibrate and connect biological and model neurons in electrophysiological setups. This covers the online detection of relevant events in the neuronal activity needed to implement the synaptic currents, the assessment of how to scale model neurons in time and amplitude, and the compensation and mitigation of problems derived from the invasive experimental techniques used such as a voltage drift compensation mechanism. The proposed tools for hybrid circuit construction were validated in a well-known central pattern generator circuit. The algorithms had been coded to work in real-time Linux systems and implemented in an in-house platform, RTHybrid, that has been developed to be used with Preempt-RT and Xenomai 3 Linux hard real-time patches. We also created modules for RTXI to make our protocols available in this hard real-time data acquisition and control application for biological research. The second part of the thesis explores the search for sequential dynamical invariants in hybrid circuits. A sequential dynamical invariant is a robust cycle-by-cycle relationship between the intervals that build a neural sequence from a network interaction. This phenomenon has been observed and characterized in biological central pattern generators. This work reproduces sequential dynamical invariants in hybrid circuits for the first time using high-performance computing resources to run thousands of simulations and genetic algorithms to tune synapses in real-time in hybrid experiments. The third part of the thesis provides additional tools to build protocols for closed-loop interactions with nervous systems and illustrates the use of the developed algorithms beyond the biohybrid circuits, including infrared activity-dependent laser stimulation and the construction of hybrots, i.e., biohybrid robots, whose locomotion is controlled by sequential dynamical invariants present in the activity of a biological circuit. In these closed-loop experiments, the developed automatic adaptations play a key role, allowing effective communication between living cells and machines. In short, this thesis provides for the first time a set of high-value tools for the automatic calibration and adaptation of closed-loop interactions between living and artificial entities, with a focus on biohybrid circuit construction. These tools have already been proven effective and validated in novel experiments. The developed software is publicly available under an open-source license, and the associated concepts and methodology aim to generalize and expand the use of biohybrid protocols in modern neurotechnology experiments and applications