Computational analysis of neural recordings
Two 2019 studies examine electrophysiological recordings using computational methods. One applies pattern recognition to large-scale local field potentials, the combined electrical signal of many nearby neurons, to distinguish signals from deep and superficial macaque cortical layers; another studies automatic cortex-wide layer identification and neuronal rhythms.
Together, the papers show how signal processing and machine learning can support analysis of neural recordings. One paper reports that an ensemble machine-learning approach identified the layers with good accuracy in both prefrontal and visual cortex; the other introduces two automatic layer-identification methods and a layer-specific pattern in neuronal rhythms across frontal, parietal, and visual cortex.
Neural decoding and technical education
In April 2019, Dr. Costilla Reyes taught a computational tutorial at MIT on neural decoding with Python; neural decoding uses machine-learning models to predict a stimulus, a behavior, or a cognitive state from recorded brain activity. The session covered spike trains (records of when individual neurons fire), local field potentials, data preprocessing, model selection, and optimization; MIT OpenCourseWare and the Center for Brains, Minds & Machines provide related learning materials.
The tutorial translated machine-learning methods into practical analysis of neural populations and computational neuroscience.
Research exchange between AI and neuroscience
At the 2019 Society for Neuroscience annual meeting, Dr. Costilla Reyes co-organized and spoke at a minisymposium on artificial intelligence and neuroscience. The institutional event report describes discussion of machine-learning analysis of neural signals recorded during cognitive tasks.
The papers, tutorial, and symposium show Dr. Costilla Reyes’s contributions to research and teaching across neural decoding, computational neuroscience, and artificial intelligence.