Fig. 1: Multiregional neural recordings and prediction of neural activity from video.
From: Brain-wide analysis reveals movement encoding structured across and within brain areas
a, Delayed response task. b, Simultaneous video and neural recording. c, Example recording configuration. d, Raster plot of recorded neurons (top) and traces of body part marker locations for a single trial. e, Three approaches to predict neural activity from video. Left: marker-based analysis. For each video frame, each of the markers (jaw, nose or tongue) is a two-dimensional vector representing the vertical and horizontal positions. Middle: embedding-based analysis. For each frame, the embedding vector is a 16-dimensional vector. Right: end-to-end learning with deep neural network. ALMc, contralateral ALM; Cb, cerebellum; CNN, convolutional neural network; Ctx, cortex; Med, medulla; MRN, midbrain reticular nucleus; PN, pons; PPN, pedunculopontine nucleus; px, pixels; SCm, superior colliculus, motor related; SNr, substantia nigra, reticular part; Str, striatum; Thal, thalamus.
