A BBD model of the medial temporal lobe

Place fields may be journey-independent or depend on the recent past or near future path. Rat hippocampus has neurons that have all three types of firing fields, as does the simulated hippocampus of Darwin X and XI.

Information from many different sensory modalities converges on the medial temporal lobe in the mammalian brain, an area that is known to be involved in the formation of spatial, multi-modal, and episodic memories. Once information reaches the medial temporal lobe it flows through a variety of nested loops before once again diverging to other cortical areas. We have hypothesized that the loops of highly processed information over multiple timescales are important to episodic memory formation.

In order to test this idea we have built two Brain-Based Device (BBD) models, Darwins X and XI, that replicate studies of rodent navigation in an artificial organism. These models have large-scale neural simulations (~100k neuronal units, 1.2M synapses) of the medial temporal lobe and surrounding cortical regions. The simulations have realistic neuroanatomy and neurophysiology, thus allowing us to ask questions about how the unique structure of the medial temporal lobe contributes to memory formation. Using an artificial organism allows us to look at the activity of every neuron in the simulated nervous system simultaneously, something that is currently impossible to do in behaving animals.

Darwin X in its open field environment

Darwin X performed a dry-land version of the Morris water-maze, a standard test of spatial memory in rodents. In this task the BBD was allowed to explore an open-field arena with visual landmarks on its edges. A hidden platform, a unconditionally rewarding stimulus that can only be sensed by touch, was placed in one location in the maze. After less than 20 trials Darwin X could navigate directly to this hidden location from any starting position in the maze. Neuronal units in the CA1 region of the simulated hippocampus developed place field activity similar to that seen in rodents without this being designed a priori into the simulation. Analysis of the simulation has shown that the tri-synaptic pathway is more prevalent in producing place field activity in early-learning trials, but less prevalent after the BBD has more experience at the task [See this paper]. The tri-synaptic pathway also produces different types of place activity than the perforant path [See this paper]. For a video of Darwin X's performance on the dry-land version of the Morris water-maze click [here].

Darwin XI in its plus maze environment

Darwin XI was designed to investigate the integration of many different kinds of sensory information into a unitary memory. Darwin XI added two new sensory inputs to the Darwin X model: artificial whiskers for texture sensing and a laser rangefinder that provides a sense of distance to obstacles. Darwin XI performed a plus-maze task which is used in rodents to investigate journey-dependent, i.e. contextual, activity in the hippocampus. Journey-dependent activity occurs when cells fire with place fields only if the animal has taken a particular path to reach that place, but not if the animal has reached that place via a different path.

We found that Darwin XI displayed journey-dependent place activity similar to that seen in rodents in its simulated hippocampus. This contextual activity emerged from the realistic neuroanatomy in the model and experience-dependent synaptic plasticity. Journey-dependent place activity was found to be caused by information flowing thorugh the tri-synaptic pathway more often than the journey-independent (non-contextual) place activity [See this paper].