Reinforcement learning in dendritic structures

The discovery of binary dendritic events such as local NMDA spikes in dendritic subbranches led to the suggestion that dendritic trees could be computationally equivalent to a 2-layer network of point neurons, with a single output unit represented by the soma, and input units represented by the dendritic branches. Although this interpretation endows a neuron with a high computational power, it is functionally not clear why nature would have preferred the dendritic solution with a single but complex neuron, as opposed to the network solution with many but simple units. We show that the dendritic solution has a distinguished advantage over the network solution when considering different learning tasks. Its key property is that the dendritic branches receive an immediate feedback from the somatic output spike, while in the corresponding network architecture the feedback would require additional backpropagating connections to the input units. Assuming a reinforcement learning scenario we formally derive a learning rule for the synaptic contacts on the individual dendritic trees which depends on the presynaptic activity, the local NMDA spikes, the somatic action potential, and a delayed reinforcement signal. We test the model for two scenarios: the learning of binary classifications and of precise spike timings. We show that the immediate feedback represented by the backpropagating action potential supplies the individual dendritic branches with enough information to efficiently adapt their synapses and to speed up the learning process.

Reinforcement learning in dendritic structures

The discovery of binary dendritic events such as local NMDA spikes in dendritic subbranches led to the suggestion that dendritic trees could be computationally equivalent to a 2-layer network of point neurons, with a single output unit represented by the soma, and input units represented by the dendritic branches. Although this interpretation endows a neuron with a high computational power, it is functionally not clear why nature would have preferred the dendritic solution with a single but complex neuron, as opposed to the network solution with many but simple units. We show that the dendritic solution has a distinguished advantage over the network solution when considering different learning tasks. Its key property is that the dendritic branches receive an immediate feedback from the somatic output spike, while in the corresponding network architecture the feedback would require additional backpropagating connections to the input units. Assuming a reinforcement learning scenario we formally derive a learning rule for the synaptic contacts on the individual dendritic trees which depends on the presynaptic activity, the local NMDA spikes, the somatic action potential, and a delayed reinforcement signal. We test the model for two scenarios: the learning of binary classifications and of precise spike timings. We show that the immediate feedback represented by the backpropagating action potential supplies the individual dendritic branches with enough information to efficiently adapt their synapses and to speed up the learning process.

Identification of brain structures involved in micturition with functional magnetic resonance imaging (fMRI)

OBJECTIVE: The voluntary control of micturition is believed to be integrated by complex interactions among the brainstem, subcortical areas and cortical areas. Several brain imaging studies using positron emission tomography (PET) have demonstrated that frontal brain areas, the limbic system, the pons and the premotor cortical areas were involved. However, the cortical and subcortical brain areas have not yet been precisely identified and their exact function is not yet completely understood. MATERIALS AND METHODS: This study used functional magnetic resonance imaging (fMRI) to compare brain activity during passive filling and emptying of the bladder. A cathetherism of the bladder was performed in seven healthy subjects (one man and six right-handed women). During scanning, the bladder was alternatively filled and emptied at a constant rate with bladder rincing solution. RESULTS: Comparison between passive filling of the bladder and emptying of the bladder showed an increased brain activity in the right inferior frontal gyrus, cerebellum, symmetrically in the operculum and mesial frontal. Subcortical areas were not evaluated. CONCLUSIONS: Our results suggest that several cortical brain areas are involved in the regulation of micturition.

Whole body postmortem angiography with a high viscosity contrast agent solution using poly ethylene glycol as contrast agent dissolver

Postmortem minimal invasive angiography has already been implemented to support virtual autopsy examinations. An experimental approach in a porcine model to overcome an initially described artificial tissue edema artifact by using a poly ethylene glycol (PEG) containing contrast agent solution showed promising results. The present publication describes the first application of PEG in a whole corpse angiographic CT examination. A minimal invasive postmortem CT angiography was performed in a human corpse utilizing the high viscosity contrast agent solution containing 65% of PEG. Injection was carried out via the femoral artery into the aortic root in simulated cardiac output conditions. Subsequent CT scanning delivered the 3D volume data of the whole corpse. Visualization of the human arterial anatomy was excellent and the contrast agent distribution was generally limited to the arterial system as intended. As exceptions an enhancement of the brain, the left ventricular myocardium and the renal cortex became obvious. This most likely represented the stage of centralization of the blood circulation at the time of death with dilatation of the precapillary arterioles within these tissues. Especially for the brain this resulted in a distinctively improved visualization of the intracerebral structures by CT. However, the general tissue edema artifact of postmortem minimal invasive angiography examinations could be distinctively reduced.