Focused ultrasound neuromodulation of mediodorsal thalamus disrupts decision flexibility during reward learning

When learning to find the most beneficial course of action, the prefrontal cortex guides decisions by comparing estimates of the relative value of the options available. Basic neuroscience studies in animals support the view that the thalamus can regulate this activity within and across the prefrontal cortex. We studied a group of patients (n=37) undergoing unilateral MR guided focused ultrasound for essential tremor, performing the restless bandit, a reward reinforcement learning task, immediately before and after thalamotomy. Thalamotomy significantly impaired the proportion of switch choices during the task without affecting overall performance. This effect was observed when the task was delivered to co-incide with maximal vasogenic thalamic oedema but not in a control group tested on the same day of their treatment. A reinforcement learning model fitted to the patients choices replicated the effect of thalamotomy when the model increased exploitation of the bandits learnt value estimate. This shift in the explore-exploit trade-off, manifesting as reduced choice flexibility, co-varied with the pattern of post-operative oedema extension into mediodorsal nucleus. These findings confirm a causal role of the thalamus and specifically the mediodorsal nucleus, in regulating the extent to which value estimates are used to guide decisions and learning from reward.