On the role of cortex-basal ganglia interactions for category learning: A neuro-computational approach
Francesc Villagrasa, Javier Baladron, Julien Vitay, Henning Schroll, Evan G. Antzoulatos, Earl K. Miller and Fred H. Hamker
Journal of Neuroscience 18 September 2018, 0874-18; DOI: https://doi.org/10.1523/JNEUROSCI.0874-18.2018
In addition to the prefrontal cortex (PFC), the basal ganglia (BG) have been increasingly often reported to play a fundamental role in category learning, but the systems-level circuits of how both interact remain to be explored. We developed a novel neuro-computational model of category learning that particularly addresses the BG-PFC interplay. We propose that the BG bias PFC activity by removing the inhibition of cortico-thalamo-cortical loop and thereby provide a teaching signal to guide the acquisition of category representations in the cortico-cortical associations to the PFC. Our model replicates key behavioral and physiological data of macaque monkey learning a prototype distortion task from Antzoulatos and Miller (2011). Our simulations allowed us to gain a deeper insight into the observed drop of category selectivity in striatal neurons seen in the experimental data and in the model. The simulation results and a new analysis of the experimental data, based on the model’s predictions, show that the drop in category selectivity of the striatum emerges as the variability of responses in the striatum rises when confronting the BG with an increasingly larger number of stimuli to be classified. The neuro-computational model therefore provides new testable insights of systems-level brain circuits involved in category learning which may also be generalized to better understand other cortico-basal ganglia-cortical loops
A computational model of visual categorization in cortex that has properties similar to our lab’s results. It must be true.
Abe, Y., Fujita, K., & Kashimori, Y. (2018). Visual and Category Representations Shaped by the Interaction Between Inferior Temporal and Prefrontal Cortices. Cognitive Computation, 1-16.
Wutz, A., Loonis, R., Roy, J.E., Donoghue, J.A., and Miller, E.K. (2018) Different levels of category abstraction by different dynamics in different prefrontal areas. Neuron published online Jan 25 2018.
Categories can be grouped by shared sensory attributes (i.e. cats) or by a more abstract rule (i.e. animals). We explored the neural basis of abstraction by recording from multi-electrode arrays in prefrontal cortex (PFC) while monkeys performed a dot-pattern categorization task. Category abstraction was varied by the degree of exemplar distortion from the prototype pattern. Different dynamics in different PFC regions processed different levels of category abstraction. Bottom-up dynamics (stimulus-locked gamma power and spiking) in ventral PFC processed more low-level abstractions whereas top-down dynamics (beta power and beta spike-LFP coherence) in dorsal PFC processed more high-level abstractions. Our results suggest a two-stage, rhythm-based model for abstracting categories.
Coarse visuospatial categories are represented in the posterior parietal cortex whereas fine-scale discrimination are in primary visual cortex with the latter depending on feedback from the former.
Li, Y., Hu, X., Yu, Y., Zhao, K., Saalmann, Y. B., & Wang, L. (2017). Feedback from human posterior parietal cortex enables visuospatial category representations as early as primary visual cortex. Brain and Behavior.
New manuscript submitted to bioRxiv:
Neuronal rhythms orchestrate cell assembles to distinguish perceptual categories
Morteza Moazami Goudarzi, Jason Cromer, Jefferson Roy, Earl K. Miller
Categories are reflected in the spiking activity of neurons. However, how neurons form ensembles for categories is unclear. To address this, we simultaneously recorded spiking and local field potential (LFP) activity in the lateral prefrontal cortex (lPFC) of monkeys performing a delayed match to category task with two independent category sets (Animals: Cats vs Dogs; Cars: Sports Cars vs Sedans). We found stimulus and category information in alpha and beta band oscillations. Different category distinctions engaged different frequencies. There was greater spike field coherence (SFC) in alpha (~8-14 Hz) for Cats and in beta (~16-22 Hz) for Dogs. Cars showed similar differences, albeit less pronounced: greater alpha SFC for Sedans and greater beta SFC for Sports Cars. Thus, oscillatory rhythms can help coordinate neurons into different ensembles. Engagement of different frequencies may help differentiate the categories.
A recurrent network model captures the dynamics of frontal and parietal cortex activity during a categorization task. It reveals cortical motifs that underlie computations for categorical decision-making.
Chaisangmongkon, W., Swaminathan, S. K., Freedman, D. J., & Wang, X. J. (2017). Computing by Robust Transience: How the Fronto-Parietal Network Performs Sequential, Category-Based Decisions. Neuron, 93(6), 1504-1517.
Antzoulatos, E. G., & Miller, E. K. (2016). Synchronous beta rhythms of frontoparietal networks support only behaviorally relevant representations. eLife, 5, e17822.
Categorization has been associated with distributed networks of the primate brain, including the prefrontal cortex (PFC) and posterior parietal cortex (PPC). Although category-selective spiking in PFC and PPC has been established, the frequency-dependent dynamic interactions of frontoparietal networks are largely unexplored. We trained monkeys to perform a delayed-match-to-spatial-category task while recording spikes and local field potentials from the PFC and PPC with multiple electrodes. We found category-selective beta- and delta-band synchrony between and within the areas. However, in addition to the categories, delta synchrony and spiking activity also reflected irrelevant stimulus dimensions. By contrast, beta synchrony only conveyed information about the task-relevant categories. Further, category-selective PFC neurons were synchronized with PPC beta oscillations, while neurons that carried irrelevant information were not. These results suggest that long-range beta-band synchrony could act as a filter that only supports neural representations of the variables relevant to the task at hand.
Stanley, D.A., Roy, J.E., Aoi, M.C., Kopell, N.J., and Miller, E.K. (2016) Low-beta oscillations turn up the gain during category judgments. Cerebral Cortex. doi: 10.1093/cercor/bhw356 View PDF
Synchrony between local field potential (LFP) rhythms is thought to boost the signal of attended sensory inputs. Other cognitive functions could benefit from such gain control. One is categorization where decisions can be difficult if categories differ in subtle ways. Monkeys were trained to flexibly categorize smoothly varying morphed stimuli, using orthogonal boundaries to carve up the same stimulus space in 2 different ways. We found evidence for category-specific patterns of low-beta (16–20 Hz) synchrony in the lateral prefrontal cortex (PFC). This synchrony was stronger when a given category scheme was relevant. We also observed an overall increase in low-beta LFP synchrony for stimuli that were near the category boundary and thus more difficult to categorize. Beta category selectivity was evident in partial field–field coherence measurements, which measure local synchrony, but the boundary enhancement was not. Thus, it seemed that category selectivity relied on local interactions while boundary enhancement was a more global effect. The results suggest that beta synchrony helps form category ensembles and may reflect recruitment of additional cortical resources for categorizing challenging stimuli, thus serving as a form of gain control.
An excellent, comprehensive review of the neurobiology of decision-making by David Freedman and John Asaad.
Neuronal Mechanisms of Visual Categorization: An Abstract View on Decision Making
David J. Freedman and John A. Assad, Annual Review of Neuroscience, 2016
Tsutsui et al shows how the prefrontal cortex integrates rule and category information for a behavioral decision.
Tsutsui, Ken-Ichiro, et al. “Representation of Functional Category in the Monkey Prefrontal Cortex and Its Rule-Dependent Use for Behavioral Selection.” The Journal of Neuroscience 36.10 (2016): 3038-3048.
Miller Lab alumnus David Freedman and colleagues present a model that shows how categorical neural activity can develop through learning. As a result of top-down influences from decision neurons, categorical representations develop in neurons that show choice-correlated activity fluctuations. They test the model via recordings from parietal cortex.
Choice-correlated activity fluctuations underlie learning of neuronal category representation
Tatiana A. Engel, Warasinee Chaisangmongkon, David J. Freedman & Xiao-Jing Wang
Braunlich et al compared stimulus identity vs categorization tasks using fMRI in humans. They applied a Constrained Principal Components Analysis. They found evidence for two distinct frontoparietal networks. One that rapidly analyzes the stimuli and a second one that more slowly categorizes them.
Task Dependence of Visual and Category Representations in Prefrontal and Inferior Temporal Cortices
Jillian L. McKee, Maximilian Riesenhuber, Earl K. Miller, and David J. Freedman
Visual categorization is an essential perceptual and cognitive process for assigning behavioral significance to incoming stimuli. Categorization depends on sensory processing of stimulus features as well as flexible cognitive processing for classifying stimuli according to the current behavioral context. Neurophysiological studies suggest that the prefrontal cortex (PFC) and the inferior temporal cortex (ITC) are involved in visual shape categorization. However, their precise roles in the perceptual and cognitive aspects of the categorization process are unclear, as the two areas have not been directly compared during changing task contexts. To address this, we examined the impact of task relevance on categorization-related activity in PFC and ITC by recording from both areas as monkeys alternated between a shape categorization and passive viewing tasks. As monkeys viewed the same stimuli in both tasks, the impact of task relevance on encoding in each area could be compared. While both areas showed task-dependent modulations of neuronal activity, the patterns of results differed markedly. PFC, but not ITC, neurons showed a modest increase in firing rates when stimuli were task relevant. PFC also showed significantly stronger category selectivity during the task compared with passive viewing, while task-dependent modulations of category selectivity in ITC were weak and occurred with a long latency. Finally, both areas showed an enhancement of stimulus selectivity during the task compared with passive viewing. Together, this suggests that the ITC and PFC show differing degrees of task-dependent flexibility and are preferentially involved in the perceptual and cognitive aspects of the categorization process, respectively.
We (Antoulatos and Miller) show increased beta-band synchrony between (but not within) the prefrontal cortex and striatum during category learning. By the time the categories were fully learned, the beta synchrony became category-specific. That is, different patterns of prefrontal cortex-striatum recording sites showed increased beta synchrony for one category or the other. Thus, category learning may depend on formation of oscillatory synchrony-aided functional circuits between the prefrontal cortex and striatum. Further, causality analysis suggested that the striatum exerted a greater influence on the prefrontal cortex than the other way around. This supports models positing that the basal ganglia “train” the prefrontal cortex (Pasupathy and Miller, 2005; Seger and Miller, 2010).
Antzoulatos, E.G. and Miller, E.K. (2014) “Increases in functional connectivity between the prefrontal cortex and striatum during category learning.” Neuron, 83:216-225 DOI: http://dx.doi.org/10.1016/j.neuron.2014.05.005 View PDF
For further reading:
Pasupathy, A. and Miller, E.K. (2005) Different time courses for learning-related activity in the prefrontal cortex and striatum. Nature, 433:873-876. View PDF »
Antzoulatos,E.G. and Miller, E.K. (2011) Differences between neural activity in prefrontal cortex and striatum during learning of novel, abstract categories. Neuron. 71(2): 243-249. View PDF »
Seger, C.A. and Miller, E.K. (2010) Category learning in the brain. Annual Review of Neuroscience, Vol. 33: 203-219. View PDF »
Roy et al show that the activity of neurons in the prefrontal cortex (pFC) are linked to categorical decisions. Monkeys were trained to categorize a set of computer-generated images as “cats” vs “dogs”. Then, they were shown ambiguous images were centered on a category boundary, that is, they were a mix of 50% of cats and dogs and therefore had no category information. The monkeys guessed at their category membership. Activity to the same ambiguous image differed significantly, depending on the monkey’s decision about the image’s category. Thus, pFC activity reflects categorical decisions.
Van der Linden et al used computer generated images to study categorization in the human brain. They found that the frontal cortex showed sensitivity to the features diagnostic for the categories, which is consistent with results from animal studies at the neuron level.
Jack Gallant and crew used FMRI to examine scene processing in the human brain. They found that scenes activated many regions of anterior visual cortex and that the scene categories capture the co-occurrence of the objects that compose the scenes.
Max Riesenhuber and colleagues used EEG to examine the time course of shape and category signals in the human brain. Neural adaptation for category changes was seen in frontal cortex and then subsequently in temporal cortex. This supports the hypothesis that shape categories are formed by shape signals from temporal cortex that converge and form explicit category representations in frontal cortex. A late category signal in temporal cortex is consistent with category signals feeding back from frontal to temporal cortex.
A Neuron Preview for Miller Lab graduate student Simon Kornblith’s paper on a network for scene processing:
Scene Areas in Humans and Macaques by Epstein and Julian
Here’s the original post on Simon’s paper and a link to it:
A Network For Scene Processing
Pannunzi et al propose a model of visual category learning in which bottom-up sensory inputs to the inferior temporal cortex are sculpted by top-down inputs from the prefrontal cortex (PFC). The PFC improves signal to noise by enhancing the category-relevant features of the stimuli.
Miller Lab work cited:
Freedman, D.J., Riesenhuber, M., Poggio, T., and Miller, E.K. (2001) Categorical representation of visual stimuli in the primate prefrontal cortex. Science, 291:312-316. View PDF »
Freedman, D.J., Riesenhuber, M., Poggio, T., and Miller, E.K (2003) A comparison of primate prefrontal and inferior temporal cortices during visual categorization. Journal of Neuroscience, 23(12):5235-5246. View PDF »
Meyers, E.M., Freedman, D.J., Kreiman, G., Miller, E.K., and Poggio, T. (2008) Dynamic population coding of category information in the inferior temporal cortex and prefrontal cortex. Journal of Neurophysiology. 100:1407-1419. View PDF »
Muhammad, R., Wallis, J.D., and Miller, E.K. (2006) A comparison of abstract rules in the prefrontal cortex, premotor cortex, the inferior temporal cortex and the striatum. Journal of Cognitive Neuroscience, 18: 974-989. View PDF »
Seger, C.A. and Miller, E.K. (2010) Category learning in the brain. Annual Review of Neuroscience, Vol. 33: 203-219. View PDF »