RESEARCH

The brain possesses a remarkable ability to change as a result of internal and external demands. My research attempts to provide a mechanistic understanding of how the brain changes in response to visual learning and visual impairments. To address these questions, I combine advanced multimodal brain imaging, non-invasive brain stimulation, psychophysics, and computational analysis including machine learning.

I investigate the mechanisms of brain plasticity and neurodegeneration in three areas: 1) visual perceptual learning, 2) blindness, and 3) glaucoma. My work on visual learning has uncovered the key neural mechanisms that regulate plasticity in the visual cortex of healthy-sighted people. Meanwhile, my work in the clinical population has approached this topic at the broader systems level, focusing on how blindness and glaucoma affect the entire brain. These two approaches provide an integrative view of plasticity and neurodegeneration. I expand on each of these topics below.

1. Plasticity in Visual Perceptual Learning

(1) Changes in spontaneous brain activity patterns

The neural circuits involved in visual learning undergo a consolidation process, which makes the learning resilient against interference and enhances performance. This consolidation process requires following awake and/or sleep periods in order to unfold. For example, when the neural processes were disrupted by transcranial magnetic stimulation after training, learning was abolished (Bang et al., 2019, Commun Biol). Building on this, I investigated which neural mechanisms contribute to this consolidation process.

My research uncovered that reactivation of a trained visual feature during subsequent wakefulness plays a key role in the consolidation of visual learning (Figure 1; Bang et al., 2018, J Neurosci). When subjects were trained on a visual feature (Gabor orientation), the brain activity patterns representing the trained orientation were re-expressed in V1 shortly after the offset of training. Critically, the strength of reactivation measured by an increase in de-codability of the trained stimulus in V1 was associated with later performance improvement.

However, reactivating all visual information is costly because the brain is continuously bombarded with visual information during wakefulness. Thus, it is likely that the brain has developed a mechanism that determines which experience should be reactivated or ignored. Motivated by this, I examined whether reactivation occurs similarly for new vs. extensively trained visual features (Gabor orientations). Replicating my previous results, exposure to a new orientation led to reactivation. However, exposure to an extensively trained orientation led to suppression, such that brain activity patterns representing the familiar orientation were suppressed (Figure 2; Bang et al., 2021, Commun Biol). Interestingly, reactivation was strongest in V1, whereas suppression had a similar strength across V1, V2, and V3. This pattern is consistent with the notion that reactivation may be a local process in V1, whereas suppression may be a top-down process.

In addition to spontaneous brain activity changes during wakefulness, I further investigated sleep oscillations. My work demonstrated that slow spindles, a prominent oscillation during sleep, are involved in the consolidation process of visual learning (Bang et al., 2014, Vision Research).

(2) Changes in the excitatory-to-inhibitory ratio (E/I ratio)

My work shows that the E/I ratio in the early visual areas is involved in the consolidation process of visual learning (Bang et al., 2018, Nat Hum Behav; Shibata, Sasaki, Bang et al., 2017, Nat Neurosci). The E/I ratio is obtained by dividing the concentration of glutamate, a key excitatory neurotransmitter, by that of g-aminobutyric acid (GABA), a key inhibitory neurotransmitter. Specifically, the E/I ratio in the early visual areas was found to be associated with the degree of plasticity of visual learning. When the visual learning was in a vulnerable state to interference (e.g., shortly after training), the E/I ratio in the early visual areas was significantly higher than baseline (excitatory-dominant state). However, this increased vulnerability and the E/I ratio tapered off following 3.5 hours after training. These results suggest that the consolidation process, which stabilizes the learning unfolds during which the elevated E/I ratio returns to baseline. Furthermore, my work demonstrated that this consolidation process involving the E/I ratio can occur more than once, especially when the visual memory is retrieved (Bang et al., 2018, Nat Hum Behav).

Taken together, my studies on visual learning demonstrate that consolidation is a dynamic, multifaceted process, which occurs during the time window where reactivation/suppression and the sleep spindles appear, and the increased E/I ratio returns to baseline.

2. Plasticity in Blind individuals

Plasticity in the brain is differentially impacted by an individual’s age at the onset of the blindness e.g., the degree of plasticity is greater the earlier one becomes blind. However, what drives the varying degrees of plasticity remains largely unclear. One possible explanation given attributes the mechanisms for the differing levels of plasticity to cholinergic signals originating in the nucleus basalis of Meynert (NBM). This is based on the fact that NBM can modulate cortical processes such as plasticity, attention, and sensory encoding through its widespread cholinergic projections. Thus, I investigated whether early and late blind individuals present differences in the structural and functional properties of NBM. My work demonstrated that NBM presents diverging patterns of plasticity between early and late blind individuals (Bang et al., 2022, Brain Communications). Specifically, I found that NBM has reduced directionality of water diffusion in both early and late blind individuals compared to sighted individuals, which may arise from a higher proportion of crossing fibers or reduced axonal maturation. However, counterintuitively, NBM presented enhanced functional connectivity at both global and network (visual, language, and default-mode networks) levels in early blind individuals, but not in late blind individuals during rest. This observation suggests that NBM develops a stronger cholinergic influence on the neocortex of early blind individuals, which may explain discrepancies in the plasticity across early- and late-onset blindness.

3. Neurodegeneration in Glaucoma patients

(1) Neurochemical changes

Glaucoma is an age-related neurodegenerative disease of the visual system, affecting both the eye and the brain. Recent studies demonstrated that glaucoma involves accumulation of amyloid b and tau in the visual pathway, which can impair glutamatergic and GABAergic systems. In line with this, my work demonstrated that both GABA and glutamate levels in the visual cortex decrease with increasing severity of glaucoma regardless of age (Bang et al., 2022, bioRxiv). Further, the reduction of GABA but not glutamate predicted the neural specificity, which is the degree to which neural representations of different stimuli can be distinguished and is thought to underlie efficient sensory and cognitive functions. This association was independent of the impairments on the retina structure and age. Our results suggest that glaucoma-specific decline of GABA undermines neural specificity in the visual cortex and that targeting GABA could improve the neural specificity in glaucoma.

(2) Changes in the cerebrospinal fluid (CSF)

Retinal ganglion cell death in glaucoma was previously thought to be caused by elevated intraocular pressure. However, recent studies suggested that a pressure imbalance between aqueous humor and CSF may play a critical role in the development of glaucoma. My work demonstrated that early stage of glaucoma involves elevated power of CSF and enhanced coupling between CSF and global brain activities, which is potentially the driving force of the CSF (Bang et al., 2022, ARVO abstract). This finding raises a possibility that enhanced CSF may be involved in glaucomatous degeneration.

(3) Changes in the sleep-regulating systems

Glaucoma patients have a higher prevalence of sleep disorders, suggesting that glaucoma may share a common mechanism with sleep disorders. Critically, the ventrolateral preoptic nucleus (VLPO), a major sleep-inducing area located in the anterior hypothalamus, receives input from photosensitive retinal ganglion cells (ipRGCs), which are known to be damaged in glaucoma. Thus, it is reasonable to suggest that the loss of ipRGCs may result in impaired function of VLPO in glaucoma. My work found that the functional connectivity between VLPO and the subcortical arousal systems/cortical areas are impaired in glaucoma and that this alteration is associated with the retinal nerve fiber layer thinning (Bang et al., 2021, ARVO abstract; funded by BrightFocus Foundation).