Neural Mechanisms of Visual Field Recovery after Perceptual Training in Cortical Blindness

Extensive training can improve performance on almost every visual task, through a process called visual perceptual learning (He et al., 2021). Visual perceptual learning has been applied to rehabilitate impaired vision for patients with low vision (He et al., 2021). In addition, visual perceptual learning can partially restore the impaired visual field in patients with cortical blindness: partial or complete loss of vision in one-half of the visual field following damage to the postchiasmatic visual pathway, including the primary visual cortex (V1, also known as striate cortex) (Melnick et al., 2016).

Two hypotheses, which are not mutually exclusive, have been proposed regarding the neural mechanisms by which perceptual training restores visual field in cortical blindness. First is the spared-V1 hypothesis. It has been demonstrated that not all neurons in V1 are destroyed by cortical damage: some residual neurons survive cortical damage and form spared islands in perilesional V1 or intact tissues of V1 adjacent to lesioned areas (Wessinger et al., 1997). The spared V1 still shows neural responses to visual stimuli presented in the perimetrically blind visual fields (Ajina and Bridge, 2018). For instance, the functional properties of the damaged visual cortex, like retinotopic organization, can be estimated by the population receptive field technique (Papanikolaou et al., 2014; Barbot et al., 2021). Since the neural responses of these residual neurons in V1 are not strong enough to elicit conscious visual percept, patients do not show conscious responses to visual stimuli presented in the perimetrically blind visual fields. Although the residual neurons in spared islands are functionally inactive, they could be stimulated and gradually reactivated by repetitive visual stimulation during extensive perceptual training, leading to functional reorganization (Huxlin et al., 2009; Das and Huxlin, 2010; Papanikolaou et al., 2014). This reorganization might allow spared V1 to encode visual sensory information more efficiently after perceptual training (Barbot et al., 2021; Fahrenthold et al., 2021). The second hypothesis regarding how visual perceptual training restores visual field is the subcortical pathways hypothesis. This hypothesis proposes that the subcortical pathways recruited during early visual processing are reorganized by perceptual training and thereby subserve vision rehabilitation for cortical blindness. Although lesions of V1 lead to the loss of conscious vision, people with cortical blindness still have considerable residual perception when visual stimuli are presented in the perimetrically blind fields (Ajina et al., 2015), a phenomenon called blindsight. The residual vision is generally attributed to projections from subcortical areas, such as LGN, superior colliculus (SC), and pulvinar. Those pathways bypass V1 and directly innervate higher visual areas, including the human middle-temporal complex area (hMT⁺, also known as V5) (Sincich et al., 2004). Visual area hMT⁺/V5 is best known for its role in perceiving moving stimuli, and it has been hypothesized to be involved in residual motion perception in blindsight (Ajina and Bridge, 2018).

To better understand the neural substrates that underlie vision restoration via perceptual training in patients with cortical blindness, Ajina et al. (2021) conducted a clinic-oriented study in which patients with hemianopia caused by stroke damage to V1 were included. These patients were trained to perform a contrast detection task involving drifting Gabor patches in each daily session for 3-6 months (Ajina et al., 2021). Notably, the large stimuli with low spatial frequency and high temporal frequency adopted here can elicit blindsight. Before and after training, behavioral performance on both trained and untrained psychophysical tasks and visual field were measured; importantly, to understand the neural substrates underpinning visual field amelioration induced by visual perceptual learning, fMRI and structural MRI data were also collected.

Consistent with previous studies (Huxlin et al., 2009; Saionz et al., 2020), Ajina et al. (2021) found that, after extensive training on the contrast-detection task, performances on both trained and untrained tasks (e.g., motion detection and motion direction discrimination) were improved in cortical blindness patients, demonstrating a prominent generalization of training effect. More importantly, Ajina et al. (2021) found significant increases in visual sensitivity of the trained region of the visual field gauged by perimetry after training, demonstrating that visual field recovery occurred. The visual field restoration, however, was limited to the trained location. The improvements in visual field sensitivity were positively correlated with the performance enhancements in the trained contrast detection task in the trained region. Together, the results of psychophysical and clinical measurements demonstrated that visual perceptual learning has the potential to serve as a restorative therapy for cortical blind patients.

Concerning the fMRI results, Ajina et al. (2021) found that training increased BOLD responses to the contrast stimuli in hMT⁺/V5 of the lesioned hemisphere, although subjects with damage to V1 did not make conscious responses to the test stimuli. This was confirmed by both whole-brain and ROI analyses. Although training increased BOLD responses in the hMT⁺/V5, these changes did not correlate with changes in performance of psychophysical tasks and visual field restoration. Notably, however, the baseline BOLD signals in hMT⁺/V5 before training were highly correlated with the amount of visual field recovery, suggesting that the extent of visual field recovery gained from perceptual training can be predicted by the pretraining activity of motion direction-sensitive visual areas (i.e., hMT⁺/V5). Importantly, extensive contrast-detection training did not change BOLD responses in the early visual cortical areas (e.g., V1), although 2 patients seemed to have spared V1.

Ajina et al. (2021) found that contrast detection training increased BOLD responses in hMT⁺/V5, but not the early visual cortex, and revealed a direct link between brain activity in hMT⁺/V5 and visual field restoration via visual perceptual training. Remarkably, previous studies have shown that neural activity in subcortical nuclei, such as LGN (Yu et al., 2016) and SC (Vaina et al., 1998), can be changed by perceptual training in healthy adults as well. Specifically, in healthy adults, researchers found that BOLD signals in the magnocellular layers of LGN increased after contrast detection training (Yu et al., 2016), whereas the extent of activation of SC changed with motion direction discrimination training (Vaina et al., 1998). It is possible that the increases in BOLD signals in hMT⁺/V5 for patients with V1 damage were driven by the strengthened LGN-hMT+/V5 or SC-hMT⁺/V5 projections, as these patients preserved LGN and SC. Thus, the neuroimaging result better supports the subcortical pathway hypothesis. Namely, the visual field recovery by perceptual training possibly results from strengthened connections between the subcortical nucleus and the extrastriate cortex. This might provide a therapeutic guide for visual field rehabilitation in cortical blindness.

Although the activity in hMT⁺/V5 increased after perceptual training, the present study provided limited information about which subcortical pathways are involved in visual field restoration through visual perceptual training. In cortical blindness, especially in people with blindsight, visual signals from the retina can directly reach the extrastriate cortex through multiple subcortical pathways, although V1 is damaged (Cowey, 2004). Among these subcortical pathways, two of the densest projections are from the dorsal LGN and the SC-pulvinar to the extrastriate cortex to area hMT⁺/V5 (Cowey, 2004; Urbanski et al., 2014; Kinoshita et al., 2019). However, it is still unclear which pathways are engaged in vision restoration after perceptual training for cortical blindness patients. To better understand this issue, in the future, comprehensive functional and structural neuroimaging techniques and multivariate analysis methods should be applied to a large cohort of patients with cortical blindness over the whole course of training (Yotsumoto et al., 2008; Yu et al., 2016; Ajina and Bridge, 2018). Such a prospective longitudinal design with multimodal neuroimaging could help us better understand the neural substrates, and particularly, to identify the subcortical pathways, that underpin vision recovery by visual perceptual training for cortical blindness (Melnick et al., 2016).