Exogenous Spatial Attention Helps Overcome Spatial Specificity of Visual Learning in the Blind Field After V1 Damage

Abstract

Visual discrimination training can restore visual functions in the blind field of participants with stroke-induced V1 damage. However, single-stimulus training in this population is limited by spatial specificity. Thus, it requires iterative training over several months to achieve improvement at more than one blind-field location, particularly at sites further from the blind field border (i.e., deeper in the blind field). With neurotypical observers, exogenous spatial attention (SA) facilitates transfer of learning to untrained locations. Here, we asked if SA pre-cues could induce transfer of training deeper into cortically blinded (CB) fields. Twenty CB participants trained on a global motion discrimination task either using a single [primary] stimulus and no cues (Task 1), a single primary stimulus with a large pre-cue deep in the blind field (Task 2), two identical stimuli (primary and deep) with small pre-cues just above them (Task 3), or a single stimulus randomly alternating at a primary and deep blind-field location, forewarned by a small pre-cue above them on each trial (Task 4). Training on Task 1 induced reliable improvements at the primary location, but no transfer of learning deeper in the blind field. The addition of SA pre-cues in Tasks 2-4 induced transfer in more than half the participants, although threshold improvements at primary locations were smaller than for Task 1. We conclude that directing exogenous SA deep in the blind field attracts attention automatically in CB patients and facilitates transfer of learning towards cued locations, even without V1 processing for those regions of space.

Keywords: Hemianopia; Motion perception; Perceptual learning; Stroke; Transfer; Vision restoration.

Fig. 1

Participant brain scans, visual fields, and training locations. Presence of stroke-induced occipital damage was confirmed via brain imaging. T1 images are presented for CB01–08, CB12, CB13, CB14, and CB20. Flair images are presented for CB09, CB11, CB16, CB18, and CB19. Diffusion-weighted images (DWI) are presented for CB10 and CB17. A computed tomography (CT) image is used for CB15. Radiological conventions are followed, with right side of the brain presented on image left. Composite binocular maps of Humphrey visual fields collected at the baseline visit are shown below each brain image. Areas of light grey indicate relatively intact vision, while black indicates relative blindness. Grey circles on each composite map denote the location and size of the primary training location (nearest to vertical blind field border), subsequent locations of home training once performance improved to criterion at the primary location, and the deep (furthest from border) training location (in Tasks 3 and 4). CB06’s visual field is enlarged, with shaded circles corresponding to sequential training locations to illustrate the process of moving training deeper at the primary location as home performance improves (illustrated in adjacent plot). In Task 2, the deep training location is indicated with a dotted circle and a grey arrow to represent the fact that a random dot stimulus was never presented there during training. CB01–07 trained on Task 1, and thus only have primary training locations. CB08–12 trained on Task 2, CB10 and CB13–16 trained on Task 3, and CB17–20 trained on Task 4. CB10 trained on two different SA tasks (Tasks 2 and 3) in two different quadrants of the visual field

Fig. 2

Task designs. A Task 1: after fixation for 1000 ms, a stimulus appeared at the primary training location, consisting of a group of dots moving randomly left or right for 500 ms, with participants asked to indicate the direction of motion perceived, with no attentional manipulation. B Task 2: after fixation, participants were first presented with a large, 5 deg diameter flashing disc that served as a spatial attention pre-cue to a region of space ~ 7 deg deeper in the blind field than the random dot stimulus presented at the primary training location, near the blind field border. Primary training stimulus parameters were identical to those in Task 1. C Task 3: after fixation, participants were presented with an SA pre-cue consisting of two small, 2 deg diameter, flashing white discs ~ 7 deg apart, just above two stimuli consisting of identically moving dots—one at the primary and one at the deep training locations. Participants were asked to discriminate the perceived direction of motion in either stimulus. D Task 4: pre-cues and stimulus parameters were identical to those in Task 3, except that only one valid pre-cue and one stimulus were presented per trial, alternating between primary and deep locations, with participants indicating the direction of motion in the stimulus only at the cued location, on each trial. E Mapping procedure: regardless of training task, all participants were mapped in lab using Task 1 to select ideal training locations. Initial stimuli were placed just inside the blind field border (grey circle), and then moved laterally deeper into the blind field by 1 deg increments until performance dropped from above chance (green circle) to chance (red circles). The first location with chance performance (dotted red circle) was selected as the initial training location

Fig. 3

Sample home training performance from each training group. In general, training locations (primary and deep) were initially unable to generate a normalized direction range (NDR) threshold less than 100%. At primary training locations, performance gradually improved with increasing session number, leading to a reduction in NDR thresholds. This was generally not observed for the majority of deep locations (as shown for CB19). The exact number of sessions required for thresholds to improve at each location varied somewhat by patient and training task (see complete home training data set in Figshare), but with few exceptions, the total number of sessions trained at home was comparable across training cohorts (see Table 1)

Fig. 4

Comparison of pre- and post-training performance. A Across all cohorts, prior to the onset of training, the initial (nearest vertical blind field border) primary locations’ NDR thresholds were very high (grey columns). Following home training, performance at these primary locations improved across all cohorts (black columns), and at the group level, they reached performance similar to that in the intact field (white columns). B Plot of performance at deep blind-field locations (see Fig. 1) using the same conventions as in A. Data show profound impairment pre-training, denoted by high NDR thresholds in all training cohorts. Despite three people who improved (green dots), training with SA pre-cues generated no group-level improvement at deep locations, remaining impaired relative to the intact field. Data are not shown for participants trained on Task 1, who were neither trained, pre-cued, nor tested at such locations. C Plot of distance of primary and deep blind-field locations from the nearest blind field border (in any direction) computed across all groups, showing that primary training locations were significantly closer to intact vision than deep locations. Importantly, the three deep locations that improved following training with SA pre-cues (green circles) were not systematically closer to intact vision than unimproved locations. White/green dots = individual participants’ data. Error bars = standard deviations. * = p < 0.05. ns, not significant; N/A, not applicable

Fig. 5

Trained and transfer locations. A Humphrey composite maps of the central visual field of all participants showing locations not trained but which exhibited transfer of learning during post-tests (yellow circles). Locations directly trained with random dot stimuli (solid grey circles—the same as in Fig. 1) or large SA pre-cues (arrowed dotted grey circles in Task 2) are also shown. Dotted yellow circles for CB09 and CB10 in Task 2 indicate improvement at the deep, cued, but untrained location, which is thus considered a transfer location. B Plot of NDR thresholds at untrained, transfer locations (yellow in A) versus distance from nearest training location across all SA cohorts (Tasks 2–4). Linear regression analysis shows significant worsening of “transfer” thresholds as distance increases from the primary training location