Optic Tract Shrinkage Limits Visual Restoration After Occipital Stroke

Supplemental Digital Content is available in the text.

One factor likely to affect the potential to recover vision after V1 damage is the state of the residual visual system. Without intervention, the perimetrically defined visual deficit of patients >6 months poststroke slowly expands and worsens over time 10,18,21 -a phenomenon attributed to retrograde degeneration of early visual pathways. Studies in humans and nonhuman primates have long shown the existence of trans-synaptic retrograde degeneration in the dorsal lateral geniculate nucleus and retina after occipital lesions. [22][23][24][25][26][27][28][29] This is thought to occur in sequence, with dorsal lateral geniculate nucleus relay cells dying first, followed by loss of retinal ganglion cells. In macaque monkeys, the magnitude of trans-synaptic retrograde degeneration varies, appearing to correlate with the size of, as well as time since the V1 lesion. 25,30 In humans, structural magnetic resonance imaging (MRI) analyses have shown that the optic tract (OT) ipsilateral to occipital cortex damage is often reduced in size, 23,27,28,[31][32][33] as are the thicknesses of the retinal ganglion cells and nerve fiber layers over corresponding regions of the retina in each eye. 28,34 Importantly, these human studies also showed considerable interindividual variability in the severity of degeneration. We now posit that since early visual neurons are responsible for conveying the bulk of sensory input to the rest of the visual system, their loss could critically limit the ability of patients with V1 damage to recover visual functions. This was recently shown to be true with respect to spontaneous recovery, which occurs with marked individual variability during the subacute period, <6 months poststroke. 34 However, to what extent retrograde degeneration limits vision restoration induced by training interventions in the chronic (>6 months) poststroke period remains to be determined and is the primary question addressed with the present study.

Participants and Study Design
The present analyses were performed on magnetic resonance imaging data from 36 CB patients ( For each patient, we documented the nature of the stroke (ischemic or hemorrhagic), and the presence of several riskfactors, including diabetes, hypertension and migraine. Finally, we excluded patients with ocular disease (other than refractive error), neglect and neurological disease unrelated to the occipital stroke. Humphrey visual field (HVF) automated perimetry was collected in all Rochester patients save one, in one Oxford patient, and was not collected in controls.
The training portion of the study was performed in a subset of chronic patients from Rochester only. Changes in HVF performance were used as the outcome measure (see Data Supplement for details of restoration training and HVF analysis protocols), which was correlated with MRI-derived OT metrics (detailed in the Data Supplement). Eligibility for the training arm of the study required that chronic CB participants meet additional criteria: (1) reliable 24-2 and 10-2 Humphrey visual field perimetry tests, with reliability defined as <20% fixation losses, false-positive and false-negative errors in either eye, (2) fixation precision better than ±1 degree relative to a fixation spot during psychophysical testing, measured using an Eyelink-1000 eye tracker (SR Research, Canada), and (3) ability to perform months of daily, in-home training, followed by a return laboratory visit in which performance on the home-training tasks was verified with fixation enforced using the Eyelink-1000 eye tracker. Only 14 of the original chronic participants met these criteria. During their return visit to the Rochester laboratory, 3 of the 14 trained patients declined (claustrophobia) or failed to meet eligibility criteria for a repeat MRI (incompatible device implantation during the training period). Thus, while we could measure training-induced changes in all 14 patients using psychophysics and HVF tests, we were unable to obtain a posttraining measure of OT laterality in these 3 patients.

MRI Procedures
For University of Rochester participants, MRI data were acquired on a 3T Magnetom Trio scanner (Siemens, Erlangen, Germany) using a 32-channel head coil. High-resolution T1-weighted anatomic volumes (MP-RAGE, magnetization prepared rapid gradient echo) were collected with a voxel size of 1 mm 3 . For University of Oxford participants, MRI data were acquired with a 3T Prisma scanner (Siemens, Erlangen, Germany) at the Oxford Centre for Functional MRI of the Brain (FMRIB). Whole-head, structural, T1-weighted scans were collected axially at a resolution of 1 mm 3 (MP-RAGE; TR: 1900 ms; echo time: 3.97 ms; flip angle: 8°). The OT volume analysis was adapted from prior published work. 23,32 To account for differences in head orientation in the scanner, FMRIB software library (FSL) image analysis software (http://www.fmrib.ox.ac.uk/fsl) was used to reorient the OTs parallel to the anterior-posterior axis (y axis) in standard space (1 mm), and the scans were resampled parallel and perpendicular to the OTs. Details on methodology to quantify OT shrinkage are provided in the Data Supplement, and illustrated in Figure I in the Data Supplement.

Statistical Analyses
Standard parametric tests (ie, repeated-measures ANOVAs, paired t-tests) were used to assess within-subject differences. For independent sample comparisons, unpaired t-tests were used when contrasting 2 groups; ANOVAs were used to contrast 3 or more groups. Partial eta-squared and Cohen d values were calculated to assess effect size for ANOVAs and t-tests, respectively. Linear regressions were used to model the relationship between explanatory variables and dependent outcomes, with r values and 95% confidence intervals (CI 95 ) for rho provided, and significance estimated using a t test.

OT Laterality in CB Patients
We first contrasted OT volumes in CB patients and visually intact controls, computing a laterality index (LI) for each person ( Figure 1A). Control and subacute CB participants had an LI ≈0 for all OT pixel intensity thresholds ( Figure 1A, gray, and open red symbols, respectively). In contrast, chronic CB participants LI values ( Figure 1A, filled red symbols) became increasingly positive at higher pixel intensities. More positive LI values denoted a higher number of brighter voxels in the contralesional relative to the ipsilesional OTs. As a group, the LI max (LI value for maximal OT voxel brightness) of chronic CBs was greater than in subacute CBs (t 34 =−5.27, P<0.0001, CI 95 =±0.073, ±0.149) or in visually intact controls (t 32 =−4.6, P<0.0001, CI 95 =±0.059, ±0.149). Importantly, this was not explained by chronic patients having larger visual defect sizes (area of 24-2 HVF with a pattern deviation <−5 dB, t 30 =0.47, P=0.641, CI 95 =±89, ±127), or severity, measured by perimetric mean deviation (PMD) averaged between the 2 eyes (t 30 =1.15, P=0.259, CI 95 =±1.9, ±2.6).
When considering all CB patients (subacute and chronic), time since stroke and OT laterality were positively correlated ( Figure 1B: r=0.3566, CI 95 for rho=0.032 to 0.613, t 42 = 2.23, P=0.032). LI max values in subacute CB patients were close to zero (open symbols, mean±SEM=−0.016±0.08, Figure 1B), resembling control participants (LI max =0.043±0.02). In contrast, chronic patients had some of the highest, most positive LI max values (eg, right-most solid red symbols in Figure 1B), albeit with a large range of LI max values evidenced as early as 7 months poststroke.
To examine factors contributing to the heterogeneity of OT LI max among chronic patients, we separated them into 2 groups: those with LI max >0.256 were termed lateralized ( Figure 1C); those with LI max <0.256 nonlateralized ( Figure 1D). LI max of 0.256 was chosen because it is 2 SDs greater than the LI max of controls. Using this classification, mean LI max in lateralized chron-ics=0.66±0.06 (SEM), while mean LI max in nonlateralized chronics=0.01±0.04.

Chronic CB: OT Laterality Correlates With Deficit Severity and Size
In chronic patients, LI max was no longer correlated with time since stroke ( Figure 2A Finally, since a large volume of V1 is devoted to the representation of the central 10° vision, [35][36][37][38][39] we asked if those participants whose deficit impacted this central region were more likely to be lateralized. As shown in Figure III Figure IV in the Data Supplement). Overall, our data suggest that independent of initial deficit size, and given extensive visual training in the blind field, initial structural integrity of the OTs at training onset may impact training efficacy in chronic CB patients. Given baseline variability in LI max , we posit that baseline LI max is a key factor driving individual variability in training outcomes for this patient population.

Chronic CB: Visual Training Does Not Appear to Change OT Laterality
Our final question was whether visual training in the chronic period altered retrograde degeneration at the level of the OTs. Our laboratory has conducted a range of training   studies where measurable improvements in visual perception were elicited in chronic CB patients. 10,14,40 Here, we analyzed the subset of 11 chronic CB patients who underwent ≈1 year of visual training, together with pre-and posttraining MRI. All 11 exhibited improvements that could be verified using fixation-enforced psychophysical testing and HVF perimetry; 7 of them were lateralized at baseline (ie, before training onset), and 4 were nonlateralized. Visual discrimination training in the blind field induced improvement ≥6 dB across an area of the HVF 101±24 deg 2 (SEM), which translated to a 1.3±0.3 dB increase in the binocular PMD. Consistent with earlier observations that pretraining LI max was negatively correlated with HVF improvement, nonlateralized chronic patients exhibited a larger area of HVF improvement than lateralized patients ( Table 2), although both groups trained for a similar number of sessions (t 9 =−1.2, P=0.26, CI 95 =±174, ±545). PMD change, which reflects vision averaged across the entire visual field, was not sufficiently sensitive to detect this blind-field specific improvement with our present, small sample size ( Table 2).
To assess the effect of visual training on OT laterality, we analyzed LI 85 rather than LI max , accounting for the fact that not every patient had voxel intensities >85% of maximum at both time-points. We found no significant difference between pre-and post-training LI 85 across the 11 patients imaged ( Figure 4A

DISCUSSION
The present study builds upon previous findings about retrograde degeneration following damage to the adult, occipital cortex, and the fundamental observation that this degeneration follows a distinct time course and pattern. 22,23,25,28,32 We not only show strong evidence of unilateral shrinkage of the OT ipsilateral to the damaged V1 in a large cohort of chronic CB patients but also highlight the lack of measurable OT shrinkage in subacute patients.
Differences in deficit size were ruled out as a possible explanation, confirming the notion 28,41 that trans-synaptic, retrograde degeneration must require ≈6 months post-V1 stroke to be evidenced at the level of the human OTs.

Factors Driving Heterogeneity of Chronic CB Patients
A clear observation in the present study was that chronic CB patients are not a uniform population in terms of OT degeneration, with 60% showing significant OT lateralization, but 39% having LI max similar to visually intact controls (and subacute CB patients). Whether preservation of the OT -sometimes years after a stroke-in a proportion of chronic CB patients truly signifies decreased or absent retrograde degeneration remains to be determined. However, since we found no mention of this dichotomy in prior literature, 22,23,25,28,32,33,42,43 we began to interrogate its possible causes. We ruled out time poststroke as a significant predictor of chronic LI max . Only one of the chronic patients in our entire sample had documented diabetes (Type 2) and they were nonlateralized. Additionally, 10/17 (59%) of the lateralized chronic patients and 7/11 (64%) of the nonlateralized chronic patients had hypertension. The incidence of migraine was similarly small across groups: 4/17 (24%) in lateralized patients and 3/11 (27%) in nonlateralized participants. All in all, in our sample, diabetes, hypertension, and migraine did not seem to be significant predictors, nor did they correlate with shrinkage of the ipsilesional OT after occipital stroke. Instead, chronic patients with larger, more severe visual deficits (measured using HVF and Goldmann perimetry) had larger LI max . Thus, greater damage to V1 caused increased degeneration of the ipsilesional OT, with deficit (and likely lesion) size accounting for much of the variance in chronic LI max . However, several other factors, which could have contributed to individual differences in retrograde degeneration after V1 damage, could not be evaluated here. We can only speculate that the relative amount of damage sustained to gray and white matter, the exact nature of treatments received in the clinic (eg, tPA), as well as genetic and epi-genetic influences on cellular repair and inflammatory responses could all influence the rate of degeneration and plastic potential of early visual pathways after occipital strokes. Nonetheless, a key question emerging was: to what extent did the degeneration sustained impact the amount of recovery these individuals could achieve with restorative interventions.

Pretraining LI Predicts Training Efficacy in Chronic CB
A recent study 34 reported a correlation between preserved, early visual circuitry (retinal ganglion cell layer thickness and stimulus-based early visual cortex activity) and the amount and location of spontaneous visual improvements during the subacute, poststroke period. However, over the last few decades, a major focus in the field has been to investigate visual training approaches to restore vision in chronic CB patients. These patients are of interest because they no longer exhibit spontaneous recovery, 4 but intensive visual training can recover a range of visual abilities in the blind-field of some (but not all) these individuals (reviewed in study by Melnick et al, 15 Matteo et al, 44 Perez et al, 45 and Pouget et al 46 ). Such variability in the extent of recovery was clearly evidenced in the present cohort: given comparable amounts of daily training over a long period of time, chronic CB patients with nonlateralized OTs (LI max close to 0) at the onset of training attained significantly larger HVF improvements than lateralized, chronic CB patients. Thus, baseline pretraining LI max emerges as a biomarker able to individually predict the efficacy of visual training for recovering vision in the chronic period poststroke.
Several factors have been postulated as key for training-induced visual field recovery in chronic CB, including the preservation and re-activation by training of perilesional V1 cortex weakened, but not destroyed, by the stroke. 47 The present findings support this hypothesis, showing that patients with the greatest preservation of retino-geniculostriate circuits sustain the greatest perceptual benefit from visual training. This is not to say that chronic patients with greater degeneration of early circuits cannot benefit from restoration training-they can, but the amount of benefit attained for the amount of effort expended is less. Perhaps these patients' visual recovery relies to a larger extent on extra-geniculo-striate pathways that convey information to extrastriate visual cortical areas for processing. [48][49][50][51][52][53][54][55][56] Our findings would suggest that recruitment of these alternative pathways by training may be less efficient for recovering conscious vision than recruitment of residual V1.

Visual Training May Not Change OT Laterality in Chronic CB
Finally, we asked if visual training administered in the chronic poststroke period changes LI. In patients with lateralized OTs, an interesting outcome would be to see training decrease LI, suggesting that training can reverse structural and/or functional processes associated with retrograde degeneration. For patients with nonlateralized OTs, an equally interesting outcome would be to see training prevent them from becoming lateralized, presumably by blocking further degeneration. All chronic patients who were trained attained some measure of visual recovery inside their blind field-even those who only began training a year or more after their occipital stroke (see also study by Huxlin and Cavanaugh, 10 Cavanaugh et al, 11 Das et al, 13 Huxlin et al, 14 Saionz et al 18 ). However, we saw no overall change in LI 85 as a function of training. Due to our reduced sample size, this result was clearest in lateralized patients (n=7), where visual training neither decreased nor increased OT laterality. Among the 4 nonlateralized patients, 1 converted to lateralized (suggesting that training did not prevent degeneration from progressing), but the remaining 3 did not. This is an extraordinarily difficult experiment to carry out: patients need to meet MRI eligibility twice-≈1year apart on average-in addition to performing rigorous, correctly executed visual training in their blind-field daily for ≈1 year on average. Nonetheless, additional studies using greater sample sizes are needed to sufficiently power any claims about the effect of training on the OTs in chronic patients.

Implications for Future Work
The present findings illustrate some important structural limitations of restorative plasticity in adult, chronic stroke patients. Follow-up studies that include larger and more homogenous groups of chronic patients are needed to verify the functional implications of these results. Additionally, from a rehabilitation perspective, OT shrinkage could be measured and used to identify chronic patients who may benefit from different training strategies.
However, our results also highlight potential opportunities offered by intervening in the subacute rather than chronic poststroke period. Could training administered early as opposed to late after stroke, stop or slow the progress of retrograde degeneration? This makes intuitive sense, as retrograde degeneration in humans is well underway and clearly evidenced by structural changes relatively early in the chronic period. 25,28,34,57 Intervening in the subacute period may catch the visual system in a state where degeneration is not yet complete. Therefore, we posit that future studies should investigate if early intervention in acute and sub-acute CB populations can slow the rate or magnitude of degeneration, keeping LI max closer to zero for longer and thus, further improving the potential for training induced visual recovery once participants reach the chronic poststroke period. Even if the progression of degeneration cannot be stemmed, an intervention during the subacute period would take advantage of the greater integrity of retinal ganglion cells and the white matter tracts (including the OT) conveying information to subcortical centers to attain larger training-induced visual improvements. We recently verified this hypothesis, showing that visual training in subacute CB elicits more rapid and spatially extensive recovery than identical training in the chronic period. 18 However, the possibility also remains that interventions started before (rather than after) retrograde degeneration is fully evidenced, may act to slow or stop its progress. If true, this would reinforce the notion that time is indeed vision after occipital strokes. More importantly, early intervention could create a larger pool of chronic patients with increased potential for further, traininginduced recovery.