Visual Rehabilitation in Chronic Cerebral Blindness: A Randomized Controlled Crossover Study

Abstract

The treatment of patients suffering from cerebral blindness following stroke is a topic of much recent interest. Several types of treatment are under investigation, such as substitution with prisms and compensation training of saccades. A third approach, aimed at vision restitution is controversial, as a proper controlled study design is missing. In the current study, 27 chronic stroke patients with homonymous visual field defects were trained at home with a visual training device. We used a discrimination task for two types of stimuli: a static point stimulus and a new optic flow-discontinuity stimulus. Using a randomized controlled crossover design, each patient received two successive training rounds, one with high contrast stimuli in their affected hemifield (test) and one round with low-contrast stimuli in their intact hemifield (control). Goldmann and Humphrey perimetry were performed at the start of the study and following each training round. In addition, reading performance was measured. Goldmann perimetry revealed a statistically significant reduction of the visual field defect after the test training, but not after the control training or after no intervention. For both training rounds combined, Humphrey perimetry revealed that the effect of a directed training (sensitivity change in trained hemifield) exceeded that of an undirected training (sensitivity change in untrained hemifield). The interaction between trained and tested hemifield was just above the threshold of significance (p = 0.058). Interestingly, reduction of the field defect assessed by Goldmann perimetry increases with the difference between defect size as measured by Humphrey and Goldmann perimetry prior to training. Moreover, improvement of visual sensitivity measured by Humphrey perimetry increases with the fraction of non-responsive elements (i.e., more relative field loss) in Humphrey perimetry prior to training. Reading speed revealed a significant improvement after training. Our findings demonstrate that our training can result in reduction of the visual field. Improved reading performance after defect training further supports the significance of our training for improvement in daily life activities.

Keywords: perimetry; rehabilitation; stroke; training; vision.

Figure 1

Study design and training stimuli. (A) Perimetry was performed at different time points. A-intake = effect no training; B-A = effect training 1; and C-B = effect training 2. (B) Sequence of screenshots for static point and optic flow discrimination task during two trials. Each trial starts with a single fixation point (2 s), followed by the stimulus (7 s). The dashed circle represent the black disk (itself invisible against the black background) on which a white optic flow pattern rotated clockwise or counterclockwise.

Figure 2

Introduction of a new method to describe visual field changes in terms of equivalent cortical surface gain (ECSG) based on the cortical magnification factor. (A) mm ECSG for 2° of field increase across the entire hemifield, starting at a pre-training defect border of 2°, 5°, 10°, 15°, 20°, and 25° eccentricity. The corresponding area of a 2° field increase (red) is shown for a defect border starting at 2° and 25°, respectively. (B) Goldmann perimetry during intake, prior to training (measurement A), after 40 h of training the affected hemifield (measurement B) and after 40 h training the intact hemifield (measurement C) in one subject (J22). The graphs below show the visual field change, when no intervention was done (A-Intake), for defect training (B-A) and for intact training (C-B). Red represent field increase and blue represent field decrease. The corresponding ECSG values are 1.27, 9.12, and −1.30, respectively. Note that, this patient with right occipital lesion shows a visual field defect crossing the vertical midline. This is reproduced in all four Goldmann measurements and also visible in all three Humphrey measurements.

Figure 3

Results of Goldmann perimetry following each training round. Defect reduction measured with Goldmann perimetry (ECSG: mean ± SEM for 25 patients) following no intervention, defect training, and intact training. The effects of the training rounds were assessed with respect to the preceding perimetry measurement. Note that, patient J18 and J23 were excluded for this analysis.

Figure 4

Results of Humphrey perimetry following each training round. (A) Mean sensitivity change in dB (± SEM) per hemifield. (B) Mean sensitivity change in dB (± SEM) for directed and undirected training combined for both hemifields. (C) Sensitivity increase measured with Humphrey perimetry (dB) after directed defect training as a function of number of non-responsive elements prior to training (A measurement). Note that patients J09, J10, J13, J14, and J22 were excluded for the Humphrey hemifield analysis (A), because this analysis requires three measurements with adequate fixation. For J14 and J22, data were included for the defect training round analyses (B,C), because their pre- and post-defect training perimetry was reliable.

Figure 5

Method to compare shifts in the visual field border based on Goldmann and Humphrey perimetry (A) Definitions of the border of the defect (G, H₀, H₁, and H_mean), using the data of one subject (J13). (B) Border difference between Goldmann (G) and Humphrey perimetry prior to training (H₀ or H₁ or H_mean). Mean ± SD for all patients. Prior to training H_mean defect border corresponds best to the Goldmann border, as it does not deviate from 0 in contrast to H_0−G (p = 0.012) and H_1−G (p = 0.000).

Figure 6

Indication for recovery potential based on the difference between Goldmann and Humphrey perimetry prior to training. (A) Regression between defect reduction measured with Goldmann perimetry (ECSG) after defect training and the difference between the Humphrey and Goldmann perimetry prior to the training (A measurement: border of G – border of H_mean expressed in ECSG). Clearly, the defect reduction was linearly related to the border difference between Humphrey and Goldmann perimetry prior to training. Patients’ border shift by Goldmann perimetry equals 3.9405 + 0.436 ECSG_Hmean−G. (B) The same analysis for the sensitivity increase (dB) in the affected hemifield by defect training.

Figure 7

Reading performance following each training round. (A) Reading speed improvement (mean ± SEM, n = 26). (B) Regression between reading performance and defect reduction measured with Goldmann perimetry (ECSG) after defect training (n = 24). Clearly, the defect reduction was linearly related to the reading improvement after defect training (% change WPM = 3.78 + 1.82 Goldmann ECSG). (C) The same analysis for the intact training revealed no such relation.