Functional MRI of visual cortex predicts training-induced recovery in stroke patients with homonymous visual field defects

Abstract

Post-chiasmatic damage to the visual system leads to homonymous visual field defects (HVDs), which can severely interfere with daily life activities. Visual Restitution Training (VRT) can recover parts of the affected visual field in patients with chronic HVDs, but training outcome is variable. An untested hypothesis suggests that training potential may be largest in regions with 'neural reserve', where cortical responses to visual stimulation do not lead to visual awareness as assessed by Humphrey perimetry-a standard behavioural visual field test. Here, we tested this hypothesis in a sample of twenty-seven hemianopic stroke patients, who participated in an assiduous 80-hour VRT program. For each patient, we collected Humphrey perimetry and wide-field fMRI-based retinotopic mapping data prior to training. In addition, we used Goal Attainment Scaling to assess whether personal activities in daily living improved. After training, we assessed with a second Humphrey perimetry measurement whether the visual field was improved and evaluated which personal goals were attained. Confirming the hypothesis, we found significantly larger improvements of visual sensitivity at field locations with neural reserve. These visual field improvements implicated both regions in primary visual cortex and higher order visual areas. In addition, improvement in daily life activities correlated with the extent of visual field enlargement. Our findings are an important step toward understanding the mechanisms of visual restitution as well as predicting training efficacy in stroke patients with chronic hemianopia.

Keywords: Rehabilitation; Stroke; Training; Visual field defects; fMRI.

Fig. 1

Visual discrimination task. Patients trained either with static point stimuli or optic flow stimuli (moving dots that were presented continuously on screen). The stimulus was presented on the border area of the visual field defect (gray shaded area) and scaled with eccentricity. Patients reported whether the stimulus location was clockwise or counterclockwise located with respect to the line extending from the fixation target (static point stimulus) or rotated clockwise/counterclockwise (optic flow pattern). The target location was cued by the line or background contraction pattern in order to make a directed covert attention shift.

Fig. 2

Scanning paradigm. (A) Setup for wide-field retinotopic mapping inside MRI scanner. (B) Retinotopic mapping stimulus presented inside the MRI scanner. The ring stimulus extended from the fixation target (centered in front of the right eye position) to 45 deg of eccentricity. The wedge stimulus rotated counterclockwise around the fixation target.

Fig. 3

(A) Comparison of Humphrey and neural perimetric coverage maps. This subject (J01) has been diagnosed with a complete left sided hemianopia based on Humphrey perimetry. The lower panel is normalized to this subject’s maximum visibility. An ‘x’ marks the Humphrey test locations with no response (i.e. 0 dB). (B) However, based on the patients’ coverage map resulting from retinotopic mapping the hemianopia is partial with a largely spared upper left quadrant. Following to the hypothesis of Papanikolaou et al. this upper quadrant may be susceptible for functional recovery. The lower panel is down sampled from the upper panel to match the resolution of the Humphrey map and thresholded at 0.5. An ‘x’ marks the Humphrey grid locations with a maximum amplitude of the Gaussian pRF across all voxels inside this grid location below threshold (i.e. 0.5). (C) Four categories were classified based on the lower two panels in A and B: (1) Hum + / Ret + represent coverage for both the retinotopy and Humphrey (coloured grid location in both maps); (2) Hum - / Ret - represent no coverage in either map (‘x’); (3) Hum - / Ret + represent coverage for the retinotopy map only (coloured grid in retinotopy; ‘x’ in Humphrey); (4) Hum + / Ret - represent coverage for the Humphrey map only (coloured grid in Humphrey; ‘x’ in retinotopy. Because this last category is infrequent (1.5% of the data), we excluded this category from further analysis. The three categories are shown for the patient used in this example (J01). The red circles indicate the locations with a training effect (i.e. 0 dB prior to training and > 1 dB after training). Note, the blind spot location is excluded. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 4

Training effects at different visual field locations. (A) Training effects established by Humphrey perimetry (map after training – map prior to training) averaged across all locations from the corresponding category from all patients. (B) Training effects established by Humphrey perimetry averaged across all locations from the corresponding category from all patients that show at least 1 dB visual sensitivity improvement; i.e. locations with exactly the same dB value after training (change = 0 dB) excluded. (C) Training for patients of cohort 4 who received training in the complementary halves of the scotoma (did not received training in the intact field). (D) Training effects for patients of cohort 4 established by Humphrey perimetry averaged across all locations from the corresponding category from all patients that show at least 1 dB visual sensitivity improvement; i.e. locations with exactly the same dB value after training (change = 0 dB) excluded. Note: only positive fluctuations for the green category (Hum+/Ret + ) are shown, since the other categories are also clipped at 0. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 5

As can be deducted from panel (A) and (B), the locations with a training effect of patient J01 (on average about 12 dB) were mainly induced by regions in the primary visual cortex corresponding to peripheral visual information processing (dashed line indicate the calcarine sulcus). (C) However, training effects could also be induced by areas higher up in the visual processing stream (C1-3) in other patients.