Neural Restoration Training improves visual functions and expands visual field of patients with homonymous visual field defects

Abstract

Background:

In recent years, the introduction of visual rehabilitation for patients with homonymous visual field defects has been met with both enthusiasm and caution. Despite the evidence that restitutive training results in expansion of the visual field, several concerns have been raised.

Objective:

We tested the effectiveness of a new rehabilitative protocol called “Neuro Restoration Training” (NRT) in reducing visual field defects and in restituting visual functions in the restored hemianopic area.

Methods:

Ten patients with homonymous visual field defects (lesion age >6 months) where trained in detecting low contrast Gabor patches randomly presented in the blind field, which refers to regions of 0 dB sensitivity, and along the hemianopic boundary between absolute (0 dB) and partial blindness (>0 dB). Training included static, drifting, and flickering Gabors in different blocks. Positions along the hemianopic boundary were systematically shifted toward the blind field according to the threshold reduction during the training. Before and after the training, we assessed visual field expansion and improvement in different high-level transfer tasks (i.e., letter identification and shape recognition) performed in the hemianopic boundary and in the blind field.

Results:

NRT led to significant visual field enlargement (≈5 deg), as indicated by the conventional Humphrey perimetry, and two custom made evaluations of visual field expansion with eye movement control (one static and one dynamic). The restored area acquired new visual functions such as small letter recognition and perception of moving shapes. Finally, for some patients, NRT also improved detection, either aware or not, of high contrast flickering grating and recognition of geometrical shapes entirely presented within the blind field.

Conclusion:

These results suggest that NRT may lead to visual field enlargement and translate into untrained visual functions.

Keywords

Hemianopia Neuro Restoration Training visual field training contrast sensitivity

1 Introduction

Until recently, restoration of visual functions in the case of homonymous visual field defects has been considered impossible. However, a convergence of two factors makes it conceivable nowadays: Firstly, numerous studies in mammals (Kaas et al., 1990; Eysel et al., 1999; Gilbert & Wiesel, 1992) and in humans (Ajina, Kennard, Rees, & Bridge, 2014; Dilks, Serences, Rosenau, Yantis, & McCloskey 2007; Halko et al., 2011; Henriksson, Raninen, Näsänen, Hyvärinen, & Vanni, 2007; Julkunen et al., 2006; Nelles et al., 2007; Tamietto, Pullens, de Gelder, Weiskrantz, & Goebel, 2012) have provided evidence of plasticity in the primary visual cortex after short- and long-term visual deprivation; secondly, other studies have demonstrated that repeated practice in a sensory task elicits long-term perceptual learning that diverts this phenomenon of spontaneous plasticity to a more profitable direction. Taken together, those studies lay the foundation for all the recent neurobehavioral rehabilitative techniques in vision (see Sagi, 2011, for a review). The most convincing evidence comes from the finding that intensive visual training is able to produce an enlargement of the visual field long after the period of spontaneous plasticity occurring after the lesion (Halko et al., 2011; Julkunen et al., 2006; Huxlin, et al., 2009; Kasten, Wüst, Behrens-Baumann, & Sabel, 1998; Kasten et al., 1999; Marshall et al., 2008; Mueller, Mast, & Sabel, 2007; Plow et al., 2011; Plow, Obretenova, Fregni, Pascual-Leone, & Merabet, 2012; Raninen, Vanni, Hyvärinen, & Näsänen, 2007). Associated with the visual field enlargement, there is significant improvement in detecting low-contrast stimuli at the transition zone between the totally seeing and totally blind field; moreover, for some patients, training has been shown to turn blindsight into awareness, as revealed by an increase of explicit perception of stimuli, which, despite destruction of the striate cortex, produce a visual response without awareness (Huxlin et al., 2009). Some studies have demonstrated that neurobehavioral outcomes are associated with cortical reorganization (Halko et al., 2011; Henriksson et al., 2007; Julkunen et al., 2006; Marshall et al., 2008; Raninen et al., 2007).

One open question is whether, associated with visual field enlargement, visual rehabilitation can lead to restitution of visual functions involving processing of complex stimuli such as letters and visual shapes either static or moving. This seems unlikely, based on the suggestion that training activates alternative routes of coarse visual processing with respect to the geniculo-striate pathway (Pessoa & Adolphs, 2010). However, it may be conceivable that training restores the functionality of partially preserved islands of V1 at the perimeter of the brain lesions, thus improving performance in high-resolution tasks (Das & Huxlin, 2010). To explore this possibility, we designed a new training protocol appropriate to modulate the spatiotemporal selectivity and the response to orientation, motion, and flicker of visual neurons in early central visual processing (Casco et al., 2014; Furmanski, Schluppeck & Engel, 2004; Karni & Sagi, 1991; Maniglia et al., 2016). We asked whether the rehabilitation with these stimuli was associated with two outcomes. The first outcome is expected to result from training along the vertical meridian, either within the transition zone, where the damage is partial (Sabel, Henrich-Noack, Fedorov, & Gall, 2011) or between the initial and the final hemianopic border (hB), i.e., the initial and final border between absolute (0 decibel, dB) and partial blindness (>0 dB). In the transition zone and between the initial and final hB we predicted an increase in the eccentricity at which high-resolution tasks, such as visual acuity of letters, can be performed. The second predicted outcome comes from the evidence that some untrained patients can explicitly discriminate amongst four shapes presented within the blind field (Kasten et al., 1998). In this study, recognition was below the discrimination threshold (8.3%). However, blindsight (implicit recognition) may have occurred by exploiting some residual vision in “islands” of surviving tissue inside the blind field (Sabel et al., 2011). In this case, training-dependent restitution of functions achieved with stimuli appropriate to stimulate neurons in V1 should lead to increased awareness for other stimuli and tasks presented in the blind field. Therefore, our rehabilitative approach is particularly focused on the possibility that perceptual learning may transfer to untrained visual functions. We asked not only whether stimulus detectability, as assessed in standard visual field testing, improved, but whether training on stimulus detectability transferred to other stimuli and tasks. It should be noticed that generalization of perceptual learning is not opposed to the principle of specificity of perceptual learning mentioned above. Indeed, although perceptual learning is specific for the trained elementary features, suggesting that the improvement is selective for the stimuli to which orientation, spatial frequency, and direction of motion selective channels respond, perceptual learning can transfer to higher level tasks provided that these tasks rely on the integrity of the trained lower-level mechanisms (Maniglia et al., 2011, 2016).

We aimed at evaluating whether perceptual learning on stimuli to which neurons in the geniculo-striate pathway are selective (Blakemore & Campbell, 1969; Wilson, McFarlane & Phillips, 1983; Shapley & Perry, 1986; Merigan & Maunsell, 1993; Maunsell & Newsome, 1987), produced an enlargement of the visual field in patients with visual field defects and whether this enlargement is associated with increased recognition of letters presented at the new border (hB) between absolute (0 dB) and partial blindness (>dB). Moreover, we aimed at assessing whether perceptual learning resulted in improved vision for untrained stimuli presented within the blind field. To these ends, we trained multiple retinal regions and low-level vision channels, selective to a range of spatiotemporal parameters.

2 Methods

2.1 Subjects and inclusion criteria

Ten patients with homonymous visual field defect were included in the Neuro Restoration Training (NRT) program. Patients’ clinical information is provided in Table 1. The group consisted of six females and four males, with a mean age of 43±14.5 years (range: 19–69 years). Average lesion age at onset of training was 27±31 months (range: 6–108 months). Visual field defects, as documented by neurological records (TC for P1, P3, P6, and P10; RM for P2, P5, and P7; neurological reports for P4 and P8), were caused by damages involving the retrochiasmal visual sensory pathway, extending to the occipital cortex. They all met the following inclusion criteria: damage age ≥6 months, absence of uncontrolled epilepsy and of cognitive impairments interfering with training (learning, memory, or attention deficits), sufficient fixation stability, valid baseline and post-training campimetric exams available. All but one (P10) subject had normal or corrected-to-normal vision in the sighted field. Following the trauma, P10 suffered blurred vision in the visual field contralateral to the hemianopic visual field defect.

Table 1
Patient’s information

Patient Cause Lesion Type Sex Age Months from the lesion

P1 Stroke Tempo-Parieto-Occipital F 51 36

P2 Stroke Tempo-Parieto-Occipital F 55 12

P3 Stroke Thalam-Occi-Cerebel M 41 22

P4 Trauma Fronto-Tempo-Occi-OptRad M 35 108

P5 A.V.M. Temporal Posterior F 55 8

P6 Trauma Tempo-Occipital-Cerebel F 19.5 12

P7 Stroke Occipital M 69 6

P8 A.V.M. Occipital F 29 7

P9 Stroke Fronto-Parietal-Tempo- Occi F 35 47

P10 Trauma &stroke Tempo-Occipital M 43 13

Patient	Cause	Lesion Type	Sex	Age	Months from the lesion
P1	Stroke	Tempo-Parieto-Occipital	F	51	36
P2	Stroke	Tempo-Parieto-Occipital	F	55	12
P3	Stroke	Thalam-Occi-Cerebel	M	41	22
P4	Trauma	Fronto-Tempo-Occi-OptRad	M	35	108
P5	A.V.M.	Temporal Posterior	F	55	8
P6	Trauma	Tempo-Occipital-Cerebel	F	19.5	12
P7	Stroke	Occipital	M	69	6
P8	A.V.M.	Occipital	F	29	7
P9	Stroke	Fronto-Parietal-Tempo- Occi	F	35	47
P10	Trauma &stroke	Tempo-Occipital	M	43	13

Demographic data of study sample; etiology, lesion type, sex, age of patients at baseline assessment, age of lesion at baseline assessment. A.V.M. refers to arteriovenous malformation.

The treatment protocol was reviewed and approved by the University of Padova’s Research Subjects Review Board (protocol no. 2297). The research adhered to the tenets of the Declaration of Helsinki and was conducted after obtaining the patients’ informed, written consent.

2.2 Measures of visual performance

Measures of visual performance were conducted at baseline, at the end of the training, and during the training, with intervals amongst measurements of approximately 3 months. There were between one and three evaluations during the training, depending on training duration (see Table 2).

Table 2
Training conditions

Patient Trained stimulus SF (cycle/deg) Sigma (deg) hB 1 (deg) hB 2 (deg) BLIND FIELD CATCH Dur (month)

0Hz 6°/sec 20Hz 1st last 1st last 1st last 1st last

P1 yes yes yes 2 3 15 15 10 10 15 15 –15 –15 4

7 7 –4 –4 –7 –7 –7 –7

P2 yes yes yes 1 3 7 7 7 7 9 9 –9 –9 5

–7 –7 7 7 0 0 0 0

P3# yes yes yes 3-1 1.5 5 5 15 15 20 15 20 15 3.5

–2 –2 –6 –6 –2 –6 –8 –6

P4° yes no no 1.5 4.5 10 13 – – – – – – 3

0 0 – – – – – –

P5 yes yes yes 2 3 –5 –13 –12 13 –20 –15 12 15 3

–5 –5 0 5 5 0 0 0

P6 yes yes yes 2 3 16 18 16 18 16 16 –16 –16 5

6 6 –6 –6 0 0 0 0

P7 yes yes yes 1 3 –10 –14 –10 –14 –15 –17 15 17 4.5

–2 –2 2 2 0 0 0 0

P8 yes no no 9-3 1 3 4 5 7 10 10 –10 –10 3

3 3 0 0 0 0 0 0

P9 yes yes yes 3 3 5 7 5 9 11 11 –11 –11 6.5

5 6 11 7 5 0 5 0

P10 yes yes yes 9-3 3 –5 –9 –5 –7 7 –10 7 7 12

5 5 –5 –5 5 3 5 11

Patient	Trained stimulus	SF (cycle/deg)	Sigma (deg)	hB 1 (deg)	hB 2 (deg)	BLIND FIELD	CATCH	Dur (month)
P1	yes	yes	yes	2	3	15	15	10	10	15	15	–15	–15	4
						7	7	–4	–4	–7	–7	–7	–7
P2	yes	yes	yes	1	3	7	7	7	7	9	9	–9	–9	5
						–7	–7	7	7	0	0	0	0
P3#	yes	yes	yes	3-1	1.5	5	5	15	15	20	15	20	15	3.5
						–2	–2	–6	–6	–2	–6	–8	–6
P4°	yes	no	no	1.5	4.5	10	13	–	–	–	–	–	–	3
						0	0	–	–	–	–	–	–
P5	yes	yes	yes	2	3	–5	–13	–12	13	–20	–15	12	15	3
						–5	–5	0	5	5	0	0	0
P6	yes	yes	yes	2	3	16	18	16	18	16	16	–16	–16	5
						6	6	–6	–6	0	0	0	0
P7	yes	yes	yes	1	3	–10	–14	–10	–14	–15	–17	15	17	4.5
						–2	–2	2	2	0	0	0	0
P8	yes	no	no	9-3	1	3	4	5	7	10	10	–10	–10	3
						3	3	0	0	0	0	0	0
P9	yes	yes	yes	3	3	5	7	5	9	11	11	–11	–11	6.5
						5	6	11	7	5	0	5	0
P10	yes	yes	yes	9-3	3	–5	–9	–5	–7	7	–10	7	7	12
						5	5	–5	–5	5	3	5	11

For each patient, the table reports: i) the trained stimuli: static (0 Hz), motion (6°/sec), flicker (20 Hz); ii) the spatial frequency (SF); iii) the Sigma reflecting stimulus size (stimulus size is defined by the sigma of the Gaussian envelope expressed in deg); iv) the stimulus position, expressed in degrees of visual angle along the two positions along the hemianopic borders (hB1 and hB2, first row represents x-axis, second row represents y axis); v) the stimulus position within the blind field and the seeing field where catch trials were presented; and vi) the duration of the NRT in months. When there is more than one SF (such as P3, P8, and P10), it means that different spatial frequencies were used along hB1 and hB2. In the hBs, the first number represents the eccentricity in deg along the x-axis, and the second number represents the eccentricity in deg along the y-axis. Zero is the fixation point. Negative numbers represent left (x-axis) or upper (y-axis) position in relation to the fixation point. °this patient was trained with a static and fixed-position Gabor of sigma 4.5. ^#this patient had only one position in the hB and three in the BF.

2.3 Apparatus

Stimuli were generated using Matlab Psychtoolbox (Brainard, 1997; Pelli, 1997). They were displayed on a 24-inch Asus ML248H LCD monitor with a refresh rate of 60 Hz. The screen resolution was 1920×1080. Gamma correction for each color channel was applied through calibration with the Spyder 4 Elite colorimeter (DataColor). The calibration was further verified using a Minolta LS-100 photometer, which indicated that the mean luminance was 50 cd/m². In order to represent 10.8 bits of luminance (1786 gray levels) on an 8-bit display to reach a theoretical threshold value of 0.0011 Michelson contrast with a mean luminance of 50 cd/m², we adopted a software solution called “Pseudo-Gray,” also known as “Bit-Stealing,” which was implemented via the Psychtoolbox built-in function.

2.3.1 Eye movement recording

All pre-training, intermediate, and post-training measurements were conducted with Mirametrix eye tracker turned on to control fixation. Gaze accuracy as reported by the manufacturer ranged from 0.5 to 1 degrees (deg) with a drift of less than 0.3 deg. A stimulus was not presented if the eyes deviated from fixation by more than 1.5 deg.

2.3.2 Visual field enhancement tests

The mapping of the spatial extent of each patient’s reported visual deficit was conducted before the onset, during and after the offset of the visual training with four different measurement methods:

Humphrey visual field (Humphrey perimeter (740, 745, or 750). The threshold test was blindly conducted on a 30-2 field size by an ophthalmic technician, with the display screen to allow eye monitoring turned on. For five patients that completed the training (P3, P5, P6, P7, P9), two visual fields, collected at least 4 months before NRT started, were also available, which could be compared to evaluate the hB and dB variations without NRT.

HRDP (high-resolution dynamic perimetry). This is a non-standard ophthalmic procedure. It was implemented with a very bright target (0.9 Weber contrast) in order to measure the hB. Participants sat 60 cm from the screen (24-inch Asus) and were required to maintain fixation throughout the entire measurement of the visual field. A circular white spot of 0.5 deg in diameter appeared smoothly behind one of the two invisible vertical stripes 1 deg wide (same color as the dark gray background), positioned at the left and at right border of the screen. The spot appeared randomly from the left or right border with equal probability. The white spot moved along a horizontal trajectory at 3 deg/sec for 41 deg, until it disappeared behind the invisible stripe positioned at the opposite side of the screen. Patients had to press the appropriate response-key as fast as possible when they perceived the target to disappear/reappear either from behind the blind field or from the invisible stripe in the seeing field. After 10 practice trials, a block of 96 trials randomly presented consisted of one repetition for each of the two directions and for each of the 48 equidistant starting points positioned along the vertical axis of 12.8 deg above and below the fixation cross. Mirametrix eye-tracker measured the eyes’ position at the moment of button press. The distance between the fixation cross and the eye position on the horizontal axis measured by the eye-tracker was used to adjust the point of disappearance and reappearance in the perimetry output. The very bright spot could be detected within the region of partial blindness but not in the field of total blindness. Thus, HRDP allowed for deriving the hB, by interpolating the points along the vertical axis, each resulting from averaging the positions, along the horizontal axis of target disappearance/reappearance.

Visual acuity at the hB. Visual acuity for eccentric SLOAN letters was measured before and after the NRT. The stimuli were 10 white SLOAN letters (C, D, H, K, N, O, R, S, V, and Z), randomly presented for 200 ms on a black background. The size of the letters varied according to a psychophysical adaptive procedure - maximum likelihood procedure - (Grassi & Soranzo, 2009; Green, 1990). Size letter threshold, corresponded to 55% of participants’ psychometric function. Patients had to verbally report the target letter, and the experimenter registered the answer. The session terminated after 60 trials in which the letters were randomly presented either on the sighted or on the blind field. This enabled controlling for compensatory shifts of fixation or eye movements that could reduce the measured blind field size. The initial width of the lines defining the letters was 40 arcmin. An acoustic cue was presented right before stimulus onset to reduce temporal uncertainty. No feedback was given to the patients. The letters had a vertical eccentricity of 0 deg and horizontal eccentricity was varied in independent blocks in 2 deg steps, from 0 to the eccentricity that lowered acuity to 40 arcmin.

Grating detection at hB (hB-grating). To test the hB displacement, we used a stimulus consisting of a static vertical grating presented within a square aperture of 14 deg of diameter. The spatial frequency was 1 cycle/deg and the Michelson contrast was equal to 0.9 to allow for the localization of the hB. After familiarizing with the stimulus and task, in the first block, the grating was presented within the blind field, with the external border coinciding with the vertical meridian. In successive blocks, the eccentricity was increased in 2 deg steps until accuracy reached the chance level. Stimulus duration was 200 ms. In each trial, the participant viewed a fixation mark (a white cross) followed immediately by two stimulus intervals separated by a 2 sec pause which contained the fixation. One stimulus interval was blank (i.e., the screen background), whereas the other presented the grating. The order of the intervals was random. The participant’s task was to report whether the grating was presented in the first or in the second interval (i.e., two-interval forced choice task). Acoustic feedback was given to signal wrong responses. Each eccentricity block consisted of 45 trials randomly presented; 30 of them had the grating presented within the blind field. In the remaining 15 catch trials the grating was presented contralaterally at the same eccentricity, with contrast decreasing from 0.2, as a function of the participant’s response, according to a psychophysical adaptive procedure (maximum likelihood procedure) that tracked the 75% point of the participants’ psychometric function.

2.3.3 Blind field testing

Testing of visual performance within the blindfield was conducted before, during, and at the end of training with two tests.

Blindsight testing. Participants were tested for residual spatial channels (Sahraie et al., 2003; Sahraie, Trevethan & MacLeod, 2008). Stimuli were Gabor patches consisting of a cosinusoidal carrier enveloped by a stationary Gaussian. The Gabor patch was characterized by the wavelength (λ), phase (φ), and standard deviation (σ) of the luminance Gaussian envelope in the (x, y) space of the image. Formally, each Gabor patch can be expressed as follows:

G (x, y) = cos ((2 π / λ) x + φ) e^{(x^{2} + y^{2}) / σ^{2}}

Blindsight testing was conducted with the Gabor center at a 15 deg distance from the hB (as delimited by HRDP). Spatiotemporal parameters of the Gabors targeted residual channels in blindsight: spatial frequency of 1 cycle/deg, temporal frequency of 20 Hz, size 9 deg and duration of 200 ms (Sahraie et al., 2008). Contrast was 0.9. Each block consisted of 60 test trials. At stimulus offset, the observer performed the two-interval forced choice task and then pressed the spacebar whenever he or she was aware of the Gabor presented in the blind field. Acoustic feedback was given to signal wrong responses. Randomly intermixed with the test trials, there were 15 catch trials in which the Gabor was presented contralaterally at the same eccentricity, with the contrast decreasing from 0.20, as a function of the participant’s response, according to the same adaptive procedure used for the hB-grating test.

Shape recognition testing. Eight patients were tested, following a practice block of 10 trials. Seven of them had a sufficiently large blind field to enable presentation of the stimuli completely in the blind area. Patients were evaluated in order to assess their ability to identify a geometrical white shape 5 degrees high and 2–5 deg wide, with the center displaced by 15 deg from the hB. At baseline and in any of the following tests, the stimuli were presented until response. There were 30 stimuli consisting of three repetitions of 10 shapes (circle, square, horizontal rectangle, star, oval, heart, cross, isosceles triangle, right triangle, vertical rectangle) presented in random order. Note that although some of the shapes could easily be distinguishable based on coarse channel output, others, such as the two triangles, the circle, and the oval, are not immediately discriminable, even with normal vision at 15 deg eccentricity. Observers performed a three-alternative forced choice task in which they had to indicate which of three shapes (randomly selected from the sample of 10) had been presented in that trial. At the end of testing, participants were asked to verbally report their degree of awareness of the stimuli.

2.3.4 Training sessions

After completing baseline testing, a computerized perceptual learning procedure (NRT) was set up in each patient’s home. The procedure used algorithms that controlled the eccentricity of the stimuli and their spatiotemporal parameters, but no eye-tracker was used during the training.

Stimuli were viewed in the dark at 57 cm distance from the computer monitor and a chin–forehead-rest was used to maintain the distance from the screen.

With few exceptions to meet individual therapeutic needs (see Table 2), NRT was administered following a standard protocol comprising the following parameters and conditions: Gabor presented statically, moving at 6 deg/sec or flickering at 20 Hz. Moving and flickering Gabors were always vertical whereas the static Gabors had four orientation conditions (vertical, horizontal, tilted at 45° and 135°). The spatial frequency varied between 1 and 4 cycles/deg, depending on Gabor eccentricy (higher spatial frequency in the hB and for static Gabors). Stimulus size (Table 2) was defined by the sigma of the Gaussian envelope expressed in deg. Stimulus duration was 200 ms. Participants had to perform a two-interval forced choice task, with Gabor contrast decreasing from 0.99, as a function of the observer response, according to the 3-down 1-up staircase adaptive procedure, which tracks 79.4% of the psychometric function. We defined the threshold as the average of the last six reversals. In each of the trained retinal positions, we ran a separate staircase. Two of the observers had a fixed Gabor location. For the others, the Gabor location varied randomly within each block: Two locations were chosen with about 10% of the Gabor overlapping the hB to obtain a contrast threshold in between floor and ceiling at the beginning of the training; the third location was in the blind field (eccentricity 15 deg or higher); and the fourth was symmetrical to it in the contralateral seeing field. Figure 1 shows one example of stimulus positioning.

Fig.1

The figure shows one example of stimuli position: two stimuli along the hemianopic borders, one stimulus in the blind field and one stimulus in the seeing field. Bear in mind that in a Gabor stimulus the carrier is presented within a Gaussian envelope, which smooths the contrast at the border.

One daily session consisted of five blocks and had approximately one-hour duration: In four of them, the Gabor was static and had one of the four orientations. In the fifth block, the Gabor was either flickering or moving. Participants performed four sessions in a week, and in each daily session, the orientation of the static Gabor was varied. As soon as one block was terminated, data were sent online to our laboratory, and they were analyzed weekly in order to evaluate training progression. Thresholds for the control location were monitored weekly. Because viewing was normal there, thresholds had to remain stable or be progressively reduced. Instead, if the patient executed compensative eye movements (not controlled with eye tracker during NRT) or did not maintain the appropriate viewing distance and room lighting, this would be reflected in anomalous threshold variation. E-mail feedback was sent to the patient each week containing: a) a summary of the threshold he/she had obtained, b) further suggestions to optimize viewing conditions when they did not appear adequate from the results in the control condition, and 3) eventually, when contrast thresholds were 0.2 or lower in one or more conditions, a new version of the training program was installed with the Gabors located deeper in the blind field. The chosen duration of training (Dur, Table 2) varied between 3 and 12 months and reflected a compromise between compliance and effectiveness. Indeed, such a long-lasting training needed to be sustainable and integrated into daily life activities.

2.3.5 Statistical analysis

Two-way repeated measures ANOVAs were conducted to assess the perceptual learning effects on contrast thresholds; t-tests with Bonferroni corrections were employed for pairwise comparisons. Perceptual learning results, expressed as threshold modulation - log₁₀(after/before NRT), - and the effects of perimetry data were analyzed using one-sample t-tests based on the null hypothesis of 0 effect. Transfer results were analyzed with t-tests and compared using correlation analysis.

3 Results

3.1 NRT effect on trained stimuli

Figure 2 shows the contrast threshold, obtained by first averaging thresholds obtained within a given eccentricity condition and then across eccentricity conditions, at the hB and within the BF, before and after NRT with the static, motion, and flickering Gabors. To assess the NRT effect on contrast thresholds, we conducted two-way repeated measures ANOVAs with NRT (before vs. after NRT) and Stimulus (static, motion and flicker) as factors. NRT results for stimuli presented along the hB (n = 9, P8 was excluded because she was only trained in the static condition) revealed a significant effect of NRT (F_1,8 = 36.2, p = 0.0001, partial-η² = 0.82) but not of stimulus (F_2,16 = 0.83, p = 0.45, partial-η² = 0.09) and of the stimulus×NRT interaction (F_2,16 = 0.14, p = 0.87, partial-η² = 0.02). Similar results were obtained for stimuli presented in the BF (n = 8, i.e, the patients that had stimuli presented in the BF): The effect of NRT was significant (F_1,7 = 15.5, p = 0.006, partial-η² = 0.69), but not that of stimulus (F_2,14 = 0.77 p = 0.77, partial-η² = 0.036) and of the NRT×stimulus interaction (F_2,14 = 1.42 p = 0.27, partial-η² = 0.17). The bottom panels show average NRT results expressed as threshold modulation, corresponding to log₁₀(after/before NRT), obtained at the hB (Fig. 2, low left panel) and in the BF (Fig. 2, low right panel). An NRT modulation < = >0 indicates that contrast thresholds are reduced, unaffected, or increased by NRT, respectively. A one sample two-tailed t-test revealed a significant enhancement (based on the null hypothesis of 0 enhancement) for the static (hB: t₉ = –3.6, p = 0.006; BF: t₇ = –3,25, p = 0.014) and moving stimulus (hB: t₈ = –6.97, p = 0.0001; BF: t₇ = 2.33, p = 0.05), whereas for the flicker stimulus, the enhancement was only significant when the stimulus was presented along the hB (t₉ = –3.89, p = 0.005). The enhancement did not reach significance in the control zone of normal vision (p = 0.50).

Fig.2

Initial and final contrast threshold and threshold enhancement are shown for stimuli located on the hemianopic border (left part) and within the blind field (right). ^*p≤0.5; ^**p≤0.01; ^***p≤0.001.

3.2 Visual field enlargement

3.2.1 Humphreys

Figure 3 represents results of Humphrey assessed monocularly before and after the NRT (a square indicates a 6×6 deg area). For this analysis, we chose the eye with the smaller number of fixation losses. The hB between the complete blind field (0 dB) and the field with some residual vision (>0 dB) is indicated by a dotted line (before NRT) and by continuous tick lines (after NRT). The white region indicates the visual field enlargement after NRT (the figure caption gives an example of how the hB enlargement was calculated). Each squared region contains a number, indicating the variation in sensitivity (dB) following NRT: The increase in sensitivity following NRT is indicated with black numbers >0, whereas white numbers >0 indicate a decrease in sensitivity in the perimetry performed after NRT. Individual and average hB (12 deg above and below fixation) is shown for when patients performed NRT between the two measurements (Fig. 4a) and when they did not (only five patients could be tested in this control condition, Fig. 4b). The overall hB displacement (obtained by subtracting the second measurement of hB from the first) was equal to 5.85 deg (±6.1) with NRT in between, and the difference was significant (paired sample t-test: t₉ = 3.05, p = 0.014). The average hB displacement with no NRT in between was equal to –0.3. Individual net variations in dB obtained in the second measurement with respect to the first (∑dB increase –∑dB decrease) are shown in Fig. 4c for the condition with NRT performed in between the two measurements and Fig. 4d for when NRT was not performed. NRT produced an average net increase in sensitivity (within the areas with sensitivity change) of 6.23 dB, which significantly differed from 0 (one sample t-test: t₉ = 3.46, p = 0.007). Without NRT between the two measurements, there was a small, non-significant, reduction in sensitivity. Comparison of fixation loss, false positive (key press with no stimulus present), and false negative (failure to detect previously seen stimuli) response revealed a decrease of proportion of fixation loss after NRT (from 0.13 to 0.1, t₉ = 2.15, p = 0.07) an increase in the percentage of false positive response (from 2 to 4%, t₉ = –2.1, p = 0.056) and no change in false negative responses (2 to 1.6%, t₉ = 0.3, p = 0.77).

Fig.3

In Fig. 3, the boundary of the hB, i.e., between the complete blind field (0 dB) and the field with some residual vision (>0 dB), expressed in degs (each square 6×6 degs), is indicated by a dotted line (before NRT) and by continuous tick lines (after NRT). The analysis of the hB displacement was conducted along the boundary of the four intermediate squares (where stimuli were positioned during NRT). For example, the hB position of P5 was 6 deg on average before NRT (dotted boundary) and 12 deg on average after NRT (tick boundary), with hB displacement due to NRT of 6 deg. Visual field enlargement refers to the difference between the hB before and after NRT, which was significant (p = 0.014). Sensitivity variation was also indicated with black numbers (NRT-dependent increase in sensitivity) and white numbers (decrease in sensitivity after NRT). For this analysis, we chose the eye with the smaller number of fixation losses. The defective region is always displayed to the right.

Fig.4

Figure 4a shows individual and average hB displacement in the trained locations for the trained patients. Figure 4b shows individual and average hB displacement derived from two visual field measurements conducted before NRT started in five patients. Figure 4c shows the net sensitivity variation (expressed in dB) for each trained patient within the 6×6 deg areas showing a change in sensitivity. Individual net variation in sensitivity was obtained by subtracting the total decrease in dB post NRT (average of white numbers) from the total increase in dB following NRT (average of black numbers). The group effect was significant (p = 0.007). The variation in dB when NRT did not occur in between the two measurements (5 patients) is also shown in Fig. 4d. ^*p < 0.5; ^**p < 0.01.

3.2.2 HRDP

Figure 5 shows the hB, as measured with HRDP before and after NRT. The HRDP for P8 could not be obtained because the BF of this patient had many islands of residual vision. The average amount of enlargement (4.7 deg, ±1.3) is shown in the last panel of Fig. 5. The t-test comparing hB before and after NRT was significant (t₈ = 3, p = 0.017). The second to last panel shows that the horizontal eye movement dispersion of each patient is within 1 deg most of the time. Importantly, there is a correlation between the Humphrey and the HRDP perimetries, both when the hBs obtained before NRT was compared (r = 0.82, p = 0.006) and when the hB displacement following NRT was compared (r = 0.88, p = 0.002).

Fig.5

The hB from 12.5 deg above and below fixation as measured before NRT (gray) and after NRT (black) is indicated for each patient. The average displacement from the vertical midline measured before (pre) and after (post) NRT is shown in the last panel. The difference was significant (p = 0.017). The second to last panel shows that the major part of the horizontal eye movement dispersion is within 1 deg. ^*p < 0.5.

3.2.3 Peripheral acuity

NRT-dependent increase in SLOAN visual acuity is shown in Fig. 6, expressed as a reduction of the slope of the regression line fitted on the average acuity data obtained as a function of eccentricity: from 0 to 12 deg from midline. Data for a control group of six observers with normal vision, aged 22–32 years, were also fitted with a regression line, which is shown for comparison in Fig. 6. The t-test comparison of individual slopes before and after NRT resulted in a significant reduction (t₉ = 4.23, p = 0.002). The inset shows the maximum eccentricity at which acuity reached the lowest value tested (40 arcmin): from 5.2 before NRT to 8.8 deg, after NRT (t₉ = 4.63, p = 0.001). Interestingly, the individual peripheral visual acuity (as reflected by the individual slope) correlates with the hB position after (Humphrey: 0.50, HRDP: 0.71) but not before NRT (Humphrey: 0.12, HRDP: 0.08). The difference in visual acuity obtained before (5.22) and after NRT (4.86) in the seeing field at 4 deg eccentricity was not significant (p = 0.46).

Fig.6

Figure 6 shows the regression lines fitted to the visual acuity for SLOAN letters, measured as a function of letter eccentricity, for hemianopic patients before (gray line) and after NRT (black line) and for the control group (broken line). Acuity increase, expressed as a reduction of the slope of the regression line, is significant (p = 0.002). The inset shows that the maximum eccentricity at which VA could be measured also increased after NRT. ^**p < 0.01.

3.2.4 hB-grating

Figure 7 shows the psychometric functions fitted to the proportion of correct detections obtained as a function of the distance between the boundary of the grating and the vertical meridian. The P50 indicates the point on the function corresponding to the distance at which detection performance was at chance. The NRT dependent increase of P50s was equal to 5.8 deg, corresponding to an NRT-dependent displacement of hB similar to that obtained when comparing the average individual differences (6.87 deg, t₉ = 4.62, p = 0.001). The contrast threshold for the same stimulus in the seeing field measured at the eccentricity of the hB before (0.025) and after NRT (0.031) did not differ (p = 0.34).

Fig.7

Psychometric functions are fitted to proportion of correct detection as a function of the separation between the grating boundary and the vertical meridian, measured before and after NRT. The NRT-dependent increase of P50s was significant (p = 0.001). ^**p < 0.01

3.3 Blind field testing

3.3.1 Blindsight

Blindsight could be measured, with the task designed to evaluate residual spatial channels in the BF, in the eight patients who had no residual vision in the blind field of the Humphrey test (P3 and P8 excluded, and for P1, the stimulus was presented in the upper visual field where it was not visible). Results are shown in Fig. 8, in which white stars indicate significance in the binomial test and numbers refer to the degree of awareness (0: never seen; 1: light; 2: fragment; 3: sometimes; 4: very often; 5: always). Before NRT, only one patient (P6) had above chance performance with some awareness. After NRT, the binomial test was significant for P4, P5, P6, and P10. P4 and P10 had no awareness (P10 perceived light), where P5 reported to perceive the grating. In the seeing field and at corresponding eccentricities, contrast thresholds for the same stimulus, measured before (0.035) and after NRT (0.043), did not differ (p = 0.52).

Fig.8

Percent correct in the task of detecting a low-contrast 1cpd Gabor drifting at 20 Hz, designed to evaluate residual spatial channels in the BF.

3.3.2 Shape recognition

Figure 9 shows percentage of correct identification of shapes presented entirely within the blind field for eight patients (P3 and P8 had residual vision in the blind field and were not tested). Overall, NRT increased both identification accuracy (from 39.8 to 58.14 percent, t₇ = 2.57, p = 0.037) and awareness. Amongst the eight patients tested, five (P4, P5, P6, P7, P10) showed an improvement, but for two of them, the improvement was small and there was no awareness of the shapes that they had identified correctly. Importantly, the NRT-dependent enhancement (log₁₀(after/before)) in this task had a non-significant correlation with the enhancement in the task devoted to testing residual channels (r = 0.49, p = 0.2). Indeed, four patients improved in both tasks, whereas three of them did not improve in either task.

Fig.9

Percent correct in the task of recognizing a white geometrical shape presented in the blind field increased after NRT (p = 0.037). The horizontal gray line indicates chance performance.

3.4 General discussion

After NRT, all patients achieved a significant reduction in contrast threshold for the trained Gabor stimuli located along the hB (but not for the one located in the control zone over the seeing field), indicating increased sensitivity along the boundary between partial and absolute blindness. NRT led to significant displacement of this hB (≈ 5-6 deg), i.e., a displacement toward the blind field of the vertical boundary between no vision at all and some residual vision. Not only did patients report subjective experience of the enhancement, but this enhancement is indicated by three different evaluations of hB displacement: conventional Humphrey, HRDP, and hB-grating. Importantly, the results also showed an increase in visual acuity for SLOAN letters, which, after NRT, become identifiable at increasing distances (≈ 3-4 deg on average) from the pretest hB. The improvement in contrast sensitivity for the NRT stimuli also occurred for the blind field location. For some of the patients, this improvement was associated with an improvement in the task designed to evaluate residual spatial channels producing blindsight in the blind field and in the task of recognizing geometrical shapes entirely presented within the blind field.

One fundamental issue has to do with the possibility that the evaluation of hB displacement may be invalidated by loss of fixation, practice effect, and/or false positive (the tendency to respond when stimuli were not presented) (Das & Huxlin, 2010). It has been questioned whether these artefacts could have produced an imprecise evaluation of hB displacement and of detection sensitivity change within the transition zone between the seeing and blind fields. In the present study, we attempted to avoid this potential confounding effect in various ways. First, in detection paradigms (NRT, hB-grating and blindsight Gabor), false alarms were avoided by using bias-free two-alternative forced choice task instead of a yes/no task (Azzopardi & Cowey, 1997). In these detection tasks, we also included 25% of catch trials, allowing for further control of fixation loss, practice effects, and response bias. Indeed, this confounding would have affected both measurements within the blind and seeing fields, but it did not. This confounding was also prevented when assessing identification of either letters in the hB or shapes in the blind field by using the maximum likelihood adaptive procedure with ten alternative stimuli and three alternative forced-choice procedures, respectively. Moreover, because conventional Humphrey perimetry does not provide a precise control for fixation, we designed two further tests to confirm hB displacement: HRDP and hB-grating. Both of these tests, as well as the SLOAN acuity test and the tests designed to evaluate vision in the blind field (blindsight Gabor and shape identification), were performed with eye movement control.

3.4.1 hB displacement

Differently from the standard Humphrey perimetry, which uses light stimuli of varying intensities (brightness), HRDP and hB-grating use suprathreshold contrast stimuli, which may be detected even in regions of partial blindness. Sabel, Kenken and Kasten (2004) claimed that the mismatch between the different techniques is more likely due to stimulus size rather than to whether the technique involved suprathreshold or threshold measurement. However, differently from automatic visual field testing, which assesses small variations in sensitivity within the zone with some residual vision, HRDP and hB would not detect these variations, but they would precisely define changes in the border between absolute and partial blindness. Despite this substantial difference, the correlation between Humphrey perimetry and HRDP (but not hB-grating) was very high at baseline. The correlation between the amount of recovery by NRT as measured with Humphrey and HRDP and between Humphrey and hB-grating was also high. One important result is that the hB displacement, when computed along the entire vertical meridian of Humphrey, was smaller than when computed in the trained location. This difference may be accounted for by selectivity of perceptual learning for the retinal trained region (Huxlin et al., 2009).

To sum up, three independent measurements show an NRT dependent displacement of the hB. This cannot be accounted for by spontaneous and/or use dependent short-term plasticity, considering that it occurred as a consequence of NRT protocol applied at least six months after the injury. A relevant question is whether these hB displacements can be interpreted as reflecting plasticity phenomena at the level of the neural representation of the hB. One possibility is that NRT restored the functionality of intact islands of V1 at the perimeter of the brain lesions whose function was reduced by the damage after the injury (Huxlin et al., 2009). Kasten et al. (1999) have provided a taxonomy of cortical blindness according to the extension of the transition zone in which residual visual functions remain. Type I comprises patients with a sharp transition zone between intact and completely blind visual field. Type II has a transition zone of medium extension and Type III has larger transition zones. According to Sabel and coworkers (2011), the hB enlargement occurs when the damage is partial. These restorative outcomes may also reflect plasticity phenomena such as receptive field expansion, so that neurons responding to stimuli presented at the boundary of the blind field would respond, following NRT, to stimuli presented within the transition zone of the visual field. Animal studies (Eysel et al., 1999) suggest that the cortical receptive field in the adult visual cortex may be modified by a Hebbian, LTP-like plasticity. These authors found that the area of the excitatory receptive field of a cell into a primarily unresponsive cortical region was transitorily enlarged following 10–60 minutes of co-stimulation with alternating on- and off-stimuli.

However, this cannot be the only explanation because Type III patients did not always have the largest enlargement, and some Type I patients, with sharp hB (P2, P6, P7, and P8), showed a significant hB displacement after NRT. An alternative intriguing hypothesis is that NRT with stimuli at the hB restored/unmasked/formed new callosal neural connections with the contralateral hemisphere (Dougherty, Ben-Shachar, Bammer, Brewer, & Wandell, 2005; Henriksson et al., 2007; Bridge, Thomas, Jbabdi, & Cowey 2008; Ajina et al., 2014; Berlucchi, 2014). With regard to this hypothesis, Bridge et al. (2008) found no evidence of increased cortico-cortical connectivity in V1 in the hemianopic with respect to controls, and, actually, weaker V1-V1 connections were found in the hemianopic patients. However, Perez et al. (2013) showed, for stimuli presented in the central visual field, a cerebral reorganization of the posterior areas following brain injury in hemianopic patients, which, for right brain damage, consisted of a more bilateral activation in the occipital lobes. NRT along the hB could perhaps enhance bilateral activation so that the receptive field contralateral to the BF may be recruited to perform high-resolution tasks. This explanation is more compatible with our finding that after NRT, there was an increase in the eccentricity at which visual acuity tasks could be measured (approximately 3-4 deg), a result suggesting that a small receptive field size is still functionally active at a large eccentricity.

3.4.2 Blindsight

Several studies have indicated that cortically blind patients may be able to detect, localize, and discriminate visual stimuli despite being phenomenally visually unaware of them (Cowey, 2010). Moreover, it has been shown that training of patients with blindsight increases the awareness of stimuli presented in the blind field. Raninen et al. (2007) showed that detection of flickering discs and discrimination of flickering letters at 10 deg was drastically increased by training. Unfortunately, the long exposure (2 s) and unprecise eye movement control did not ensure that only the blind field was stimulated. Sahraie and others (2003) reported a training-dependent increase in the response of residual spatial channels selective to low spatial and high temporal frequency in the blind field. Huxlin and coworkers (2009) found a striking increase in responses to coherent motion, a stimulus that can be coded in residual extrastriate areas devoted to processing complex visual motion even after V1 damage. Most of our patients were trained with stimuli presented in the blind field. After NRT, we tested them for blindsight; three of our patients (P4, P5, and P10) increased (and P6 maintained) the sensitivity for high-contrast stimuli designed to stimulate residual spatial channels in the blind field, a result accounted for by recruitment of existing extra-geniculate pathways (Das & Huxlin, 2010). Interestingly, two of them (P4 and P10) never reported awareness whereas P5 turned blindsight into awareness.

Most importantly, an improvement was also found in some patients for an untrained task consisting of the identification of simple geometrical white shapes entirely located in the blind field. For two patients, the improvement was associated with awareness. Note, however, that both these stimuli had high contrast, whereas, when contrast was low, as during NRT, the improvement obtained with flickering stimuli in the BF was small. This may suggest that although NRT is designed to produce contrast gain at an early processing level, the generalization to other task relies on higher levels of visual processing areas rather than on the response of intact neurons in V1, which behave as a contrast gain control mechanism (Papageorgiou, Papanikolaou, & Smirnakis, 2014). We speculate that higher level visual areas in the ventral stream (V2 and V4) are activated by direct input from the lateral geniculate nucleus, where selectivity for spatiotemporal frequency is found, even in the absence of input from V1 (Schmid, Panagiotaropoulos, Augart, Logothetis, & Smirnakis, 2009; Schmid et al., 2010; Henriksson et al., 2007; Goebel, Muckli, Zanella, Singer, & Stoerig, 2001). These pathways can be either direct or indirect, consistent with the reverse-hierarchy model of perceptual learning (Ahissar & Hochstein, 1996). In agreement with this model, perceptual learning may start in high-level dorsal areas as a consequence of the input received from extra-geniculate pathways bypassing V1 and then increase the functional response of retinotopic extrastriate visual areas in the ventral stream, such as V2, V3, and V4, by backward propagations from dorsal stream areas to these ventral stream areas.

4 Conclusions

This exploratory study presents some limitations; the major one is that the patient group is small and with lesion type affecting the visual cortex and higher-level visual areas. Moreover, the NRT method proposed is very demanding for the patients. However, other (shorter) types of training seem inappropriate (Vedamurthy et al., 2016) if the aim is one of inducing long-term plasticity in either the primary visual cortex or higher order visual areas. Despite these limitations, the results obtained with NRT are of great clinical relevance because they provide evidence of perceptual learning transfer from the trained elementary stimuli to the more complex tasks that patients have to perform in everyday life.

References

Ahissar

, & Hochstein

(1996) Learning pop-out detection: Specificities to stimulus characteristics. Vision research, 36(21), 3487–3500. https://doi.org/10.1016/0042-6989(96)00036-3

Ajina

, Kennard

, Rees

, & Bridge

(2014) Motion area V5/MT+ response to global motion in the absence of V1 resembles early visual cortex. Brain awu328. https://doi.org/10.1093/brain/awu328

Azzopardi

, & Cowey

(1997) Is blindsight like normal, near-threshold vision? Proceedings of the National Academy of Sciences, 94(25), 14190–14194.

Berlucchi

(2014) Visual interhemispheric communication and callosal connections of the occipital lobes. Cortex, 56, 1–13. https://doi.org/10.1016/j.cortex.2013.02.001

Blakemore

C.T.

, & Campbell

F.W.

(1969) On the existence of neurones in the human visual system selectively sensitive to the orientation and size of retinal images. The Journal of physiology, 203(1), 237–260. 10.1113/jphysiol.1969.sp008862

Brainard

D.H.

(1997) The psychophysics toolbox. Spatial vision, 10, 433–436. https://doi.org/10.1163/156856897X00357

Bridge

, Thomas

, Jbabdi

, & Cowey

(2008) Changes in connectivity after visual cortical brain damage underlie altered visual function. Brain, 131(6), 1433–1444. https://doi.org/10.1093/brain/awn063

Casco

, Guzzon

, Moise

, Vecchies

, Testa

, & Pavan

(2014) Specificity and generalization of perceptual learning in low myopia. Restorative Neurology and Neuroscience, 32(5), 639–653. 10.3233/RNN-140389

Cowey

(2010) The blindsight saga. Experimental brain research, 200(1), 3–24. 10.1007/s00221-009-1914-2

10.

Das

, & Huxlin

K.R.

(2010) New approaches to visual rehabilitation for cortical blindness: Outcomes and putative mechanisms. The Neuroscientist. https://doi.org/10.1177/1073858409356112

11.

Dilks

D.D.

, Serences

J.T.

, Rosenau

B.J.

, Yantis

, & McCloskey

(2007) Human adult cortical reorganization and consequent visual distortion. Journal of Neuroscience, 27(36), 9585–9594. https://doi.org/10.1523/JNEUROSCI.2650-07.2007

12.

Dougherty

R.F.

, Ben-Shachar

, Bammer

, Brewer

A.A.

, & Wandell

B.A.

(2005) Functional organization of human occipital-callosal fiber tracts. Proceedings of the National Academy of Sciences of the United States of America, 102(20), 7350–7355. 10.1073/pnas.0500003102

13.

Eysel

U.T.

, Scweigart

, Mittmann

, Eyding

, Qu Y. Vandesande

, Orban

, & Arckens

(1999) Reorganization in the visual cortex after retinal and cortical damage. Restorative neurology and neuroscience, 15(2, 3), 153–164.

14.

Furmanski

C.S.

, Schluppeck

, & Engel

S.A.

(2004) Learning strengthens the response of primary visual cortex to simple patterns. Current Biology, 14(7), 573–578. https://doi.org/10.1016/j.cub.2004.03.032

15.

Gilbert

C.D.

, & Wiesel

T.N.

(1992) Receptive field dynamics in adult primary visual cortex. Nature, 356(6365), 150. 10.1038/356150a0

16.

Goebel

, Muckli

, Zanella

F.E.

, Singer

, & Stoerig

(2001) Sustained extrastriate cortical activation without visual awareness revealed by fMRI studies of hemianopic patients. Vision Research, 41(10), 1459–1474. https://doi.org/10.1016/S0042-6989(01)00069-4

17.

Grassi

, & Soranzo

(2009) MLP: A MATLAB toolbox for rapid and reliable auditory threshold estimation. Behavior Research Methods, 41(1), 20–28. https://doi.org/10.3758/BRM.41.1.20

18.

Green

D.M.

(1990) Stimulus selection in adaptive psychophysical procedures. J Acoust Soc Am, 87(6), 2662–2674. https://doi.org/10.1121/1.399058

19.

Halko

M.A.

, Datta

, Plow

E.B.

, Scaturro

, Bikson

, & Merabet

L.B.

(2011) Neuroplastic changes following rehabilitative training correlate with regional electrical field induced with tDCS. Neuroimage, 57(3), 885–891. https://doi.org/10.1016/j.neuroimage.2011.05.026

20.

Henriksson

, Raninen

, Näsänen

, Hyvärinen

, & Vanni

(2007) Training-induced cortical representation of a hemianopic hemifield. Journal of Neurology, Neurosurgery & Psychiatry, 78(1), 74–81. http://dx.doi.org/10.1136/jnn2006.099374

21.

Huxlin

K.R.

, Martin

, Kelly

, Riley

, Friedman

D.I.

, Burgin

W.S.

, & Hayhoe

(2009) Perceptual relearning of complex visual motion after V1 damage in humans. Journal of Neuroscience, 29(13), 3981–3991. https://doi.org/10.1523/JNEUROSCI.4882-08.2009

22.

Julkunen

, Tenovuo

, Vorobyev

, Hiltunen

, Teräs

, Jääskeläinen

S.K.

, & Hämäläinen

(2006) Functional brain imaging, clinical and neurophysiological outcome of visual rehabilitation in a chronic stroke patient. Restorative neurology and neuroscience, 24(2), 123–132.

23.

Kaas

J.H.

, Krubitzer

L.A.

, Chino

Y.M.

, Langston

A.L.

, Polley

E.H.

, & Blair

(1990) Reorganization of retinotopic cortical maps in adult mammals after lesions of the retina. Science, 248(4952), 229. 10.1126/science.2326637

24.

Karni

, & Sagi

(1991) Where practice makes perfect in texture discrimination: Evidence for primary visual cortex plasticity. Proceedings of the National Academy of Sciences, 88(11), 4966–4970.

25.

Kasten

, Poggel

D.A.

, Müller-Oehring

, Gothe

, Schulte

, & Sabel

B.A.

(1999) Restoration of vision II: Residual functions and training-induced visual field enlargement in brain-damaged patients. Restorative Neurology and Neuroscience, 15(2, 3), 273–287.

26.

Kasten

, Bunzenthal

, & Sabel

B.A.

(2006) Visual field recovery after vision restoration therapy (VRT) is independent of eye movements: An eye tracker study. Behavioural Brain Research, 175(1), 18–26. https://doi.org/10.1016/j.bbr.2006.07.024

27.

Kasten

, Wüst

, Behrens-Baumann

, & Sabel

B.A.

(1998) Computer-based training for the treatment of partial blindness. Nature medicine, 4(9), 1083–1087. 10.1038/2079

28.

Maniglia

, Pavan

, Cuturi

L.F.

, Campana

, Sato

, & Casco

(2011) Reducing crowding by weakening inhibitory lateral interactions in the periphery with perceptual learning. PLoS one, 6(10), e25568. https://doi.org/10.1371/journal.pone.0025568

29.

Maniglia

, Pavan

, Sato

, Contemori

, Montemurro

, Battaglini

, & Casco

(2016) Perceptual learning leads to long lasting visual improvement in patients with central vision loss. Restorative Neurology and Neuroscience, 34(5), 697–720. 10.3233/RNN-150575

30.

Marshall

R.S.

, Ferrera

J.J.

, Barnes

, Zhang

, O’Brien

K.A.

, Chmayssani

, Hirsch

, & Lazar

R.M.

(2008) Brain activity associated with stimulation therapy of the visual borderzone in hemianopic stroke patients. Neurorehabilitation and Neural Repair, 22(2), 136–144. https://doi.org/10.1177/1545968307305522

31.

Maunsell

J.H.

, & Newsome

W.T.

(1987) Visual processing in monkey extrastriate cortex. Annual Review of Neuroscience, 10(1), 363–401. https://doi.org/10.1146/annurev.ne.10.030187.002051

32.

Merigan

W.H.

, & Maunsell

J.H.

(1993) How parallel are the primate visual pathways?. Annual Review of Neuroscience, 16(1), 369–402. https://doi.org/10.1146/annurev.ne.16.030193.002101

33.

Mueller

, Mast

, & Sabel

B.A.

(2007) Recovery of visual field defects: A large clinical observational study using vision restoration therapy. Restorative neurology and neuroscience, 25(5, 6), 563–572.

34.

Nelles

, De Greiff

, Pscherer

, Forsting

, Gerhard

, Esser

, & Diener

H.C.

(2007) Cortical activation in hemianopia after stroke. Neuroscience Letters, 426(1), 34–38. https://doi.org/10.1016/j.neulet.2007.08.028

35.

Papageorgiou

, Papanikolaou

, & Smirnakis

(2014) A Systematic Approach to Visual System Rehabilitation—Population Receptive Field Analysis and Real-time Functional Magnetic Resonance Imaging Neurofeedback Methods. In Advanced Brain Neuroimaging Topics in Health and Disease-Methods and Applications. In InTech. 10.5772/58258.

36.

Pelli

D.G.

(1997) The VideoToolbox software for visual psychophysics: Transforming numbers into movies. Spatial vision, 10(4), 437–442. 10.1163/156856897X00366

37.

Perez

, Peyrin

, Cavézian

, Coubard

, Caetta

, Raz

, & … Gout

(2013) An fMRI investigation of the cortical network underlying detection and categorization abilities in hemianopic patients. Brain Topography, 26(2), 264–277. https://doi.org/10.1007/s10548-012-0244-z

38.

Pessoa

, & Adolphs

(2010) Emotion processing and the amygdala: From a’low road’to’many roads’ of evaluating biological significance. Nature Reviews Neuroscience, 11(11), 773–783. https://doi.10.1038/nrn2920

39.

Plow

E.B.

, Obretenova

S.N.

, Halko

M.A.

, Kenkel

, Jackson

M.L.

, Pascual-Leone

, & Merabet

L.B.

(2011) Combining visual rehabilitative training and noninvasive brain stimulation to enhance visual function in patients with hemianopia: A comparative case study. PM&R, 3(9), 825–835. https://doi.org/10.1016/j.pmrj.2011.05.026

40.

Plow

E.B.

, Obretenova

S.N.

, Fregni

, Pascual-Leone

, & Merabet

L.B.

(2012) Comparison of visual field training for hemianopia with active versus sham transcranial direct cortical stimulation. Neurorehabilitation and Neural Repair, 26(6), 616–626. https://doi.org/10.1177/1545968311431963

41.

Raninen

, Vanni

, Hyvärinen

, & Näsänen

(2007) Temporal sensitivity in a hemianopic visual field can be improved by long-term training using flicker stimulation. Journal of Neurology, Neurosurgery & Psychiatry, 78(1), 66–73. https://doi.10.1136/jnn2006.099366

42.

Sabel

B.A.

, Kenkel

, & Kasten

(2004) Vision restoration therapy (VRT) efficacy as assessed by comparative perimetric analysis and subjective questionnaires. Restorative Neurology and Neuroscience, 22(6), 399–420.

43.

Sabel

B.A.

, Henrich-Noack

, Fedorov

, & Gall

(2011) Vision restoration after brain and retina damage: The “residual vision activation theory”. Progress in Brain Research, 192(1), 199–262.

44.

Sagi

(2011) Perceptual learning in vision research. Vision Research, 51(13), 1552–1566. https://doi.org/10.1016/j.visres.2010.10.019

45.

Sahraie

, Trevethan

C.T.

, Weiskrantz

, Olson

, MacLeod

M.J.

, Murray

A.D.

, Dijkhuizen

R.S.

, Counsell

, & Coleman

(2003) Spatial channels of visual processing in cortical blindness. European Journal of Neuroscience, 18(5), 1189–1196. 10.1046/j.1460-9568.2003.02853.x

46.

Sahraie

, Trevethan

C.T.

, & MacLeod

M.J.

(2008) Temporal properties of spatial channel of processing in hemianopia. Neuropsychologia, 46(3), 879–885. https://doi.org/10.1016/j.neuropsychologia.2007.11.008

47.

Schmid

M.C.

, Mrowka

S.W.

, Turchi

, Saunders

R.C.

, Wilke

, Peters

A.J.

, Ye

F.Q.

, & Leopold

D.A.

(2010) Blindsight depends on the lateral geniculate nucleus. Nature, 466(7304), 373–377. https://doi.10.1038/nature09179

48.

Schmid

M.C.

, Panagiotaropoulos

, Augath

M.A.

, Logothetis

N.K.

, & Smirnakis

S.M.

(2009) Visually driven activation in macaque areas V2 and V3 without input from the primary visual cortex. PLoS One, 4(5), e5527. https://doi.org/10.1371/journal.pone.0005527

49.

Schwartz

, Maquet

, & Frith

(2002) Neural correlates of perceptual learning: A functional MRI study of visual texture discrimination. Proceedings of the National Academy of Sciences, 99(26), 17137–17142. 10.1073/pnas.242414599

50.

Shapley

, & Perry

V.H.

(1986) Cat and monkey retinal ganglion cells and their visual functional roles. Trends in Neurosciences, 9, 229–235. https://doi.org/10.1016/0166-2236(86)90064-0

51.

Tamietto

, Pullens

, de Gelder

, Weiskrantz

, & Goebel

(2012) Subcortical connections to human amygdala and changes following destruction of the visual cortex. Current Biology, 22(15), 1449–1455. https://doi.org/10.1016/j.cub.2012.06.006

52.

Wilson

H.R.

, Mcfarlane

D.K.

, & Phillips

G.C.

(1983) Spatial frequency tuning of orientation selective units estimated by oblique masking. Vision Research, 23(9), 873–882. https://doi.org/10.1016/0042-6989(83)90055-X