To test for the feasibility of our training procedure, we conducted a pilot study involving 5 subjects.
Criteria used for inclusion of subject were the following: (1) presence of a bilateral AMD condition stable over the last year (table 1); (2) visual acuity ranging between 1.5 logMAR and 0.9 logMAR in the best eye; and (3) complaints of reading disturbances consisting essentially of gross difficulty to decipher long words. Six subjects were initially enrolled in the study (MT, SL, LG, AA, GM, ML). Subject ML, however, was subsequently discarded, as he showed severe signs of memory impairment three months following the training. His statement that he did not remember to have been trained in our laboratory emphatically demonstrated his memory loss. Ages ranged from 76 to 81 years (mean: 78.8 years ± 1.92 SD). In each subject a stable bilateral macular condition had been present in average for 2.3 years. Contrast sensitivity of subjects' best eye, as measured with the Pelli-Robson contrast sensitivity chart (Clement Clarke Intl, Haag-Streit, UK) was lower than the one measured in subjects without macular disorder of similar age . These clinical features are shown in table 1. A regular low-vision rehabilitation procedure, including scanning and reading exercises similar to those described by Beaunoir et al. (2000) , had been conducted by a low vision specialist, for four to ten training sessions, extending over a period of more than three months. Despite rehabilitation, subjects presented reading difficulties during the SLO evaluation as for example reading only the first five letters of a ten letter-word, or guessing, often incorrectly the end of presented words. In general they were unable to uncover the end of words (table 3). Stability of the clinical condition and previous rehabilitation training allowed the use of each subject as his own control, as no spontaneous recovery was expected.
All subjects gave their written informed consent for the procedure and testing methods, all of which were in accordance with the Declaration of Helsinki and the study had been accepted by the ethical committee for human experimentation of Geneva University Hospitals.
Reading strategies for isolated words and paragraphed text
The results of the initial evaluation step showed that for reading isolated words and paragraphed text four (MT, SL, LG, AA) out of five subjects had spontaneously developed an initial PRL that fulfilled the needs for detailed discrimination capacity although it prevented global viewing. Indeed, the PRL showed better visual acuity than the examiner's selected TRL, but was not optimally located to allow the viewing of the whole word, thus altering reading ability (figure 3A, 4, 5A).
Three subjects (MT, SL, LG) had an initial PRL to the left of the scotoma and consequently ignored the end of the words (figure 3A, 5A and 6A). Two of these subjects (SL, LG) were additionally unable to perform search movements to detect single letters and isolated words initially projected within the scotoma, and to relocate the projected image onto a healthy peripheral area. The third subject (MT) showed an additional small scotoma in the vicinity of the left border of the central scotoma on the SLO picture that repeatedly superimposed on parts of the scrutinized characters (figure 1, 5). We trained these three subjects (MT, SL, LG) to use an examiner's selected TRL above the scotoma on the SLO image (i.e. below the scotoma in the visual field).
The fourth subject (AA) had an initial PRL located above the scotoma on the SLO image, but too close to the scotoma's border, the upper part of the letters scrutinized being often projected into the scotoma. Consequently he often missed the center of presented words. We trained him to use an examiner's selected TRL located on the same side of the scotoma as the initial PRL, i.e. above the lesion on the SLO image, but farther from the scotoma's border (figure 1).
The remaining fifth subject (GM) had not developed a PRL despite previous rehabilitation. He was found to use an ill-defined area with preserved global viewing but non-optimal spatial resolution. In addition, he presented a highly instable fixation.
Based on the results of visumetry, he was trained to use an area below the scotoma on the SLO picture, which presented a better acuity than the initial PRL while still preserving global viewing capacity (figure 1).
Immediately after SLO training, analyses of reading strategies revealed that four subjects (SL, LG, AA, GM) used the examiner's selected TRL, or additional newly self-selected PRLs (figure 3B and 6B) alone or in combination with their initial PRL for reading isolated words and paragraphed text. One of these subjects (GM) (no defined PRL before SLO training) used the examiner's selected TRL alone. The remaining fifth subject (MT) showed an improved ability to scan the word but only used the initial PRL (figure 5B).
Three months following SLO training, four subjects (SL, LG, AA, GM) consistently used the examiner's selected TRL either in isolation (GM), or in combination with the initial PRL (LG, AA, SL) for reading both isolated words and paragraphed texts (table 3, figures 4 and 7). The use of newly developed self-selected PRLs was observed in three subjects (SL, AA, GM) to read isolated words and in two subjects to read paragraphed text (MT, SL). Interestingly, one subject (MT) lost the ability to perform eye movements to scan the entire words, but kept some adaptation, i.e. used newly self-selected PRLs, during text reading (table 2).
ETDRS visual acuity
ETDRS values are reported in table 2 and figure 8. Visual acuity evolution across the three evaluation periods was analyzed using a one-way repeated measure ANOVA. This analysis yielded a significant difference (F(2,3) = 17.85; p = 0.022). Values measured before and three months following the SLO training significantly improved compared to those measured before the training process (Bonferroni pairwise comparison p = 0.012), while the difference was marginally significant between acuities measured before and immediately after SLO training (Bonferroni pairwise comparison p = 0.085). However when considering individual data, the majority of improvements were no greater than the test repeatability of 0.2 [34, 35].
Threshold character size for single letters and isolated words
Threshold character sizes for single letters and isolated words are shown in table 2 and figure 9. A three-way repeated measure ANOVA with subjects (MT, SL, LG, AA, GM), stimulus types (single letters, 3 and 4 letter-words, 6 and 7 letter-words and 9 and 10 letter-words) and evaluation periods (before, at the end of training and three months later) was conducted to explore the impact of the training procedure as measured by the threshold character size.
There was a highly significant main effect of the evaluation periods (F(2, 19) = 46.04; p < 0.0005). The post-hoc analysis using Bonferoni correction showed that all subjects benefited from the training procedure both, immediately after SLO training (mean ± standard error: before = 1.255 ± 0.017, immediately after SLO training = 1.055 ± 0.016; p < 0.0005) and three months later (1.084 ± 0.014; p < 0.0005). There was, however, no significant evolution after completion of the training procedure (p = 0.563).
There was also a main effect of stimulus type confirming that threshold character size varied as a function of word length (F(3, 20) = 73.921; p < 0.0005). Threshold character size for single letters (0.9 ± 0.019) was lower than for any other word lengths and threshold for 3 and 4 letter-words (1.13 ± 0.019) was lower than those for 6 and 7 (1.23 ± 0.019) and 9 and 10 letter-words (1.26 ± 0.019; Newman-Keuls: p < 0.05). Additionally, the retention of the training gains varied with stimulus types (interaction evaluation periods and stimulus type: F(6, 38) = 3.135; p = 0.014). To further investigate the retention of the training gains on the different stimulus types we conducted a post-hoc analysis using Bonferroni adjustment for multiple comparisons with α = 0.00125. Subjects read significantly smaller character sizes immediately after the end of the SLO training procedure at all word lengths but 3/4 letter-words, despite the fact that a slight improvement was noticeable for the latter (1 letter: before = 1.02 ± 0.029, immediately after SLO training = 0.82 ± 0.022; p = 0.006; 6 and 7-letter words: before = 1.41 ± 0.062, immediately after SLO training = 1.15 ± 0.027; p = 0.008; 9 and 10-letter words: before = 1.38 ± 0.057, immediately after SLO training = 1.13 ± 0.045; p = 0.006; 3 and 4-letter words: before = 1.21 ± 0.062, immediately after SLO training = 1.10 ± 0.045; p = 0.513). Three months following the training procedure, gains were retained for single letters and 6 and 7-letter words, as shown by the non-significant difference between results obtained at the end of training and three months later (1 letter: three months later = 0.86 ± 0.016; p = 1.0; 6 and 7 letter-words: three months later = 1.13 ± 0.056; p = 1.0). In contrast, performances were not retained for 9 and 10-letter words (three months later = 1.28 ± 0.057; p = 0.02) and returned to the initial score (p = 0.32). For 3 and 4-letter words the statistical test did not reveal any additional change over the whole procedure either (immediately after SLO training – three months later: p = 1.0: before – three months later: p = 0.456).
Another main result was that subjects benefited differently from the training procedure, as shown by the interaction between evaluation periods and subjects (F(8, 38) = 5.371; p < 0.0005). Immediately after the SLO training procedure, all subjects had improved their performances regardless of the type of stimulus. Three months later subjects SL, LG and AA totally retained and even slightly improved, training gains. Subject GM did not fully preserve the improvements, even if his performances were far better than before the SLO training. Subject MT did not retain training gains at all (figure 9). Despite the fact that the interaction subjects, stimulus type and evaluation periods was not significant (F(24; 38) = 0.1014; p = 0.497), it is interesting to describe individual behavior. When looking at individual data (figure 9), it appears that all subjects had better performances three months following the training procedure than before for 3 and 4-letter words. For 6 and 7-letter words, three subjects (SL, LG, AA) progressed in deciphering capacities, one subject (GM) retained training benefits, and the remaining subject (MT) partially lost training gains. For the 9 and 10-letter words, one subject (SL) improved, three subjects (LG, AA, GM) partially lost the gains, and the remaining subject (MT) lost all benefits of the SLO training (see arrows in figure 9). Finally there was a main effect of subjects, as expected regarding the individual variations of threshold character sizes (F(4, 20) = 16.443; p < 0.0005). The interaction subject and stimulus type was not significant (F(12; 20) = 1.100; p = 0.087).
Paragraphed text reading
Threshold character size for paragraphed text and percentage of words read are shown in table 2 and figure 10. Immediately following SLO training, three subjects (SL, AA, GM) became able to read paragraphed text one character size smaller (0.1 logMAR) and one subject (LG) could read two character sizes smaller (0.2 logMAR) than evaluated before training (table 2). In all subjects, the proportion of words correctly read at the first attempt, i.e. reading accuracy, increased with training (figure 9). Three months following the completion of the SLO training procedure, the smallest readable character size for paragraphed text was similar to that of the initial evaluation in four subjects (MT, SL, AA, GM), but the percentage of words correctly read had increased (figure 9) allowing subjects to have an adequate comprehension on smaller character sizes considering that an accuracy of 85% allows a complete understanding of the text (27).
Patients' comments on the modification of their visual perception before and after the training
We asked the patients to give a subjective evaluation of the influence of the SLO training on their everyday life. All patients reported less visual fatigue and improved awareness of the location of the scotoma. MT reported that two weeks after the completion of the SLO training, she had the impression of an improvement in reading. SL noticed that SLO training enabled a more efficient reading of the panels in the street as well as bus numbers. LG could not report specific subjective improvements but said that, in general, she better coped with her visual field defect. AA applied the taught strategy to other visual tasks. She said she recovered the ability to hand-knit. After the SLO training process, GM started to use a closed circuit television (CCTV) and was able to read with it. He stated that the SLO training allowed him to locate his scotoma in his visual field and, as a result, several visual tasks were easier to accomplish.