Visual loss and getting lost in Alzheimer’s disease

Subject selection.

We studied AD (n = 11), elderly normal (EN; n = 12), and young normal (YN; n = 6) subjects in psychophysical experiments after evaluating their primary visual and neuropsychological functioning. AD subjects (age range, 60 to 82 years; mean age, 73) were recruited from the clinical programs of the University of Rochester Alzheimer’s Disease Center and had probable AD by National Institute of Neurological and Communicative Disorders and Stroke–Alzheimer’s Disease and Related Disorders Association criteria.8 EN subjects (age range, 63 to 81 years; mean age, 72) were in programs for the healthy elderly or were the spouses of AD subjects. YN subjects (age range, 19 to 25 years; mean age, 21) were undergraduates. Informed consent was obtained from all subjects. All protocols were approved by the University Human Subjects Review Board.

Each subject was interviewed and examined to exclude neurologic or ophthalmologic disorders other than AD. All subjects had normal, or corrected to normal, visual acuity by Snellen testing (left eye, right eye [mean ± SD]: YN = 13.7 ± 1.0, 13.7 ± 3.3; EN = 19.1 ± 7.3, 19.3 ± 7.8; AD = 23.3 ± 10.6, 35.0 ± 25.1) and full visual fields by confrontation testing. Contrast sensitivity profiles were tested at five spatial frequencies (0.5–18 cycles/^o, VisTech Consultants, Inc., Dayton, OH) and were in the normal range for all groups; the YNs performed better but there were no significant differences between the ENs and ADs (mean ± SD: YN = 21.7 ± 14.0; EN = 25.8 ± 14.7; AD = 31.1 ± 9.3).

Neuropsychological testing.

Neuropsychological tests were conducted on all subjects, including the Mini-Mental State Examination (MMSE),9 the North American Adult Reading test, the Boston Naming test, the Controlled Oral Word Association (F/A/S) test, the Rey Auditory Verbal Learning test (scored on fifth recall trial), clock drawing (scored by the method of Rouleau10), the grooved pegboard test (time to complete two rows with the dominant hand), and the Money Road Map test.11 The YN and EN groups were normal for age performance on all tasks. The AD group showed uniform mild impairment consistent with the diagnosis of early AD (table).

In addition, we developed an open-field test of visuospatial orientation for spatial navigation. Subjects were escorted from the hospital lobby to the laboratory, having been told that they would be asked questions about the route when they arrived in the laboratory. This route consisted of walking down four corridors, making four turns, and passing a variety of distinct landmarks. The test consisted of 12 questions that included recognizing the layout of the path, the relative length of each corridor, the direction of turns, and pictures of landmarks.

Visual psychophysical testing.

Tests of visual discrimination and visual self-movement interpretation both used the same laboratory configuration. Subjects sat in front of a large-screen computer display and were required to press one of two buttons on every trial. They were 4 feet away from an 8 × 6-foot rear projection tangent screen that covered the central 90° × 74° of the visual field. Centered visual fixation was maintained on a red light emitting diode (LED) image (0.2°). Eye position (electro-oculogram [EOG]) and head position were monitored during all trials so that gaze shifts beyond the central segment of the screen aborted that trial. We restricted gaze to the central 20° in discrimination trials and the central 30° in self-movement interpretation trials so that all foci of expansion in optic flow stimuli were included in the gaze control area. All subjects tended to maintain centered fixation on the LED image in all trials.

Trials began with an audible tone indicating that central fixation was required within 1 second. After fixation was established, a visual stimulus was presented and was followed by a pair of tones to prompt the subject’s response. The response buttons were then illuminated and the subject was given up to 8 seconds to press the button corresponding to the response chosen for that trial.

Subjects were trained on each task by presenting high coherence stimuli (patterns that were not obscured by superimposed random dots) to test their ability to see those patterns, understand the task, and respond appropriately. Two AD subjects were excluded from the study because they could not perform accurately in the training session. A third AD subject was excluded because button press errors interfered with all of his responses, even with high coherence stimuli.

Visual stimuli were generated off-line and presented by a personal computer driving a television projector (Electrohome 4100, Electrohome Limited, Kitchener, Ontario) to create a 90° × 60° image centered at eye height. The stimuli consisted of 500 white dots (2.69 cd/m²) on a black background in an animated sequence of frames presented at 60 Hz. Dot positions were specified for each frame by algorithms for each type of display.12 All stimuli had the same dot density, luminance, and contrast. All moving dot stimuli had the same average dot speed.

Visual discrimination threshold tests.

We obtained visual discrimination thresholds using optic flow stimuli superimposed on random motion. Visual discrimination thresholds for static shapes on a cluttered background, and for horizontal motion on random motion, were obtained to provide measures of more fundamental visual capacities commonly tested in psychophysical research. In these studies, subjects were instructed to push the button labeled as corresponding to the preceding stimulus.

In each discrimination task, we presented a pseudorandom sequence of stimuli with twelve repetitions at each coherence level. We chose this method of constant stimuli, rather than more efficient adaptive methods, to obtain full psychometric functions for every subject. The order of presentation for the three stimulus sets was counterbalanced so that approximately equal numbers of subjects in each group did each discrimination task first, second, or third.

We measured visual discrimination thresholds to three different types of stimuli. 1) In the shape discrimination task, dots forming the outline of a circle or a square were centered on the viewing screen and presented with randomly placed stationary dots. Between 1.2% and 11% of the dots were in the shape pattern, each stimulus was presented for 250 msec, and subjects chose whether the imbedded pattern formed a circle or a square (figure 2A). 2) In the horizontal motion discrimination task, leftward or rightward moving dots were superimposed on various numbers of dots moving in random directions. Random dot movement was created by randomly assigning each dot, in each frame, either to the patterned motion or to the random motion. Between 1% and 43% of the dots were in the horizontal motion pattern, each stimulus was presented for 750 msec, and subjects chose whether the motion was toward the left or the right (see figure 2B). 3) In the optic flow discrimination task, outward radial patterns with a focus of expansion 15° to the left or right of center were superimposed on various numbers of dots moving in random directions. Between 1% and 43% of the dots were in the radial pattern, each stimulus was presented for 750 msec, and subjects chose whether the focus of expansion was on the left or the right (see figure 2C).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. The visual discrimination stimuli used in these studies. Each stimulus consisted of a coherent pattern containing a given percentage of the total number of dots on the screen, the remaining dots being randomly distributed. (A) Static shape stimuli consisted of the outlines of circles or squares formed by dots and superimposed on randomly placed stationary dots. (B) Horizontal motion stimuli contained either leftward or rightward moving dots superimposed on randomly moving dots. (C) Radial optic flow stimuli consisted of an outward radial pattern with a focus of expansion 15° to the left or right of center with superimposed randomly moving dots.

Visual self-movement interpretation tests.

We tested self-movement interpretation based on optic flow using radial stimuli without added random motion. In this task the optic flow was made more readily distinguishable by presenting foci of expansion at the usual position ±15° from the center of the display, and at the more extreme positions of ±30°. Each stimulus was presented for 750 msec, then four arrows appeared on the screen pointing in the directions of movement simulated by the stimuli. Subjects were instructed to push one of four buttons positioned to correspond to the perceived direction of simulated self-movement. In the self-movement interpretation task, each of the four directions was presented six times.

Data analysis.

In tests of visual discrimination, the percentage of dots in the pattern ([pattern dots/pattern plus random dots] × 100) was related to the percentage of correct responses obtained from each subject. These values were fit to a cumulative probability function (Probit function) to derive a perceptual discrimination threshold. Threshold performance was considered the percentage of pattern dots at which subjects achieved 75% correct responses. In tests of self-movement interpretation, all stimuli were presented at 100% coherence and the results were scored as the average percentage of correct responses. Statistical analyses were conducted using the SAS statistical package (SAS Institute, Cary, NC).13

Visual discrimination thresholds.

A selective impairment in optic flow discrimination was seen in AD subjects (figure 3). This was confirmed by a two-way analysis of variance (ANOVA) with a significant group-by-task interaction effect (F(4,52) = 6.53, p < 0.0002). Follow-up linear contrasts determined that the source of the interaction effect was a large difference between horizontal motion and optic flow thresholds in the AD group (ANOVA: F(1,20) = 15.4, p < 0.01) but not in the YN or EN groups. Thus, the first result of this study was that AD patients are selectively impaired in the optic flow discrimination task.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3. Visual discrimination thresholds from the three stimulus sets in the three subject groups. Discrimination thresholds (ordinate) are plotted as the percentage of stimulus dots that had to be in the coherent pattern for a subject to identify the target pattern correctly in 75% of the trials (50% correct is chance performance). Data are presented as average thresholds ± the standard error of the mean for each subject group (abscissa): young normals (n = 6), elderly normals (n = 12), and Alzheimer’s disease (n = 11) subjects. Thresholds for the three discrimination tasks are plotted: static shape (squares), horizontal motion (circles), and radial optic flow (diamonds) stimuli. A significant task-by-group interaction effect was attributable to higher optic flow thresholds in the AD subjects.

In addition, there was a common trend across all three discrimination tasks with significantly larger thresholds from the AD group (one-way ANOVAs for shape F(2,26) = 9.85, p < 0.007; horizontal motion F(2,26) = 7.27, p < 0.003; optic flow F(2,26) = 12.9, p < 0.0001). All three tasks showed the same pattern of differences such that AD > EN = YN (Tukey’s studentized range [HSD], p < 0.05). There were no significant differences attributable to the order in which static shape, horizontal motion, and optic flow tasks were presented.

To see if differences between horizontal motion and radial optic flow perception were unique to subjects in the AD group, we compared these thresholds for each subject (figure 4). Subjects in all three groups showed substantial variation in the difference between radial and horizontal thresholds (T_R − T_H ± SE: YN = 1.5 ± 3.5, EN = 2.2 ± 2.2, AD = 13.6 ± 4.2), but the AD group consisted of two subgroups: five subjects with approximately equal radial and horizontal thresholds (T_R − T_H ± SE: −0.4 ± 3.9) and six subjects with much larger radial optic flow thresholds than horizontal motion thresholds (T_R − T_H ± SE: 25.3 ± 4.0). Thus, this selective impairment of optic flow discrimination was limited to about half of the AD patients tested.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4. Profile plots of the relationship between horizontal motion and radial optic flow discrimination thresholds. Each subject’s thresholds are connected by a solid line. Young normal and elderly normal subjects showed relatively small differences between their horizontal motion and radial optic flow thresholds (mean ± SE; young = 1.5 ± 3.5, elderly = 2.2 ± 2.2). Alzheimer’s subjects showed substantially larger differences between their horizontal motion and radial optic flow thresholds (mean ± SE = 13.6 ± 4.23). This difference is attributable to the fact that 6 of the 11 subjects (55%) showed much larger radial optic flow thresholds with average differences of 25.3 ± 4.0 (bold lines), whereas the remaining five subjects showed small differences between those thresholds (−0.4 ± 3.9).

Relations between discrimination thresholds and neuropsychological tests.

The two groups of AD subjects, defined by the presence or absence of impaired optic flow perception, showed no significant differences in our battery of standard neuropsychological tests of memory and cognitive function. However, regression analysis showed a significant inverse relationship between optic flow threshold and MMSE score (r = −0.53, p < 0.05). This suggests that patients with more advanced AD were more impaired on optic flow perception.

The open-field test of spatial navigation, regarding the route subjects traveled from the hospital lobby to the laboratory, revealed significant between-group differences (see table 1). Young (n = 7) and elderly (n = 4) normal subjects averaged 83% and 75% correct responses, respectively, whereas the AD group averaged only 42% correct responses (between-group ANOVA F(2,18) = 10.38, p < 0.01).

In AD patients, spatial navigation test scores were significantly correlated with optic flow thresholds, but not with the horizontal motion or shape discrimination thresholds. Poor performance on the spatial navigation test was associated with an elevated optic flow threshold, yielding a best fit line by linear regression with a slope of −0.78 and a correlation coefficient of −0.66 (p < 0.05). Multiple regression showed that optic flow threshold accounted for much more of the variance in spatial navigation scores than did horizontal motion or shape discrimination, with squared semi-partial correlations for optic flow = 0.174, horizontal motion = 0.055, and shape discrimination = 0.001.

We considered whether the spatial navigation scores merely reflected global impairment as measured by the MMSE. However, there was no significant correlation between these tests (r = 0.11). In addition, there was no benefit to adding MMSE scores to optic flow threshold in a regression model predicting spatial navigation scores. These findings suggest that optic flow discrimination thresholds were a significant, independent predictor of spatial navigation performance, supporting the view that this perceptual impairment contributes to difficulties in spatial navigation.

Visual self-movement interpretation.

Optic flow provides cues about self-movement, as well as spatial navigation. We considered that responses to optic flow might depend on whether the subjects were engaged in a task that was explicitly directed at self-movement interpretation. We used optic flow stimuli like those used in the perceptual discrimination tests, but asked our subjects to indicate the simulated direction of self-movement rather than the location of the focus of expansion. We overcame the perceptual limitations of elevated optic flow thresholds by removing the random dots from the stimuli and by presenting foci of expansion at the usual positions of ±15° and also at ±30°. Subjects were asked to respond by indicating the simulated direction of self-movement (figure 5A).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5. Studies of the ability to interpret optic flow as simulating self-movement were conducted in young normal (YN), elderly normal (EN), and AD subjects who were normal (ADN) or impaired (ADI) by optic flow discrimination testing. (A) Radial optic flow stimuli were presented at 100% coherence with foci of expansion at ±15° or ±30°, with subjects choosing which of four directions corresponded to the direction simulated by the stimulus. (B) The percent correct identifications of the direction simulated in the stimulus (ordinate) for each subject group (abscissa) (YN, n = 6; EN, n = 11; ADN, n = 5; ADI, n = 6). The graph shows the percent correct identifications averaged across all four stimuli (chance performance = 25% correct). The YN and EN groups showed good performance whereas the ADN and ADI groups were impaired.

We compiled the results of self-movement interpretation for the YN and EN groups. AD subjects were split into AD normal (ADN) and AD impaired (ADI) subgroups based on their optic flow thresholds. We found significant differences between subject groups (ANOVA: F(3,24) = 26.0; p < 0.0001), with the average percent correct responses decreasing from YN = 94% and EN = 81% to ADN = 54% and ADI = 50% (Tukey’s HSD test showed that YN = EN > ADN = ADI, p < 0.05) (see figure 5B). Thus, young and elderly normal subjects did uniformly well at interpreting optic flow as an indication of their direction of self-movement, and all AD subjects did poorly at self-movement interpretation.

The ADN and ADI groups were similarly impaired on self-movement interpretation, so this impairment was not a function of optic flow discrimination threshold. This is confirmed by the absence of a correlation between self-movement scores and optic flow thresholds in the AD group (r = 0.01). Furthermore, three AD subjects spontaneously complained that the stimuli did not suggest self-movement although they could identify the location of the focus of expansion. This supports the notion that impaired self-movement interpretation is separate from impaired optic flow perception.

Perceptual mechanisms of visuospatial disorientation.

We found a clear association between impaired optic flow perception in AD and the inability to answer questions about a recently traversed route through a complex environment as measured by our open-field test of spatial navigation. This finding supports the possibility that impaired optic flow analysis might contribute to visuospatial disorientation.

The prevalence of impaired optic flow perception in our AD patients (55%) was somewhat higher than previous estimates of the prevalence of visuospatial disorientation in AD (20% to 40%).2,18 This difference may be attributable to greater sensitivity of our psychophysical measures compared to the standard neuropsychological tests of visuospatial function.19 One implication is that tests of visual perception might detect impaired spatial processing before it would be evident in other tests.

Impaired visual self-movement interpretation in AD.

In a separate task, we used readily discriminated optic flow stimuli (100% motion coherence and widely separated foci of expansion) to test our subjects’ ability to interpret optic flow as a self-movement cue. We found a significant impairment of optic flow interpretation in AD subjects (see figure 5) that was not correlated with their optic flow discrimination thresholds or their performance on other visual and neuropsychological tests. Our EN subjects showed a trend toward impaired optic flow interpretation that is consistent with the previous report of impaired heading detection in elderly subjects viewing simulated self-movement.20 However, in our data only the AD group was significantly different from the YN group.

These findings suggest a degree of independence between optic flow perception for spatial navigation and optic flow interpretation as a cue about self-movement. This is highlighted by spontaneous complaints from three AD subjects who could see the optic flow, but could not see how it implied anything about self-movement. The inability to recognize the patterned visual motion of optic flow as a cue about self-movement might be termed visual autokineagnosia. This is consistent with the nomenclature for visual agnosias based on the presence of an interpretive deficit with preserved perception.21 In addition, this term emphasizes that this mechanism might contribute to visuospatial disorientation from other causes including focal cortical lesions or other neurodegenerative diseases. The symptoms of visuospatial disorientation may be more evident in AD because spatial memory impairments in these patients place greater reliance on visual perceptual and interpretive capacities.

Localizing visuospatial disorientation.

Our findings are consistent with the dual pathways model of visual processing embedded in Kleist’s assigning visual object agnosias to occipito-temporal areas and visuospatial agnosias to occipito-parietal areas (Kleist,22 cited in Grusser and Landis23). Ungerleider and Mishkin24 developed this model with experimental evidence for a ventral, occipito-temporal pathway for visual object recognition and a dorsal, occipito-parietal pathway for visuospatial perception, with Goodale and Milner25 emphasizing the model’s behavioral implications.

In the dorsal visual pathway, the middle temporal area (MT) is critical for the perception of uniform patterns of visual motion.26 The adjacent medial superior temporal area (MST) processes optic flow12 and its cues about heading direction27 and environmental structure28 with contributions from other sites along the dorsal pathway.29,30 This suggests that disease activity in area MT might account for elevated horizontal motion thresholds in AD,15,16 whereas the elevated optic flow thresholds found in this study might reflect effects on area MST. Likewise, the separate impairment of optic flow interpretation, seen in some AD subjects, might be attributable to disease effects on higher dorsal visual areas.

Rather than explaining these deficits as the result of focal pathology, we might view them as cortical disconnection syndromes. This is consistent with the selective effects of AD on corticocortical projection neurons31 that might cause greater deterioration of visual motion signals as they pass through the successive stages of the dorsal visual pathway. Thus, increasing impairment in horizontal motion, optic flow, and self-movement perception could reflect the cumulative degradation of visual motion signals so that lower centers are least affected and higher centers are most affected. Thus, visuospatial disorientation in AD might suggest a new model for cortical pathophysiology: the serial disconnection syndrome.