Letter cancellation test (LCT) variants are widely used pen-and-paper assessment tools in clinical and experimental psychology, but brain regions that mediate LCT performance are not well understood. An fMRI study involving elderly healthy volunteers was conducted to establish the neural correlates of the LCT using a highly novel fMRI-compatible tablet system that enables investigation of drawing behavior. The resultant brain activation highlighted parietal and frontal regions, consistent with existing fMRI literature on visual attention. This is the first fMRI study of the LCT and the results have relevance for future clinically-oriented fMRI studies of this test.
Fifteen healthy subjects (8 males; mean age 64.9 ± 1.5 years) participated in this ethically approved fMRI study and were imaged using a Siemens Skyra 3 T MRI system using our standard lab protocol4. Prior to fMRI, subjects performed the paper LCT and their completion time was recorded. A brief training session of writing and drawing tasks followed so that subjects could become proficient using the tablet in the magnet. Subjects then performed the tablet LCT during fMRI for a unique array of letters in five separate runs, each involving a different target letter, with instructions to cross out (as fast as possible while maintaining accuracy) all instances of the target letter in the array provided. Each LCT task block had a fixed duration of 60 s. Task blocks were separated by 20 s periods of visual fixation. Figure 1 shows the tablet setup and illustrative LCT responses of a single subject during fMRI5.
The fMRI data were analyzed using Optimization of Preprocessing Pipelines for NeuroImaging (OPPNI) methodology6,7,8, Analysis of Functional NeuroImages (AFNI)9 and MRIcron10. Statistical parametric maps were calculated using a General Linear Model (GLM) and thresholded using the two-tailed False-Discovery Rate (FDR) to correct for multiple comparisons11.
Subjects made both errors of omission (target letters not crossed out) as well as commission (non-target letters crossed out) when performing the tablet LCT. Overall, there were 3.2 ± 0.5 omissions (group mean ± SEM) and 1.9 ± 0.4 commissions, a statistically significant difference (p<0.01). Because not all subjects completed all tablet LCTs within the fixed block duration, comparison of errors and completion time was not attempted between paper and tablet versions of the LCT. Performance was compared instead using a “seconds per hit” metric. Figure 2 shows individual data as well as box and whisker plots of the behavioural results for both LCT versions, indicating that the subjects performed the tablet LCT significantly more slowly (p<0.001). The mean paper LCT performance was 0.98 ± 0.05 seconds per hit, and the mean tablet LCT performance was 1.95 ± 0.15 seconds per hit. A Spearman’s rank-order correlation with bootstrapping showed no significant correlation between subject performance on paper and on the tablet (rs = 0.20, p = 0.48).
Figure 3 shows the statistical brain maps of selected slice locations for the LCT vs. fixation contrast. Activity primarily involved the cerebellum, occipital areas, middle temporal lobe, precentral gyrus, postcentral gyrus, frontal gyrus, insula, cingulate gyrus, supramarginal gyrus, parietal lobule, and supplementary motor area.
Discussion
Use of the tablet affected the ability of subjects to perform the LCT compared to performance on paper. For example, the tablet resulted in elevated and more variable time to strike out letters on a “per hit” basis. The differences between paper and tablet results could arise from multiple factors such as intrinsic subject variability in task performance, and subject variability in ease of interactivity with the tablet. Regarding brain activation, the letter vs. fixation contrast activated various parietal regions of the brain, consistent with lesion studies in which patients were impaired in cancellation tasks1. This suggests that our results include activation that is crucial to LCT performance. The frontal and parietal regions that showed activation were also consistent with regions known to be involved in spatial guidance towards attentional targets12.