Fig 1.
Schematic overview of the n-back task using images of objects as stimuli.
The 0-, 1-, and 2-back tasks were performed separately in each block.
Fig 2.
Schematic overview of the searchlight-pattern classification.
Similar classifications were conducted in the HU and non-HU groups separately.
Fig 3.
No change in performance on the 0- to 2-back tasks or on the neurocognitive tests with hydroxyurea treatment.
Reaction time (A), balanced accuracy (B), commission errors (C), and omission errors (D) of the 0-, 1-, and 2-back tasks were measured pre-treatment and at 1-year follow-up. Neurocognitive tests (E) were also measured at both time points. When compared across the hydroxyurea (HU) and non-HU control groups at the 1-year follow-up time point, the balanced accuracy and commission errors of the 0-back task significantly differed. Red and gray bold lines represent the means of the HU and the non-HU control groups, respectively. Light red and light gray markers represent individual data points in the HU and the non-HU control groups, respectively. * p < .05 (Wilcoxon rank-sum test) and # p < .05 (Wilcoxon signed-rank test).
Fig 4.
A significant treatment-by-time interaction was detected in the right cuneus and angular gyrus (labeled in red) for 2- >0-back contrast.
The brain regions activated by the n-back tasks using objects as stimuli are labeled in green. Blue lines mark the Montreal Neurological Institute (MNI) coordinate [32, −64, 25].
Fig 5.
More clusters with changes in brain activation between time points during the 2-back task were identified in the non-HU control group than in the HU group.
Clusters with higher predictive power of time points were identified using a classification accuracy threshold of >.65 and a cluster-size threshold of >20 voxels. These clusters had larger differences in 2-back BOLD signals between time points. In the non-HU group, we identified 27 clusters, with most larger clusters localizing in the posterior region In the HU group, we identified 7 clusters with only 2 larger clusters localizing in the frontal region.
Fig 6.
Anatomical locations of the 9 selected clusters in the transverse plane.
Clusters with peak accuracy ≥0.7 and volume ≥520 mm3 were selected for downstream analysis. The respective main region in the brain (the highest percentage coverage in the AAL Atlas) for each cluster is shown underneath. The details of these clusters are reported in S2 and S3 Tables. The crossed lines mark the peaks of the clusters.
Fig 7.
Increased BOLD signals during working memory processing as a function of time points and task difficulty.
Nine clusters were selected from all the clusters identified via searchlight analyses. (A) Six selected clusters in the non-HU control group had higher brain activation during working memory processing at 1-year follow-up than at pre-treatment. However, only 1 cluster had higher brain activation in the HU group at 1-year follow-up compared to pre-treatment. * p < .05 [Wilcoxon signed-rank test within the HU group (red) or the non-HU control group (gray)]. (B) Increased BOLD signals were observed in most selected clusters in both groups, as the working memory load increased from 0- to 1-back. In the non-HU control group, BOLD signals did not increase further when the working memory load increased from 1- to 2-back. In the HU group, however, BOLD signals increased continuously as the working memory load increased from 0- to 2-back. # p < .05 (Wilcoxon signed-rank test).