Neural systems predicting long-term outcome in dyslexia

Abstract

Individuals with developmental dyslexia vary in their ability to improve reading skills, but the brain basis for improvement remains largely unknown. We performed a prospective, longitudinal study over 2.5 y in children with dyslexia (n = 25) or without dyslexia (n = 20) to discover whether initial behavioral or brain measures, including functional MRI (fMRI) and diffusion tensor imaging (DTI), can predict future long-term reading gains in dyslexia. No behavioral measure, including widely used and standardized reading and language tests, reliably predicted future reading gains in dyslexia. Greater right prefrontal activation during a reading task that demanded phonological awareness and right superior longitudinal fasciculus (including arcuate fasciculus) white-matter organization significantly predicted future reading gains in dyslexia. Multivariate pattern analysis (MVPA) of these two brain measures, using linear support vector machine (SVM) and cross-validation, predicted significantly above chance (72% accuracy) which particular child would or would not improve reading skills (behavioral measures were at chance). MVPA of whole-brain activation pattern during phonological processing predicted which children with dyslexia would improve reading skills 2.5 y later with >90% accuracy. These findings identify right prefrontal brain mechanisms that may be critical for reading improvement in dyslexia and that may differ from typical reading development. Brain measures that predict future behavioral outcomes (neuroprognosis) may be more accurate, in some cases, than available behavioral measures.

Conflict of interest statement

The authors declare no conflict of interest.

Figures

Fig. 1.

Change in reading skills. Change in single-word reading skills {WRMT Word Identification standard score (WID[ss]); A} and reading comprehension skills {WRMT Passage Comprehension subtest standard score (PC[ss]); B} over time for the dyslexic group that showed more (Reading Gain group, n = 13) or less improvement (No Reading Gain group, n = 12) in reading ability (slope-WID [ss]) over 2.5 y based on median split, and the control group.

Fig. 2.

fMRI and DTI predictors of reading gains in dyslexia. (A) Association between brain activation (rhyme > rest) and future reading improvement. This statistical map shows a region in right (Rt) inferior frontal gyrus (IFG) region where significant positive correlation was found with reading gains (increase in WRMT WID[ss] as a function of time) 2.5 y later in the group with dyslexia. No other significant correlations were found either in the group with dyslexia or controls. (B) Association between right IFG activation and future reading improvement. Individual contrast estimates from entire right inferior frontal cluster in Fig. 1A showed significant positive correlation with reading gains in the group with dyslexia (n = 25; red circles; mean cluster r2 = 0.46), but not in the control group (n = 20; blue circles; mean cluster r2 = 0.016). Subgroups of dyslexic and control participants matched for age are shown with larger hallow ring; regression line for age-matched subgroups are shown in dotted lines (B–D). (C) Association between white matter integrity and future reading improvement. Individual FA values of the Rt superior longitudinal fasciculus (SLF) showed significant positive correlation between gains in reading in the dyslexic group (r2 = 0.27), but not in the control group (r2 = 0.004). Dyslexia group (n = 21, 4 individuals were missing DTI data), red circles and regression line in red. Control group (n = 17, 3 individuals were missing DTI data), blue circles and regression line in blue. (D) Association between DTI white matter integrity and fMRI activation. FA values of Rt SLF and Rt IFG contrast estimates show significant positive correlation in the dyslexia group (r2 = 0.26), but not in the control group (r2 = 0.00002). Dyslexic group (n = 21), red circles and regression line in red. Control group (n = 17), blue circles and regression line in blue.

Fig. 3.

Multivariate pattern classification of reading gains in dyslexia. (A) Classification accuracy using fMRI right inferior frontal gyrus (IFG) activation only (fMRI), DTI right superior longitudinal classification (SLF) only (DTI), combination of the two (Comb), patterns of behavior (Bx), and fMRI whole-brain pattern (fMRI Pattern). Black horizontal lines above the graph show comparisons of interest that are significant (e.g., fMRI pattern showed significantly greater accuracy than Comb). (B) Association between distance from hyperplane for the whole-brain fMRI pattern classifier and gains in reading ability. (C) Brain activation patterns that discriminate between children with dyslexia who gained in reading ability vs. those who did not. Voxels in red are positive weights (those that as a pattern showed greater activation in children who improved compared with those who did not) and in blue are negative weights. (D) Overlap between results from fMRI univariate regression analysis (green) and multivariate analysis (red) in the right IFG activation is displayed (yellow).