Reading therapy strengthens top-down connectivity in patients with pure alexia

Abstract

This study tested the efficacy of audio-visual reading training in nine patients with pure alexia, an acquired reading disorder caused by damage to the left ventral occipitotemporal cortex. As well as testing the therapy's impact on reading speed, we investigated the functional reorganization underlying therapy-induced behavioural changes using magnetoencephalography. Reading ability was tested twice before training (t1 and t2) and twice after completion of the 6-week training period (t3 and t4). At t3 there was a significant improvement in word reading speed and reduction of the word length effect for trained words only. Magnetoencephalography at t3 demonstrated significant differences in reading network connectivity for trained and untrained words. The training effects were supported by increased bidirectional connectivity between the left occipital and ventral occipitotemporal perilesional cortex, and increased feedback connectivity from the left inferior frontal gyrus. Conversely, connection strengths between right hemisphere regions became weaker after training.

Keywords: alexia; connectivity; magnetoencephalography; reading disorders; stroke.

Figure 1

Schematic diagram of research protocol over four time points (t1 to t4).

Figure 2

Optimal source locations from the variational Bayesian equivalent current dipole analysis for each subject plotted on a glass brain in MNI space. The starting points for the source locations were: occipital cortex: ±15 −95 2; ventral occipitotemporal cortex: ±44 −58 −15 14; inferior frontal gyrus: ±48 28 0. The average locations (with standard deviations, SD) of the winning source locations for the patient group are also reported. OCC = occipital cortex; vOT = ventral occipitotemporal cortex; IFG = inferior frontal gyrus (n = 8).

Figure 3

(A) Lesion overlay map for the patient group, demonstrating the area of maximal overlap in occipital white matter of the left hemisphere. Crosshairs indicate the location of the left occipital (OCC) and ventral occipitotemporal (vOT) spatial priors for the source localization analysis (n = 8). (B) Reading speed slopes demonstrating the effect of word length (3–6 letters) on reading latency for the patient group (pink) and control group (blue). Each cross represents data from an individual patient.

Figure 4

The effect of reading training on word reading speed. (A) Average word reading speed across all patients at each time point (t1–t4) for trained (pink line) and untrained (blue line) words. Training occurred between t2 and t3. (B) Reading speed slopes showing the effect of word length before training (the average of t1 and t2, dotted lines) and after training (t3, solid lines) for trained (pink) and untrained (blue) words. (n = 9).

Figure 5

Results of the DCM analysis demonstrating the modulatory effects of stimulus type (trained versus untrained words at t3) on connection strengths. Numerical values represent the posterior mean of the gain on connection strength: values significantly greater or smaller than 1 indicate connections that are stronger for trained words (in pink) or stronger for untrained words (in blue), respectively. Significance threshold: P > 0.9 (n = 8). OCC = occipital cortex; vOT = ventral occipitotemporal cortex; IFG = inferior frontal gyrus.

Figure 6

Global field power during word reading in the patient group (trained words, pink solid line; untrained words, pink dotted line) and a group of age-matched healthy controls (blue) using data from Woodhead et al. (2012).