Brain Biomarkers of Response to Treatment for Apraxia of Speech

For each participant, three target sounds were chosen: single consonants, vowels, or clusters at the word level. Based on these target sounds, 10 word items were generated for each target sound that served as Trained Items. Participants were tested on lists of Trained Items (probe trials) during repeated sessions in pre-intervention stage to establish baseline performance (percent of each word list repeated correctly) and during every other treatment session (again, percent of each word list repeated correctly). The change in percent correct list repetition from pre-intervention (5 probe trials) to the end of intervention (final 3 probe trials) was calculated as individual treatment effect sizes using the Busk & Serlin (1992) d2 statistic, which involves subtracting the difference between mean performance at end of the intervention minus pre-intervention, divided by the pooled standard deviation of the two phases. The larger the d2 effect size, the larger the effect of the treatment.

To test generalization, we assessed the change in performance (percent correct repetition) on lists of 10 Untrained Items that had the same speech production targets as each patient's lists of 10 Trained Items, balanced for syllabic structure, word frequency, grammatical form class, and stress pattern. Just as for Trained Items, participants were tested on the lists of Untrained Items (probe trials) during repeated sessions in the pre-intervention stage to establish baseline performance and during every other treatment session (percent of each word list repeated correctly). The change in percent correct list repetition from pre-intervention (5 probe trials) to end of intervention (final 3 probe trials) was calculated as individual treatment effect sizes using the d2 statistic: the difference between mean performance at end of the intervention minus pre-intervention, divided by the pooled standard deviation. The larger the d2 effect size, the larger the effect of the treatment.

To measure neuroplasticity associated with the speech treatment protocol, the investigators calculated the percent change in fractional anisotropy (FA) from pre- to post-treatment for each of the eight fiber tracts in the left hemisphere. The average percentage change in FA across all eight fiber tracts was calculated, which could thus range from 0-100%. A positive change indicates an increase in white matter integrity, and a negative change indicates a decrease in white matter integrity. A number close to 0 indicates minimal/no significant change.