A Canonical Correlation Approach Towards Identifying Latent Neurofeedback Responses in Patients with Schizophrenia

Patients with schizophrenia exhibit frontal gamma dysregulation, and associated impairments in cognitive function. To improve self regulation of frontal gamma activity, we designed a double-blind, placebo-controlled randomized clinical trial to test a novel neurofeedback (NFB) protocol, which rewards active maintenance of current or higher levels of frontal gamma coherence (gcoh-NFB). We report here unique treatment responses among participants with high versus low baseline working memory (WM) function. At baseline, among All participants, greater gamma dysregulation (higher resting gcoh) positively correlated with greater WM impairment. Among Active participants, completing gcoh-NFB training lowered the elevated baseline gcoh in participants with lower baseline WM function, and, conversely, increased the relatively lower baseline gcoh in participants with higher baseline WM function. In contrast, completing placebo-NFB produced no gcoh changes, regardless of baseline WM function. Compared to Placebo participants, all Active participants exhibited improved WM with training. The differing gcoh responses to NFB among Active participants suggested multiple neurophysiologic and WM responses during treatment. We selected canonical correlation analysis (CCA) to further evaluate potential latent and divergent responses. CCA identified two statistically significant canonical components; the stronger component representing the expected positive interactions between training-related WM responses, and the weaker component representing diverging interactions between training-related NFB and WM responses. Coefficients for the stronger (but not the weaker) component efficiently separated participants into distinct clusters with high, versus low, baseline WM, suggesting this response as the primary driver of divergent yet equally therapeutic effects observed for patients with low or high baseline WM function.