Post-stroke apathy biotypes and their relation to the effort-based reward decision network: a resting-state fMRI study

Abstract

Abstract Post-stroke apathy (PSA) is a highly heterogeneous disorder that affects approximately 30% of stroke survivors. Nevertheless, comprehensive understanding regarding neurobiological mechanisms the heterogeneity of PSA is lacking. We hypothesized that the effort-based reward decision network (ERDN) may play a critical role in PSA heterogeneity. Therefore, we prospectively recruited 190 patients with acute ischemic stroke and 50 demographically matched healthy controls. Sparse canonical correlation analysis (SCCA) was employed to elucidate the associations between symptoms of PSA and patterns of resting-state functional magnetic resonance imaging (rsfMRI) functional connectivity. Through the application of hierarchical clustering, we successfully identified four distinct PSA biotypes based on their unique connectivity profiles. Biotype 1 had high levels of both apathy and depression at baseline. Biotype 2 had consistently higher levels of apathy but lower levels of depression at baseline. Biotype 3 had low levels of apathy and depression at baseline and follow-up. Biotype 4 had higher levels of depression but lower levels of apathy at baseline and follow-up. Furthermore, biotype1, 2 and 3 had varying degrees of increased scores on different dimensions of apathy relative to the overall mean. We used machine learning to evaluate the predictive performance of the ERDN connectivity model compared to the whole-brain connectivity model. Our results indicate that the ERDN model exhibited similar or superior predictive capabilities compared to the whole-brain model in biotype 1 and 2. The multiscale rsfMRI parameters in the ERDN were investigated further. Only biotypes 1 and 2 deviated from the overall mean in terms of graph-theoretic parameters, with biotype 1 having lower values and biotype 2 having higher values. On the other hand, these biotypes displayed distinct characteristics in terms of their functional separation parameters. Our study emphasizes the importance of ERDN in PSA heterogeneity and provides new insights for future research and therapeutic targets.