Improving the prediction of language recovery in stroke patients by including risk and protective neurocognitive factors.

Aphasia is a language disorder most commonly caused by a cerebrovascular accident, or stroke, in the left hemisphere. Approximately 15-45% of patients in acute stroke settings have aphasia, which impairs their communication and dramatically affects their quality of life. In the first months after stroke, there is often spontaneous and intervention-induced recovery, but for 26-43% a chronic language deficit remains. To date, individual language recovery in patients with aphasia remains hard to predict. Yet, the identification of early predictors of aphasia recovery could not only manage patients’ expectations, but also maximize the recovery potential through well-targeted individualized interventions. While it is established that neuroplasticity is an important driver of language recovery, it is unknown which neural reorganization processes are effective and which are maladaptive for good language outcomes. Improved knowledge on how the brain reorganizes after stroke, especially during the first months, will be helpful to inform neuroimaging-based prediction of long-term language outcomes.

The present dissertation had two overarching aims. The first aim was to identify early markers that improve the prediction of language recovery in stroke patients. Several predictors of language recovery after stroke have been identified, yet there is a lack of predictors that can be measured immediately after stroke - i.e., in the clinically relevant acute phase. Moreover, previous research has mainly focused on the influence of brain damage and language deficits on language outcomes. What has largely been missing is the focus on possible protective factors as potential predictors of recovery, including intact gray and white matter, intact general cognitive abilities and their corresponding neural substrates. Our hypothesis was that statistical learning, a mechanism upon which we rely heavily in daily life to learn the structure inherent to our complex linguistic environment, and the hippocampus, as one of the potential neural substrates of statistical learning, could support language (re)learning in aphasia. The second aim was to gain insight in neuroplastic changes during recovery and their associations with language. Past research has been mostly restricted to the study of neuroplasticity in cortical regions years after the stroke has occurred, while there is a lack of knowledge on how the brain reorganizes in the first months after stroke. Moreover, language is not organized in isolated brain regions, but rather in a complex structural network of interconnected cortical regions. We therefore explored early plasticity in long-ranging white matter connections involved in language processing, i.e., the dorsal arcuate fasciculus (AF) and the ventral inferior fronto-occipital fasciculus (IFOF). These two overarching aims were addressed in four different studies.

In the first study (Chapter 4), we conducted a thorough review of the literature to obtain a complete picture of what is known on neuroplasticity patterns in the post stroke aphasia population. We found that stroke-induced neuroplasticity related to language interventions was not limited to language structures in the left hemisphere, but involved a bilateral network of structures that support language from a broader cognitive perspective. In our second study (Chapter 5), we showed that it is possible to construct tasks that can capture non-linguistic statistical learning in healthy elderly and that are at the same time feasible in difficult-to-test populations, such as patients with aphasia. Our third and fourth study entailed a longitudinal follow-up of patients with post stroke aphasia. In our third study (Chapter 6), we showed that we were indeed able to measure non-linguistic statistical learning in patients with aphasia in the subacute phase, which was intact compared to a healthy older control group. According to our expectations, the subacute behavioral statistical learning results were associated with acute hippocampal measures. Moreover, the acute volume of the left hippocampus significantly contributed to the prediction of long-term language outcomes over and above traditional predictors. In our last study (Chapter 7), we established that the acute connection strength of the AF and IFOF was associated with the severity of the acute language deficit, but not predictive of later language outcomes over and above information on the initial language impairment. Concerning neuroplasticity, we observed white matter neurodegeneration in the first months after stroke, with changes in the AF being associated with worse language outcomes.

In sum, the work that was carried out in the context of this dissertation highlights the potential importance of an individual’s (intact) cognitive capacity for compensation and corresponding neural correlates for obtaining a more reliable prediction of language recovery after stroke. Moreover, we have provided new insights in (changes in) connectivity of damaged and undamaged language pathways in patients with aphasia in the first months after stroke, as well as if/how such measures are related to language outcomes at different stages of recovery. Although the current observational and longitudinal data do not allow us to draw firm causal conclusions, they are a vital first step to explore the nature of brain-behavior associations over time. The ultimate end goal is to obtain evidence-based prognostic models which would be beneficial for patients, relatives, clinicians, and even society.