The PLORAS database

All the behavioural and structural neuroimaging data reported throughout the current thesis were retrieved from the Predicting Language Outcome and Recovery After Stroke (PLORAS) database (Seghier et al., 2016). No fMRI data from either stroke patients or neurologically-normal controls acquired as part of the PLORAS project are analysed or indeed discussed in subsequent sections. The reader is referred to the experimental chapters where the study-specific patient selection criteria are specified.

The PLORAS database includes brain images, demographic information and the results of behavioural assessments that have been collected from hundreds of stroke patients at the Wellcome Centre for Human Neuroimaging since 2003 with the goal of

58

understanding, characterising and modelling the most important sources of inter-patient variability in lesion-outcome associations (Price et al., 2010a). Patients can take part in the project independently of the extent and location of the brain damage they have acquired, the presence or absence of aphasia and the amount of time elapsed since stroke onset. Inclusion criteria to the PLORAS database include: (i) a demonstrable previous medical history of stroke; (ii) no record of concomitant neurological or psychiatric illness (e.g., dementia or depression); and (iii) being able to provide written informed consent.

The structural brain scans are run through an automated lesion identification algorithm (see below for details) whose output provides detailed information about the location and extent of damage in patients with and without acquired language disorders. The lesion sites that are found to consistently result in very specific functional impairments (i.e. critical lesion sites) are then used to predict outcome in new patients and also offer unique insights into the functional anatomy of language. Those patients who perform unexpectedly well on a number of language tasks despite relatively focal damage to previously identified critical regions or relatively large left or right hemisphere strokes are invited to participate, upon obtaining informed consent, in a multifactorial fMRI paradigm designed to tease apart a whole range of language processing levels. The selection criteria for the fMRI component of the PLORAS project are both study- specific and constantly being updated as our knowledge about how very specific language impairments relate to lesion site progresses over time. For example, in the fMRI study by Seghier et al. (2014), the goal was to investigate how speech production can be supported after left putamen damage. Accordingly, patients with relatively focal left putamen damage were contacted to complete the PLORAS fMRI paradigm so as to identify brain areas with abnormally high activation during successful reading aloud and picture-naming responses. In short, the fMRI data allow us to explain inter-patient variability in lesion-deficit mappings by pinpointing which of the preserved brain regions are functioning normally or abnormally, and how they connect to one another. The fMRI

59

data can therefore be used to understand the neural pathways that support normal behaviour following brain damage. Neurologically-normal individuals are also invited to participate in the fMRI component of the project to define a range of normal responses and establish the degree to which each patient’s fMRI pattern departs from that normal range. When possible, the behavioural and neuroimaging data are collected at multiple time points after stroke so as to sample the entire recovery process. By integrating cross- sectional and longitudinal behavioural, demographic and neuroimaging data, the main factors that determine long-term outcome and recovery after stroke start to emerge. This knowledge is then progressively translated into structure-function-recovery rules in the form of a clinical prognosis tool which is expected, in the future, to generate the most likely recovery trajectory for a new patient based on the site and extent of the stroke lesion, thereby informing clinical prognosis and therapy planning (see Figure 2.1).

2.1.1 My contribution to the PLORAS database

Over the course of the last 5 years, I have actively contributed (as part of a team effort) to the continuous expansion of the PLORAS database by participating in the recruitment, assessment and scanning of tens of stroke patients. I have also been involved in the collection and analysis of neuroimaging data from patients with brain tumours with the goal of better understanding how the gradual expansion of tumours versus the acute onset of strokes impacts upon the language system. Finally, I have set up a project in Chile (my mother country) which aims to enable, in the medium term, the realisation of studies investigating the effects of cultural (UK versus Chile) and linguistic (English versus Spanish) variables on language outcome and recovery after stroke. This has entailed translating the Comprehensive Aphasia Test (see below) into Spanish, sorting out the logistics of the project, securing ethical clearance, testing approximately 100 Chilean stroke survivors and obtaining their clinical scans (i.e. computerized tomography).

60

Figure 2.1: The PLORAS approach. The structural brain scan of a new patient is converted into a 3-dimensional image encoding the location and extent of damage incurred. The lesion image is then compared to those of all other patients in the database. The speech scores of the patients with similar lesions (both in terms of size and location) and demographic details are subsequently retrieved. By plotting the behavioural scores of the matching patients against time post-stroke, a prediction for the new patient is generated which indicates their likely time course of recovery with a confidence rating. Reproduced with permission (Elsevier) from Seghier et al. (2016).