Close
Patient recruitment
Patients are recruited from a variety of sources, including the Clinical Research Network, regional stroke clubs, speech and language therapy teams, newspaper advertisements and events for stroke survivors and professionals working in stroke, e.g. UK Stroke Assembly and UK Stroke Forum.
Time post-stroke
This is a critical factor for predicting how well a patient will have recovered. Patients are therefore recruited at all stages post stroke from acute (within first month) to chronic (more than one year). The effect of time post stroke can then be examined across different patients with similar lesions.
Structural brain images
We obtain lesion information from all our participants using high resolution T1 weighted anatomical whole brain images acquired with 1.5T or 3T MRI. Although the predictors and critical lesion sites have been developed on the basis of these high resolution images, we anticipate that the system will be sufficiently robust to generate predictions for patients whose clinical brain scans were acquired with low resolution MR contrasts, or possibly computed tomography (CT). It will be critical to test this carefully as we develop the PLORAS procedures over time.
At this stage, MRI scans from participants' local hospitals cannot be used, i.e. participants are required to have a scan at our laboratory. When we have further tested the accuracy of our predictions in the controlled environment of our laboratory, it should become possible for MRI scans taken at local hospitals to be used. Indeed, this is how we envisage the system working in order that future stroke patients around the UK who experience aphasia can be given a prediction about the recovery of their speech by the doctors/therapists at their local hospital. Appropriate goals and realistic expectations can then be set for the individual.
Language assessment
The language assessment we use for all stroke patients is the Comprehensive Aphasia Test (CAT) [1] which comprises a cognitive screen and language battery, including tests of comprehension and production of both single words and sentences presented in both auditory and written format. The CAT is administered to all patients irrespective of whether they report any aphasia as a result of their stroke.
Aphasia recovery self-rating
We have developed a simple questionnaire for capturing each patient's perspective on whether they currently have any difficulties with speaking, understanding, reading and writing. Patients are also asked to rate their ability in each of these dimensions at one week, month and year post stroke. This provides us with some insight into whether they have experienced any aphasic symptoms prior to meeting us. Brief information on amount, nature and timing of speech and language therapy is also collected.
Repeat assessments
To monitor the effect of time post-stroke within patients, we repeat structural imaging and behavioural assessments in any patient with a language deficit recruited within the first year of their stroke. Assessments will be made every six months between acute and chronic stages. These data will also be used to examine how brain structure and function change during the course of recovery.
Participants recovering from aphasia may be invited to return for subsequent follow up scans and/or language assessments.
Duration of patient participation
This is a cross-sectional study therefore participants are only required to make a one-off visit to us. However, some participants who demonstrated a range of language difficulties at their initial appointment are invited back for repeat assessments, to provide longitudinal data. Participants will be invited to come to our laboratory in central London when it is convenient for them (travel expenses paid).
Participants have a structural brain scan that takes approximately 13 minutes. Prior to or after the brain scan, the participant's language and cognitive abilities will be tested. These tests may take up to two hours, but usually take 60 minutes. A typical appointment lasts 2.5 - 3 hours in total with tea breaks, etc, included.
Lesion extraction
Given our large sample sizes, it is essential to have an efficient, reliable, automated lesion identification algorithm. The method we have developed generates two 3D-lesion images in standard (MNI) space for each patient [2]. The first lesion image codes the degree of abnormality of tissue (relative to typical values in a large cohort of healthy controls). The second is thresholded to create a binary 3D-lesion image. The correspondence between the lesion images and the raw brain scan is checked manually for each patient. To date, the images generated have provided robust and replicable results. Nevertheless, they will be refined when both possible and necessary.
Lesion similarity
Rule-Based Similarity: Rules are based on critical lesion sites and (if known) the pathways that, when left intact, can support subsequent recovery. Each rule is effectively a validated hypothesis concerning the functional contribution of a region, or network of regions. When a new patient's lesion is found to invoke a particular rule, our predictions will be based on other (predictor) patients who invoke the same rule.
Behavioural predictions
In the PLORAS system, predicted prognoses are similarity-weighted averages of predictor patients' scores, at a range of times post-stroke. The measures of speech production and comprehension that we are using are based on combinations of scores from the Comprehensive Aphasia Test. Our experience is that these scores are very strongly correlated with patients' experiences of life; and impairments on our single-word production measures strongly predict how well a patient can describe a complex scene. We are therefore confident that our behavioural measures are a reliable proxy for the practical things that patients might be expected (and might want) to do in their daily lives. This will be investigated over the course of the project.
References
1. Swinburn, K., Porter, G. and Howard, D. 2004. Comprehensive Aphasia Test. Psychology Press: Hove.
2. Seghier, M., Ramlackhansingh, A., Crinion, J., Leff, A. and Price, C. 2008. Lesion identification using unified segmentation-normalisation models and fuzzy clustering. Neuroimage 41, 1253.