Abstract
Introduction: Post stroke emotionalism (PSE) is a common but poorly understood condition. The value of altered brain structure as a putative risk factor for PSE alongside routinely available demographic and clinical variables has yet to be elucidated.
Methods: 85 patients were recruited from acute inpatient settings within 2 weeks of stroke. PSE was diagnosed using a validated semi-structured interview and standardised measures of stroke severity, functional ability, cognition, mood and quality of life were obtained. Neuroimaging variables (intracranial volume and volumes of cortical grey matter, subcortical grey matter, normal appearing white matter, cerebrum, cerebrospinal fluid and stroke; white matter hyperintensities; and mean cortical thickness) were derived using standardised methods from Magnetic Resonance Imaging (MRI) studies. The relationships between PSE diagnosis, brain structure, demographic and clinical variables were investigated using machine learning algorithms to determine how well different sets of predictors could classify PSE.
Results: The model with the best performance was derived from neuroradiological variables alone (sensitivity = 0.75; specificity = 0.8235), successfully classifying 9/12 individuals with PSE and 28/34 non-PSE cases.
Conclusions: Neuroimaging measures appear to be important in PSE. Future work is needed to determine which specific variables are key. Imaging may complement standard behavioural measures and aid clinicians and researchers.
Methods: 85 patients were recruited from acute inpatient settings within 2 weeks of stroke. PSE was diagnosed using a validated semi-structured interview and standardised measures of stroke severity, functional ability, cognition, mood and quality of life were obtained. Neuroimaging variables (intracranial volume and volumes of cortical grey matter, subcortical grey matter, normal appearing white matter, cerebrum, cerebrospinal fluid and stroke; white matter hyperintensities; and mean cortical thickness) were derived using standardised methods from Magnetic Resonance Imaging (MRI) studies. The relationships between PSE diagnosis, brain structure, demographic and clinical variables were investigated using machine learning algorithms to determine how well different sets of predictors could classify PSE.
Results: The model with the best performance was derived from neuroradiological variables alone (sensitivity = 0.75; specificity = 0.8235), successfully classifying 9/12 individuals with PSE and 28/34 non-PSE cases.
Conclusions: Neuroimaging measures appear to be important in PSE. Future work is needed to determine which specific variables are key. Imaging may complement standard behavioural measures and aid clinicians and researchers.
Original language | English |
---|---|
Article number | 120229 |
Journal | Journal of the Neurological Sciences |
Volume | 436 |
Early online date | 21 Mar 2022 |
DOIs | |
Publication status | Published - 15 May 2022 |