Machinelearning‐basedmultimodal prediction of language outcomes in chronic aphasia

Abstract

Recent studies have combined multiple neuroimaging modalities to gain further understanding of the neurobiological substrates of aphasia. Following this line of work, the current study uses machine learning approaches to predict aphasia severity and specific language measures based on a multimodal neuroimaging dataset. A total of 116 individuals with chronic left‐hemisphere stroke were included in the study. Neuroimaging data included task‐based functional magnetic resonance imaging (fMRI), diffusion‐based fractional anisotropy (FA)‐values, cerebral blood flow (CBF), and lesion‐load data. The Western Aphasia Battery was used to measure aphasia severity and specific language functions. As a primary analysis, we constructed support vector regression (SVR) models predicting language measures based on (i) each neuroimaging modality separately, (ii) lesion volume alone, and (iii) a combination of all modalities. Prediction accuracy across models was subsequently statistically compared. Prediction accuracy across modalities and language measures varied substantially (predicted vs. empirical correlation range: r = .00–.67). The multimodal prediction model yielded the most accurate prediction in all cases (r = .53–.67). Statistical superiority in favor of the multimodal model was achieved in 28/30 model comparisons (p‐value range: <.001–.046). Our results indicate that different neuroimaging modalities carry complementary information that can be integrated to more accurately depict how brain damage and remaining functionality of intact brain tissue translate into language function in aphasia.

Mapping lesion, structural disconnection, and functional disconnection to symptoms in semantic aphasia

“…This is consistent with our observation that the small clusters of structural disconnection associated with deficits in semantic cognition largely followed lesioned substrates, although a different pattern was found for executive dysfunction. Hope et al (2018) similarly found limited predictive value of indirectly measured structural disconnection beyond lesion location for predicting aphasia severity, while other studies have suggested unique contributions from such measures (Del Gaizo et al, 2017; Kristinsson et al, 2021). It may be that structural disconnection is unlikely to explain unique variation in behavioural performance beyond lesion location when patients have been selected to show particular deficits associated with areas of cortex that are lesioned, as was the case for semantic deficits in our sample.…”

Edge-centric analysis of stroke patients: An alternative approach for biomarkers of lesion recovery

“…Recent literature has developed strategies to combine different data modalities (anatomical, functional, behavior, etc.) of stroke patients ( Iorga et al, 2021 , Kristinsson et al, 2021 , Pustina et al, 2017 ). While this study does not include anatomical information of the patients, future multimodal studies could adapt the presented methods to include the structural constraints of each individual subject.…”

Cerebral perfusion in post-stroke aphasia and its relation to residual language abilities

“…Furthermore, a perfusion protocol is easier to administer, and can be more readily standardized across sites and hence included in routine clinical assessments. Perfusion measures, while previously largely overlooked in multimodal neuroimaging studies (with the notable exception of Kristinsson et al, 2021), may offer valuable prognostic indicators of recovery potential and should be routinely included in future studies investigating the neural mechanisms of post-stroke recovery.…”

“…Furthermore, a perfusion protocol is easier to administer, and can be more readily standardized across sites and hence included in routine clinical assessments. Perfusion measures, while previously largely overlooked in multimodal neuroimaging studies (with the notable exception of Kristinsson et al, 2021)…”