Stroke: Multimodal neuroimaging and behavioral data to predict speech rehabilitation in stroke

Stroke is a leading cause of severe and complex long-term disability, affecting all areas of social participation
.

Aphasia is one of the most devastating consequences of stroke, affecting approximately one-third of stroke survivors

However, recovery from aphasia after stroke is individualized and influenced by a variety of factors, making it difficult for clinicians to determine prognosis
.

Previous studies have shown that different factors may partially or independently explain the degree of spontaneous or treatment-related language recovery after stroke

Specific verbal or nonverbal ability and demographic may also play a role in the amount of recovery of language ability .
However, there is no consensus in the literature on the role of demographic data in treatment outcomes .

statistical consensus

In addition, better brain structural integrity and local functional activity and connectivity were positively associated with the degree of spontaneous and treatment-related language recovery
.

Importantly, most of these studies investigated the value of a single variable or combined only a few of them

Therefore, it remains unclear (1) how each of the above factors compares in predicting natural language recovery and recovery after rehabilitation; and (2) whether a combination of factors is superior in prediction relative to a single type of factor
.

This question is important because clinicians need to know what types of data are necessary to provide patients with accurate prognosis

Recent advances in machine learning have enabled multivariate analysis of multimodal neuroimaging data to be applied to predict language impairment at a single time point after brain injury or, in longitudinal studies, natural language recovery over time after stroke
.

However, these studies have several limitations: the recovery period investigated varied among participants, the amount of recovery each received was not controlled and only one type of imaging data (ie, functional or structural) was included

A total of 55 people with chronic post-stroke aphasia underwent a series of standardized assessments as well as structural and fMRI scans, and received speech therapy for 12 weeks
.

Using pre-treatment behavioral, demographic, structural, and functional neuroimaging data, support vector machines and random forest models were constructed to predict responsiveness to treatment

Their SVM model trained on measures of aphasia severity, demographics, anatomical integrity, and resting-state functional connectivity achieved the best predictive performance (F1=0.

94)

Compared with SVM models trained on all feature sets (F1 = 0.

82, P < 0.
001) or a single feature set (F1 range = 0.
68-0.
84, P < 0.
001), the predictive performance of this model is significantly superior
.
Of all random forest models, training on resting-state fMRI connectivity data yielded the best F1 score (F1=0.
87)
.

Of all random forest models, training on resting-state fMRI connectivity data yielded the best F1 score (F1=0.
87)
.

The significance of this study lies in the discovery that while behavioral, multimodal neuroimaging data, and demographic information are complementary in predicting recovery responses in chronic aphasia after stroke, resting brain functional connectivity after stroke is a predictor of A particularly important factor in treatment responsiveness , either alone or in combination with other patient-related factors
.

Resting brain functional connectivity after stroke is a particularly important predictor of treatment responsiveness

Original source:
Billot A, Lai S, Varkanitsa M, et al.
Multimodal Neural and Behavioral Data Predict Response to Rehabilitation in Chronic Poststroke Aphasia.
Stroke .
Published online January 26, 2022:STROKEAHA.
121.
036749.
doi:10.
1161/STROKEAHA.
121.
036749

Multimodal Neural and Behavioral Data Predict Response to Rehabilitation in Chronic Poststroke Aphasia.
Stroke commented here