Predicting Clinical Outcome of Stroke Patients with Tractographic Feature

Area of Science:

• Neuroscience
• Medical Imaging
• Neurology

Background:

• Stroke lesion volume is the standard for predicting patient outcomes.
• Undetected neural disruptions in other brain regions can also impact stroke recovery.
• Existing methods may not fully capture these secondary effects.

Purpose of the Study:

• To introduce a novel tractographic feature for predicting stroke patient outcomes.
• To assess the efficacy of this feature in capturing potentially damaged brain regions.
• To compare its predictive performance against traditional stroke volume metrics.

Main Methods:

• Developed a tractographic feature using stroke lesion and average connectome data from healthy subjects.
• Integrated information on affected functional regions into the feature.
• Validated the feature on the Ischemic Stroke Lesion Segmentation 2017 benchmark dataset.

Main Results:

• The tractographic feature demonstrated higher accuracy in predicting modified Rankin Scale (mRS) grades compared to stroke volume and state-of-the-art features.
• Achieved a lower average absolute error in mRS grade prediction than the stroke volume feature.
• Successfully identified and incorporated the impact of neural disruptions in connected brain regions.

Conclusions:

• The tractographic feature offers a more comprehensive approach to predicting stroke outcomes.
• It complements existing stroke volume measurements by accounting for broader neural impacts.
• This novel feature shows significant potential for improving clinical trial assessments and patient prognostication.