Know your tools--concordance of different methods for measuring brain volume change after ischemic stroke

Longitudinal brain volume changes have been investigated in a number of cerebral disorders as a surrogate marker of clinical outcome. In stroke, unique methodological challenges are posed by dynamic structural changes occurring after onset, particularly those relating to the infarct lesion. We aimed to evaluate agreement between different analysis methods for the measurement of post-stroke brain volume change, and to explore technical challenges inherent to these methods.

Fifteen patients with anterior circulation stroke underwent magnetic resonance imaging within 1 week of onset and at 1 and 3 months. Whole-brain as well as grey- and white-matter volume were estimated separately using both an intensity-based and a surface watershed-based algorithm. In the case of the intensity-based algorithm, the analysis was also performed with and without exclusion of the infarct lesion. Due to the effects of peri-infarct edema at the baseline scan, longitudinal volume change was measured as percentage change between the 1 and 3-month scans. Intra-class and concordance correlation coefficients were used to assess agreement between the different analysis methods. Reduced major axis regression was used to inspect the nature of bias between measurements.

Overall agreement between methods was modest with strong disagreement between some techniques. Measurements were variably impacted by procedures performed to account for infarct lesions.

Improvements in volumetric methods and consensus between methodologies employed in different studies are necessary in order to increase the validity of conclusions derived from post-stroke cerebral volumetric studies. Readers should be aware of the potential impact of different methods on study conclusions.