Fast semi-automated lesion demarcation in stroke.

摘要: Lesion-behaviour mapping analyses require the demarcation of brain lesion on each (usually transverse) slice individual stroke patient's image. To date, this is generally thought to be most precise when done manually, which is, however, both time-consuming and potentially observer-dependent. Fully automated methods have been developed address these issues, but are often not practicable in acute research where for patient only a single image modality available differs over patients. In current study, we evaluated semi-automated approach, so-called Clusterize algorithm, patients scanned range common modalities. Our results suggest that, compared standard manual demarcation, algorithm capable significantly speeding up commonly used modalities, without loss either precision or reproducibility. For three investigated datasets (CT, DWI, T2FLAIR), containing total 44 images obtained regular clinical setting at admission, reduction processing time was average 17.8 min per advantage increased with increasing volume (up 60 min largest volumes our datasets). Additionally, that performance chronic dataset 11 T1 comparable its datasets. We thus advocate use integrated into simple, freely SPM toolbox, precise, reliable fast preparation imaging data lesion-behaviour analyses.