Automated White Matter Lesion Segmentation in MRI using Box-Whisker Plot Outlier Detection

We report a new automated method for White Matter Lesions (WMLs) Segmentation in cranial MR Imaging. WMLs are diffuse white matter abnormalities which are often presented as hyperintense regions. In our approach, the presence of these abnormalities are detected as outliers in the intensity distribution of the FLAIR sequence using the his-togram tails analysis and Box-Whisker plot technique. In addition, our method includes post-processing to reduce False Positives attributed to MRI artefacts commonly observed in FLAIR sequences. We validated our approach using 19 cases of cranial MRI. A high correlation is seen between our automated approach and the results of a manual visual scoring approach performed by an expert radiologist.