Deep learning-based quantification of white matter hyperintensity applicable to real-world clinical FLAIR images

White matter hyperintensity (WMH) in the brain is known to correlate with cognitive prognosis in many diseases; automated quantification tools for WMH have been developed, but most have been used to quantify study data from specific diseases imaged with a single scanning protocol. The low accuracy of these tools when used for clinical data with diverse scan protocols and diseases has been a problem in clinical applications. To overcome this limitation, we developed a deep-learning-based WMH quantification model for real-world clinical FLAIR images with high heterogeneity. The results show the potential of this method as a clinical tool.