Anisotropic diffusion filtering of computed tomography images and improving diagnosis of stroke

Stroke is a devastating neurological event and one of the leading causes of death and disability. Rapid diagnosis and treatment ensure the best outcome for afflicted patients. The Alberta Stroke Program Early Computed Tomography Score (ASPECTS) was developed to assess the extent of damage in ischemic stroke. Despite extensive training, significant operator variability exists in scoring. The anisotropic diffusion filter is a noise-reducing filter that preserves edges and fine detail information. Acceleration of the algorithm with specialized hardware yielded algorithmic speed-ups of up to 50 times. The anisotropic diffusion filter was also applied retrospectively to acutely acquired, noncontrast computed tomography scans to reduce operator variability for ASPECTS. It was shown that within-reader variability in scoring was significantly reduced without affecting between-reader variability or accuracy of scoring. These results are significant as they shown that off-the-shelf graphics hardware can be used for image processing to improve clinical diagnosis of acute stroke.