A study for semi-automatic diagnosis support of strokes from 2D MRI/FLAIR sequences

The MRI/FLAIR sequence is being used for the diagnosis of strokes. The bleeding area shows a high intensity in most cases. However, a doctor must check each image which can take a lot of time. We found that in this process a skilled doctor typically checks the gap area (internal cavity) of the brain to give a diagnosis. This fact is confirmed by a questionnaire of several neuro surgeons. In order to extract the gap area, we utilize a region growing method and a 2D deformable model. We also improve the preciseness by eliminating high intensity areas near the bulb of the brain. Finally, we calculate the degree of riskfactor by counting pixels with high intensity in the gap area. From our experience, we found that the histogram of strokes had a similar pattern, and it could be used for diagnosis of strokes. Finally, we visualize the distribution of the bleeding area of the stroke, and evaluate the results in several ways.