Morphological image analysis for studying Alzheimer's disease

One of the hallmarks of Alzheimer's disease is the presence of a brain lesion known as a senile plaque (SP), which appears as a circular cluster of small objects when viewed through the microscope. To quantitatively study their number and distribution, the author has developed an image analysis system that will automatically detect and count these lesions. The morphological operation of closing can group a cluster of small objects into a singular one, and thus allow detection of the SP. Since the optimal structuring element size for grouping the SP is different in each image, the Bayesian decision-making method is used to dynamically determine the size. A study comparing computer and human detection of SP was performed by evaluating 50 images from five different patients. The results of this comparison demonstrate that the computer counts correlate well with the neuropathologists counts (correlation coefficient=0.81).<
>