The plaques are segmented by applying gaussian blurring to the amyloid signal and using local hysteresis thresholding to approximate the extent of each plaque. The segmentation is then refined using morphological operations to remove any small holes in the identified plaques and to reject the smallest plaques.

The segmented plaques are used as starting points to identify concentric ring-like regions-of-interest (ROIs) with increasing radial distance from the core of the plaques.

To generate the ROIs, each plaque is first expanded using binary dilation with a disc structuring element of 5µm. We then subtract the initial plaque segmentation from the dilated plaque to obtain a ring-like ROI that contains all pixels that are within 5µm of a given plaque. This process is repeated multiple times to generate radial ROIs around all plaques. These radial ROIs are used later to accurately localize the neuroinflammation around the plaques and to quantify microgliosis and astrogliosis as a function of the distance to the β-amyloid pathology.