Extraction of skeletal patterns from magnetic resonance images using mathematical morphological filters

It is recognized that bone quality and strength are affected by trabecular bone structure. The purpose of this study was to establish a method for the three-dimensional analysis of trabecular bone structures for use in the evaluation of bone quality. A mathematical morphological filter was applied to magnetic resonance (MR) images of the radius, tibia, calcaneus, and the third lumbar vertebra, and their trabecular bone structures were extracted as binary skeletal patterns. A superimposition of MR image and binary skeletal pattern was conducted in order to demonstrate that the filter accurately extracted trabecular bone structure. The filter successfully extracted trabecular bone structures from MR images as distinctive binary skeletal patterns. The skeletal patterns were obtained as a set of subset images and sumset images that comprised spatial frequency components that agreed with the equation and structural element of the filter. The extracted skeletal images exhibited characteristic patterns corresponding to the type of bone, slice direction, and slice position. It was concluded that mathematical morphological filter processing of MR images is a useful method of evaluating bone quality through the three-dimensional analysis of trabecular bone structures.