Review on Segmentation of Computer-Aided Skeletal Maturity Assessment

Bone age assessment (BAA) is an examination of ossification development with the purpose of deducing the skeletal age of children to monitor their skeletal development and predict their future adult height. Conventionally, it is performed by comparing left-hand radiographs to standard atlas by visual inspection; this process is subjective and time-consuming; therefore, the automated inspection system to overcome the drawbacks is established. However, the automated BAA system invariably confronts with problem in segmentation, which is the most crucial procedure in the computer-aided BAA. Inappropriate segmentation methods will produce unwanted noises that will affect the subsequent processes of the system. The current manual or semi-automated segmentation frameworks have impeded the system from becoming truly automated, objective, and efficient. The objective of this thesis is to provide a solution to the mentioned unsolved technical problem in segmentation for automated BAA system. The task is accomplished by first applying the modified histogram equalized module, then undergoing the proposed automated anisotropic diffusion, following by a novel fuzzy quadruple division scheme to optimize the central segmentation algorithm, and finally, the process ends with an additional quality assurance scheme. The designed segmentation framework works without the need of resources such as training sets and skillful operator. The quantitative and qualitative analysis of the resultant images have both shown that the designed framework is capable of separating the soft tissue and background from the hand bone with relatively high accuracy despite omitting the above-mentioned resources.