Development of quantitative computed tomography image resolution enhancement based on fully automated femur segmentation and topology optimization

Abstract

Osteoporosis and hip fracture have imposed societal, economic, and individual burden with population aging. Although much research has been studied to reduce their occurrence, above all things, prevention though early diagnosis is the best cure due to their irreversibility. For accurate diagnosis, the high resolution trabecular architecture image of region of interest is essential. For this reason, this thesis proposed a novel method which enhances a quantitative computed tomography image resolution based on fully automated femur segmentation and topology optimization. First, a fully automated segmentation method was proposed to extract a three-dimensional (3-D) proximal femur from the raw QCT data. Watershed algorithm obtained the segmented patches regardless of unclear boundaries caused by limited spatial resolution. Then, the femur areal information was collected through load path algorithm and adaptive thresholding. By merging two complementary information, a 3-D proximal femur could be extracted from the raw QCT data. The proposed segmentation method was validated by comparing the results of different methods. Secondly, a novel method for QCT image resolution enhancement was proposed based on topology optimization. A patient-specific 2-D finite element (FE) model was constructed using the BMD information on the coronal planes of the 3-D segmented femur. The Young’s modulus of each finite element was assigned using the Young’s modulus-BMD relationship reported in the literature. Compliance minimization topology optimization was performed under daily loading conditions with the BMD deviation constraint, which minimizes the difference of estimated BMD from the original data. Through the above process, trabecular architecture with a resolution of 78μm could be successfully obtained for the three regions of interest from the QCT data of $625$\mu m$ resolution. This thesis proposed quantitative computed tomography image resolution enhancement based on fully automated femur segmentation and topology optimization. With further research, it would contribute to the development of patient-specific medications including early diagnosis for osteoporosis and its prognosis for timely medical interventions.