Auto-ROIs of bone scan image using thin-plate spline with specific bone landmarks

Bone scan image (BSI) is the most widely available technique for detecting cancer that has spread (metastasized) from the original location, such as breast or prostate cancer. One benefit of this process is as strong prognostic indicator of survival longevity for cancer patient. However, some challenging issues should be considered, i.e. variability in image brightness, abnormal bone shape or structure, asymmetry gesture position, and image noise due to muscle and fat tissue make the process to determine the region of interest (ROI) become more difficult, not to mention the required-time when inspecting the ROI of each bone part manually. The purposes of this study are to provide an effortless and meticulous ROI labeling as prerequired in clinical application. Finally, to propose a novel framework for determination of multiple ROI automatically by utilizing specific bone landmarks, in the hope of that could contribute to clinical observation in metastasis analysis. The results show that the proposed framework is promising method with high sensitivity (0.96), specificity (0.97), precision (0.94), Sorensen-Dice Index (0.96) and image content index (0.96).