Literature Review: Segmentation Methods and Feature Extraction in Bone Imaging

Abstract

This study presents a systematic literature review on the development of segmentation and feature extraction methods in bone imaging, which play a crucial role in improving the accuracy and efficiency of medical image analysis. The review follows the PRISMA guidelines to ensure that the literature selection process is transparent, structured, and replicable. Out of 200 initially identified studies, six articles met the inclusion criteria after undergoing the stages of identification, screening, eligibility assessment, and final inclusion. The findings reveal that traditional segmentation methods—such as thresholding, watershed, and active contour—remain widely used but exhibit limitations when applied to bone images with complex structures. Deep learning–based approaches, particularly U-Net, have emerged as a dominant trend due to their ability to produce more precise segmentation and support automated feature extraction. Commonly used feature extraction techniques include GLCM, LBP, HOG, and CNN-based deep features. Overall, recent studies emphasize the importance of combining preprocessing, adaptive segmentation, and robust feature extraction to enhance the detection of bone structures, including micro-fractures. This review also highlights the need for more comprehensive datasets and broader clinical validation to ensure that these techniques can be optimally implemented in computer-aided diagnostic systems.