SVM-based Detection of Bone Fracture using Weighed Local Binary Pattern Features

The work presented in this research outlines a new method for detecting bone fractures, which is based on a machine learning model. To achieve this goal, Support Vector Machine (SVM) classifiers are trained on a dataset of images depicting bone fractures and normal bones enabling them to categorize bone images into two classes - bone fracture and non-bone fracture images. The results have shown that the system attains a great performance of 98.42% accuracy, suggesting high correctness of fracture detection cases. Moreover, the articulated system also comprises state-of- the- art image preprocessing methods such as removal of noise, any adaptive histogram equalization, additional sharpening and edge detection of images for the purpose of improving features. Among key developments, is the introduction of a unique feature which is termed as Weighed Local Binary Pattern (WLBP) for the inquisitive information containing textures and patterns in the images. Therefore, the present study tends to provide a very strong as well as an accurate model for fracture assessment, which is likely to improve the quality of the diagnosis and the patients’ care.