engleski

Automatic Radius Bones Fracture Detection using Machine Learning

Radius bone fracture is one of the most common occurring injury in humans . The usual procedure of detecting radius bone fracture is by examination of its X - ray image by radiologist. Besides denoising and brightness and contrast adjustments, the examination process of an X-ray image is not automated or enhanced by any software assistance. The aim of this work is to create a fully automated fracture detection software for calculating probability of existence of radial bone fracture on X-ray images. The preprocessing of an X-ray image is done by edge detection of the radius bone with contour generation. Method used for edge detection and contour generation is adaptive thresholding [1]. Proposed approach for detecting fracture uses the difference between the estimated unbroken radius bone line and the real contour of a bone. The estimation of the unbroken bone is calculated using a combined bootstrap method with polynomial regression of contour points [2]. After calculating the differences, the error threshold is set dividing fractured bone images from non-fractured ones. Also, this approach enables detecting the exact area of the fracture. Afterwards, this method is evaluated based on precision its achieves on detecting fractures on previously unseen X-ray images.

Radius bone fracture ; X - ray, Polynomial regression ; Machine learning

nije evidentirano