Abstract:
Computer Aided Diagnosis (CAD) has been a major research subject for diagnosing
pathologies using computer vision and artificial intelligence. Human body is composed of 206
long, short and irregular bones. Bones are very prone to common pathology known as fractures.
There are several etiologies of bone fractures. The bone fractures are of various types ranging
from highly devastating comminuted fractures to hair line fractures. An algorithm has been
proposed in this study to detect bone fractures using image processing techniques. For that
purpose, MATLAB v. R2015a was used to execute the said task. 126 plain radiographic X ray
images, acquired from a public sector hospital of Islamabad, containing long bone fractures were
classified into “ground truth annotated” and their counterpart “test” data sets. The “test” images
were preprocessed by contrast adjustment and noise removal followed by segmentation into
background and foreground by Active Contour Model. Hough transform is applied, as a feature
extraction technique, after that, for detecting vertical lines in the image which lead to
identification of bone fractures. The results were calculated through Jaccard Index and finally the
mean precision value for each image was calculated. The percentage precision was equal to
88.52% which is highest or equal to any bone fracture detection algorithm proposed so far up to
best of our knowledge. The proposed algorithm open new grounds for CAD analysis of bone
fractures, reducing the load of radiology departments of public sector hospitals.
