Ground Truth Annotation, Analysis and Release of Data set of Radiographic Images of Bone Fractures

Bone is the tough, locomotive tissue of the body which is subjected to fractures very often. Advancements in Computer Aided Diagnosis (CAD) have leaded the researchers to develop robust algorithms to reduce human error in assessing different types of pathologies. This paper focuses on the data set release of the plain radiographic images of bone fractures which include both “Ground Truth Annotated” and “Non-Annotated Images”. The data was collected from the Radiology Department of Pakistan Institute of Medical Sciences, Islamabad, Pakistan in the form of digital images which were 8,345 in number, out of them 735 images were of bone fracture. The bone fractures were annotated manually to provide “truth” for the spatial location of bone fracture. The type of bone affected, area affected of the bone fractured, type of fracture, potential laterality of the affected bone and x-ray projection had also been identified and reported in the form of a data sheet. These features are expected to help training classifiers for CAD analysis of bone fractures. Frequency analysis of essential features had been performed to accentuate the data. The data set will provide grounds for the research community to manipulate for various CAD operations which may include estimation of precision of the developed CAD operators