Computer-aided bone fracture identification based on ultrasound images

Clinical diagnosis is increasingly being supported by complementary imaging methods, many of them based on ionizing radiation such as X-ray. Bone fractures are traditionally diagnosed through X-ray. On the other hand ultrasound imaging represents a cheap and fast alternative, particularly when stress fractures are considered. Framed by a computer-aided system a 3-stage procedure for automatic identification of bone fractures on ultrasound images is proposed. Combined methods of speckle reduction, bone line detection and fracture identification, tuned by well performed variables, enabled correct classification of bone fractures in 89% of the 44 analyzed images. Stress fractures were all correctly identified in comparison with the clinical records therefore proving the ultrasound imaging primacy against X-ray for this type of pathology and encouraging its broader application on other type of bones.