Leveraging Semantic Information for Sonographic Wrist Fracture Assessment Within Children

Abstract

The use of ultrasound for the diagnosis of pediatric distal forearmfractures provides a radiation-free, rapid, and inexpensive alternative to radiography. Computer-aided examination and diagnosis support may contribute to the increasing popularity of fracture sonography. Although machine learning approaches are considered the tool of choice for medical image processing, the success of datadriven methods is highly dependent on the quality and quantity of image data. Both conditions are not necessarily met in the field of pediatric bone sonography, so supporting measures for the application of deep learning techniques are required. One possible solution is the incorporation of additional semantic information. In this work, we investigate to what extent the use of existing state-of-the-art frameworks together with segmentations of anatomical structures can increase the classification accuracy of the detection of distal forearm fractures in children ausing ultrasound.