Applying 3D Models of Thoracic Structures to the Interpretation of 2D Chest X-rays
The aim of the research is to apply 3D models of thoracic structures to the interpretation of 2D chest X-rays in order to explore the potential for computer aided detection.
In 2011, 8.7 million people fell ill with tuberculosis (TB), including 1.1 million with HIV and an estimated 1.4 million died from the disease. Although incidence is low in the UK, increasing mobility has led to a sharp rise and some London boroughs now have incidence rates of over 100 per 100,000. One of the primary tools in the diagnosis of TB is chest radiography. However there are difficulties with this technique: the interpretative skills it requires are often absent in low resource settings, and even when skilled radiologists are available, the images can be hard to interpret. We believe that these difficulties can be, at least in part, addressed with computer aided detection (CAD). We think CAD has several potential uses: supporting screening programmes where large numbers of images are taken from people in high risk populations with the aim of identifying TB early; supporting interpretation of radiographs taken to investigate suspected cases of TB; and supporting identification of TB from X rays taken for clinical purposes where a diagnosis of TB had not previously been suspected. We believe the sensitivity and specificity of CAD can be considerably enhanced by using temporal sequences of images. Such sequences are increasingly available for individuals who are screened more than once, or who have repeated X Rays to investigate clinical symptoms. Analysis of sequences of images allows temporal changes to be detected and used as cues in discriminating between disease and non-disease cases.