Detection of Normality/Pathology on Chest Radiographs using LBP

Since the discovery of X-rays and their applications, medical imaging has been a great help for radiologists in the diagnosis of diseases. In recent years, there has been a great effort in the computer vision community in the research of systems for the analysis and extraction of useful information from medical images. In this scenario, we have designed, implemented and validated a novel method to detect normality/pathology in chest radiographs, which constitutes the core of a computer-aided detection (CADe) system. Although the problem addressed is very complex and little explored, our approach is completely automatic, starting from the location of areas of interest using template matching techniques. The main novelty of our contribution is the application of a transformation known as local binary patterns (LBP) to these areas. LBP histograms are then used as input features for a classification system, which is ultimately responsible for the decision of normality/pathology. The results of our preliminary experiments are quite promising. With success rates in the best cases close to 90%, we believe that increased performance could be obtained with bigger training sets and more advanced classification systems, which will make these systems to be fully viable in the near future.