Jaeger.info

My new project is computer-aided diagnosis on radiographs. The goal of this project is to be able to capture and process digital radiographs in remote and difficult to access areas using a portable x-ray machine, and to assist local doctors with various degrees of experience in their diagnosis. Special emphasis lies in the development of a reliable system to automatically differentiate normal from pathologic chest x-rays to screen for the presence of tuberculosis or other lung disease. Despite some preliminary work published in the last ten years, computer-aided diagnosis in chest radiography can be considered a largely unsolved problem. To solve this problem, and to reach a performance relevant to clinical practice, extensive research is required to develop new image processing techniques that can cope with the low-contrast and noisy images encountered in practice. Other problems that need to be addressed include, but are not limited to segmentation and feature computation. The textural abnormalities in chest radiographs, in particular those associated with pulmonary tuberculosis, need to be quantified in a way that allows reliable discrimination between normal and abnormal cases. Typical abnormal findings in chest radiographs, in particular findings associated with pulmonary tuberculosis, are for instance, nodules, pleural effusions, or cavitations. It is very unlikely that a single feature type will be powerful enough to capture all the abnormal findings encountered in clinical practice. Different feature sets need to be experimented with, including features already proposed for computer-aided medical diagnosis, as well as features from other image processing domains and entirely new features yet to be developed.