Caelum Research Corporation proposes to develop a computer-aided detection (CAD) system for detecting minimal tuberculosis (TB) on chest radiographs. Among infectious diseases, TB is the leading killer of adults in the world today and has infected one third of the world's population. In addition, TB currently kills more adults each year than AIDS, malaria, and tropical diseases combined, and kill almost 170,000 children annually. TB will also become of increasing importance in the United States, due to the increasing frequency of multi-drug resistance. Finally, AIDS and TB form a lethal combination, each speeding the other's progress. TB is the leading cause of death among people who have AIDS, accounting for about 15% of AIDS deaths worldwide. In this proposal, Caelum describes computer-aided methods for identifying and eventually removing (through image processing) the shadows created by the bones of the chest. This proposal also describes methods for identifying the small nodules and subtle changes in lung texture that occur with minimal tuberculosis. Our work will be guided by the findings reported by experienced radiologists. In this Phase I project, we will use an advanced bi-direction wavelet technique to extract the ribs and clavicle bones. We will also use a nodular modeling and enhancement technique to facilitate the analysis of lung structures. Based on the enhanced chest radiograph, we will develop computer algorithms, using pattern analysis and neural networks, in order to detect radiographic signs of TB infection.
The specific aims of this Phase I program are: (1) delineation of posterior ribs in chest radiography; (2) delineation of the clavicle; (3) intensity reduction of clavicle bones and ribs using contralateral rib registration and modeling techniques; and (4) analyses of TB patterns and textures on chest radiographs. The proposed project represents Caelum's commitment to continue working with the clinical and research, communities to develop CAD systems not only for lung cancer but also for other lung diseases. Early detection of those with minimal TB has been shown to be the most effective method of TB control. Caelum's TB-CAD system will be an inexpensive way to enhance the detection of minimal TB. It will allow early treatment for disease control and should help decrease mortality from this common, lethal disease.
