Cancer and cancer related diseases account for a significant amount of mortality in the world. One of the leading causes of cancer related deaths is lung cancer. One of the best ways to battle cancer is to diagnose it in its developing stages. In the case of lung cancer, this early diagnosis can be accomplished by chest radiographs. However, the evaluation of X-ray images may prove to be insufficient. To combat this, digital chest tomosynthesis is being developed. Promising increased visibility with insignificant increase in irradiation. The process creates many images though, and doctors around the world are usually overworked, leaving them prone to making errors, possibly letting preventable conditions develop. If the diagnosis process could be automated, that would reduce the load on the doctors and potentially increase early detection rates.
This thesis explores the possibility of accomplishing this task with my current capabilities. Introduces the problem, explores some possibilities, and ultimately makes some remarks about future directions.
No real solution was found, but some potentially interesting ideas were investigated that may prove useful to solve some of the underlying problems that are inherently part of chest tomosynthesis lung region detection.