Clinical chest CAD system for lung cancer, COPD, and osteoporosis based on MDCT images

Lung cancer kills more people than any other cancer worldwide. Lung cancer screening using low-dose CT have been performed in many countries. Comparative reading of current and past CT images is important for evaluation of pulmonary nodules in lung cancer CT screening. However, primary problem in comparative reading is mismatch of slice and nodule positions caused by lung variation. It is hard for physicians to manually match slice positions, nodule positions, and evaluate the nodule's degree of change. A system to assist smooth comparative reading is necessary. We proposed a comparative reading system for lung cancer CT screening. A distinctive feature is highly accurate matching method of region of interest based on thoracic organs registration. Pulmonary blood vessels registration using analysis of the tree structure is performed. The system is evaluated by 1 mm and 2 mm slice thickness CT images obtained from lung cancer CT screening. We show how it is useful for lung cancer CT screening. (author)