Automated Segmentation Of Lung Fields From Thoracic CT

The mortality rate for lung cancer is higher than that for other kinds of cancers around the world. At the same time, it appears that the rate has been continually increasing in recent years. Conventional chest radiograms have been used to screen lung cancer and computed tomography is widely used to help detect lung cancer. The doctors analyze CT images for making a correct diagnosis and locate accurately the lung cancer. They are both bald works and need facing a great of data everyday so that making a incorrect analysis is inevitable because of a great of reading. Computer aided detection or diagnosis(CAD) is put forward for that situation and the result after it analyses much data can help the doctors to safely efficiently analyse data. Lung segmentation is a very important basic work of any CAD and it has much value of study how can quickly segment accurate lung.In this thesis, we studied the problem of the quickly lung segmentation in pulmonary computed tomography images. The main works include the following:(1) A lung segmentation is put forward after we studied many CAD for lung cancer. Its main steps include image pretreatment , removing the trachea and main bronchi , gaining lungs and repairing defective lungs . There are tow methods of repairing defective lungs in this thesis . They are falling-ball method and Snake method, and they have good effect on a measure of defect.(2) A part of the initial contour maybe is inside the segmentation goal but the other is outside when the image is processed by Snake model. For that , this thesis puts forward an advanced Snake model that loosens the condition of the initial contour very much and has good abilities to resist noises.Six cases total 169 CT images and are processed through this method. The rates of accurate segmentation lung are 93.47%(43/46) for the whole the lung and the tumour cohere a little (width no more than 10 pixels) with the part which is out of...