Research And Implementation Of Lung Nodules Detection System Based On The Lung Segmentation

Lung cancer is a malignant tumor with the highest rate of prevalence andmortality worldwidely, cause more than million deaths every year. Early detection andtreatment of lesions in the early stage of lung cancer can improve the survival rate ofpatients suffering from lung cancer greatly. The characteristic of early stage lung canceris lung nodules. Usually, lung nodule is lesions like circle with diameter less than3cmin the lung.The traditional diagnosis of lung cancer is analysising lung CT scan imagesby doctors with the naked eyes, then get the final ccnclusion with the help of medicalknowledge. With the development of modern CT technology, a single CT slice becomesthinner, the resolution of CT images becomes increasingly higher, which with anincreasing numbers of slices in one scan. Facing mass CT images from a large numberof patients everyday, the doctor’s workload is increasing, and more and more possiblymakes mistakes, or even delays the treatment of disease. To overcome this difficultyand free doctors from the heavy burden of processing slices, researchers started todevelop lung CAD (computer assist diagnose), known as computer-aided diagnosissystem, which can help radiologists automatically or semi-automatically analysis anddiagnosis of CT images, also can greatly improve the efficiency of the radiologist.Lung nodules with the characteristics of complex shape, and easy to adhere topulmonary tissue, make it difficult to make accurate judgments even for experiencedphysicians. Now,we still need the help of observing the medical history as well aspuncture and other medical means to the final diagnosis. So, Assisted detection can onlylabel out the possible pulmonary nodule regions to provide an auxiliary to help doctor’sdiagnosis. The first step of lung nodules detection is lung segmentation. The so-calledlung segmentation is to determine the boundaries of the lungs. Only with the segmentthe lung parenchyma, can we process targeted treatment on the lung CT images. Thelung CT images as study object, automatic detection of lung lesions as target, we finishthe following work:(1) With the study of theoretical knowledge and literature, we proposed a methodof lung segmentation. The main steps follow below: image preprocessing, binaryzation,removal of the trachea/bronchus, segment the left and right lung, and the repair of lungedge. Focus on the limitation of the commonly used global threshold method inbinarization, we find peak between the crest as threshold; morphological repair and fast boundary march method are used to repair the boundary of lung. The result shows, wecan get a satisfying segmentation.(2) Base on the segmention of lung image, we detect out the lung nodules, andstatistics the nodules information according to the gray and morphologicalcharacteristics of lung nodules.(3) Coding in Matlab, we develop a Lung Nodules Detection System, with thesefunctions: reading CT files; showing CT information; lung segmentation; labelling outlung nodules; showing nodules information and DataBase operation to assist doctors todiagnosis.The results show that our system has its practical value.