Research On The Segmentation Of Lung Region Based On CT Images

In recent years, as computer technology continues to update and develop, the advantages of large medical equipment in clinical diagnostic imaging have increasingly reflected. Wherein, X-ray computer tomography (CT) technology used in the lung imaging is the most common instrument for the lungs and surrounding tissue to provide a wide dynamic range and high spatial resolution images, which is widely used in assessing and diagnosing many lung diseases, such as lung cancer, pulmonary embolism and lung nodule detection. After performing chest computed tomography, the automatic segmentation of lung extraction is the basic premise of computer-aided diagnosis (CAD) technology to chest CT images for subsequent analysis, which is of great importance for the early detection of lung disease and lung function analysis of the local area (such as analysis of the density of the lungs, trachea and lung diaphragm, etc.), and reducing a lot of unnecessary computation time and un-correlation result. In addition to being an important preprocessing step of computer-aided diagnosis applications, the first step of many three-dimensional image data visualization algorithm also need an extracted lung region.The requirement of accurate segmentation of lung regions is twofold. Firstly, pathology, which continues to promote the development of chest CT computer-aided diagnosis technology, still occupies a dominant position, having an important impact on the lung segmentation results, making segmentation method set to adapt to clinical expected shape and size of the lung region in CT images. Secondly, the final lungs segmentation result must be completed, to ensure that the edge of the lungs can be accurately detected. Because some abnormal tissue(such as pulmonary nodules) is existed in the most peripheral lung region, if the entire lungs do not be completely segmented, these abnormal tissue will be lost and affects subsequent quantitative analysis of lung volume data, causing disease misjudgment, directly affects clinical practical value. Therefore, the design of the lung segmentation algorithm should be universally adaptive, which can segment lung parenchyma from the chest CT image accurately and quickly, and overcome omission problem during the segmentation process for to be used in normal clinical cases. For special needs, modify the algorithm according to the different circumstances.As medical images with big noise, complex image structure and individual differences, the application of commonly used image segmentation methods often can not accomplish segmentation tasks alone. In response to these problems, in order to achieve smoothed segmented lung region CT images accurately and efficiently, improve clinical efficiency of computer-aided diagnosis, this paper carried out relevant research on the segmentation algorithm of the lungs, mainly discussed the key technologies in the process of lung parenchyma segmentation, proposed a automatically lung region segmentation method based on fused multi-algorithm for chest CT images:(1) Preprocess to enhance image quality. Firstly select median filtering and histogram equalization method in the original CT image to reduce image noise and improve image contrast, provide a guarantee for the subsequent segmentation process.(2) The initial segmentation. Successive two region growing technique using in lung CT images has removed the background and tracheal/bronchial interference information, extracted lung images. In this step, according to the anatomy characteristics of the lung CT image, seed point has been selected automatically, reducing errors caused by manual intervention.(3) Separation of left and right lungs. According to the lungs segmentation results in last step, using line scan method in the lung upper field to get the number of edge points and taking the middle points of lines to fit for generating boundary information, successfully separate the left and right lungs. This step is operated in the local area, which can reduce processing time and improve the segmenting efficiency.(4) Repair the edge of lungs. For the holes in the internal lung parenchyma and gaps on the edge of lungs caused by missing nodules, this paper uses mathematical morphology to refine, which overcomes the independence of the traditional repairing method—rolling ball algorithm, which results depending on the selected radius size. Finally get a complete and accurate lung image.By testing this method in25groups of chest CT sequence images, then compare and evaluate the experimental results between which are obtaiained by an expert with three comparative evaluation criteria, the final results provened by automatic computer segmentation and those manual segmentation results obtd that this method can separate the lung parenchyma out from the background and achieve a satisfactory results similar with manual segmentation results for subsequent research and process.