Research On Medical Image Line Structure Detection Algorithm Based On Geodesic

In image processing, the line structure is a form of elongated structures, Line structure detection is also called the curve structure detection in specific cases, A curve structure is a straight line or curve that has a certain scale. Line structure detection technology plays an important role in digital image processing, This paper will focus on the application of line structure detection technology in the field of medical image processing.Due to the variability of tissue characteristics in medical images, the fuzziness of the boundary between different tissues or tissues and the complexity of the distribution of the fine structure, The medical image technology has been limited and The traditional approach has been unable to meet. Local information acquisition with respect to the traditional method, The geodesic is also integrated with the global information, In the field of image processing, it is mainly used in the field of image processing, By using the characteristics of the prior information, it can predict the next step based on the existing information. In this paper, we take the midline structure of medical image as the research object, Based on geodesic theory, a new algorithm for line structure detection is proposed. To provide important reference for functional region division and focus location. The main work includes:1. Aiming at the characteristics of thin and thin lines in medical image, This paper is based on local linear growth, proposing an improved algorithm for the detection of local area growth line structure of the polling local area is. First select a seed point on the target, Using the fast marching algorithm to construct the geodesic distance field, Then, in the local area, the polling of the ground wire is carried out. To detect the branch points as the new seed point to update the local area geodesic distance field, Sequentially iterative growth detection, and finally the fracture part of the segmentation results are processed by linear interpolation, and Get a more complete segmentation results finally.2. For the detection of multi branch (tree structure) line structure, In order to reduce the computation of global search, This paper is based on the toe-finding algorithm, proposing an improved histogram peak branch location algorithm. Based on geodesic distance field, Extracting the geodesic line of the specified distance, Expand it into a histogram, Analyze the arrival time and calculate the area peaks to identify and locate the position of each branch. Then the influence area is delineated to vote the local area growth detection, Finally, the termination condition is determined by the difference of the branch end histogram and the iteration depth, ending the Multiple branch detection of tree structure.3. Aiming at the Tubular curve structure organization which contains width information, in this paper, we combinate the Oriented flux gradient, proposing a detection algorithm of tubular curve structure based on tubular geodesic line. The radius of each position point of the tubular structure in the center line is obtained by the Oriented flux gradient. Then, a fast marching algorithm is used to calculate the pipe line. At the same time, the position of the path on the path is measured and the radius of the curve structure is estimated. Obtaining The envelope of the curve family of the center line and the corresponding radius, the Tubular curve structure.4. In order to verify the effectiveness of the algorithm, we establish a comprehensive evaluation system in this paper. By using the Lobe and Lung Analysis 2011 (LOLA 11) and CT data in MRI database and Digital Imaging and Communications in Medicine(DICOM) database, the evaluation and comparison of qualitative and quantitative detection results are carried out. And calculate the TPR, FDR, detection accuracy and the average running time of these four evaluation criteria.By comparing with other algorithms as well as with Truth Ground, The algorithm presented in this paper has good performance in segmentation. Especially in the anti noise performance has certain advantages, and it can meet the needs of clinical applications...