Research Of Key Algorithms For Computer-aided Diagnosis Application Of Pulmonary Cancers And System Implemention

Pulmonary cancer has become one of most harmfully malignant tumor to human, and the leading cause of deaths among all type of cancer, early detection and treatment of pulmonary cancers are effective ways of improving the survival rate. With the advent of the low dose multislices CT, it has become an important tool for early diagnosis of pulmonary cancer, however, the CT data which the radiologists need to diagnose has increased sharply. Computer-aided diagnosis (CAD) technic has become an important area of mediacal imaging and computer science to help the radiologists detecting the lung nodules more intuitive、to cut down the ragiologists’work and reduce the misdiagnosis rate. Study shows that CAD system plays an active role in improving the diagnostic accuracy and reducing the rate of misdiagnosis.In mediacal diagnosis, using the information-rich3D data instead of simple2D images has become trends of mediacal image processing. Computer graphics processor (Graphics Processing Unit, GPU) is a highly parallel stream processor, it is different with the serial computation mode of CPU, GPU has a stronger ability of floating point computing, and now programmable parallel computing platform has come true. And GPU programming used in image processing can break the bottleneck of three-dimensional mediacal image processing speed, meet the real-time requirement of practical application.This thesis has research some key algorithm about lung cancer CAD system, mainly include four aspects:(1).The low-dose lung CT images have been processed, use the Otsu and Ball-snake algrothim to segment the lung parenchyma, analyz and compare the results. Use the Fuzzy C mean clustering algorithm to obtain more complete lung segmention template, remove the excess interference region, improved ball-rolling method to repair corrosive edge of polumonary parenchyma. This algorthim can automatically get a more complete lung parenchyma region.(2).The nodule features in three-dimensional space have been analysed. Research three-dimensional multi-scale Hessian matrix to filter and disanti-geometric diffusion filter in three-dimensional space, and detect the nodules base on sharp index. (3).After analysis the existing volume redering algorithm and the method of acceleration. Accelerating of raycasting algorithm base on parallel processing of GPU has been reasearched. The algorithm can quickly reconstruct large three-dimensional volume data. Meet the requirement of real-time processing in practical application.(4). A lung cancer CAD system platform has been established, it can process the DIOCM data of medical image. The key technologies of lung segmentation, three-dimensional reconstruction and lung nodule detection are put in this platform to meet the needs of practical applications。Experiments show that the proposed lung segmentation algorthim can automatically obtain the pulmonary parenchyma, lung nodules detection algorthims get high accuracy, three-dimensional reconstruction achieves the requiment of real-time, meet the requirements of practical applications, and have practical application value.