| Pneumoconiosis is a global occupational disease, which is cause by long-term exposure to high concentrations of dust, so the disease is very high in the coal mining industry. According to statistics, the incidence of pneumoconiosis has accounted for 84.22% of all occupational incidence, so shocking figures have to cause our attention for pneumoconiosis prevention and care of pneumoconiosis patients. This paper will examine the application of computer aided diagnostic technique to the detection of pneumoconiosis, to reduce the doctor work burden, improve the diagnostic accuracy and reduce the costs of patient diagnoses. This paper proposes a new method, the study area is located at portions between the ribs, rather than the entire chest radiograph, so as to make the research targets more accurately. The main research contents includes the following steps:First, we use the graph-cut segment the lung, and use Guassin unsharp masking threthold algorithm to segment the regions of ribs, then the intercostal region is divided into 32 X 32 pixels pieces. Second, we use the calculation of these small images GLCM four directions, and the analysis of the eigenvalues of these matrices to extract the texture features of the image. Third, using the method of random forest algorithm classification small images, then analyzing the results of this combination to arrive at whether X-ray chest radiograph is the pneumoconiosis X-ray chest radiograph. Finally, According to the above three parts of research, design a computer-aided pneumoconiosis diagnosis system. In this paper, using 150 chest X-ray chest radiograph on the system tested, the results show that the classification accuracy reached 91%,true positive reached 96%, true negative reached 86%, all of this show that the results of this study can effectively distinguish between normal X-ray chest radiograph and pneumoconiosis X-ray chest radiograph, could help doctors diagnosed pneumoconiosis more efficiently. |
PDF Full Text Request