| Summary: | 碩士 === 國立成功大學 === 醫學工程研究所 === 87 === Purpose: To improve the image quality of portal films in order to make the interpretation more easily. We also aim to develop a system for evaluation of various enhancement methods objectively.
Introduction: It is generally admitted that took a portal film remains the most important procedures of assuring geometric accuracy during radiotherapy. Unfortunately, it is difficult to interpret a portal film due to the poor image quality of these films. The most convenient method is to apply digital processing technique in order to improve quality of portal films.
Material and Methods: This study used Konica MG x-rays films and Kodak X-Omatic "PORTAL VERIFICATION" cassette as portal film preparation. A contrast-detail phantom was put ahead of above cassette and exposed with 6-MV x-rays at 7 MU. After film development, these portal films were digitized with scanner into a 800 686 pixels with 8-bit grayscale images. The enhancement methods in this study included histogram equalization, contrast-limited adaptive histogram equalization (CLAHE), frequency Gaussian filter plus contrast-limited adaptive histogram equalization (CLAHE), and complex methods (frequency Gaussian filter, spatial histogram equalization, spatial Gaussian filter, and AHE in series) four methods. The objective evaluation system was based on t-test statistical analysis. This method compared the statistical difference of grayscales compositions between foreground and background groups of pixels. The foreground group was defined as the pixels of the holes on phantom and the background group was defined as the pixels of the square surrounding these testing holes. The statistical cut-off level is defined as 5.0.
Results: The subjective image qualities of enhanced images, including all above four methods, were superior to the original image. The ratio of successful identification of holes on phantom for original, HE, CLAHE, frequency Gaussian filter plus CLAHE, and complex method were 23.2%, 21.4%, 25.0%, 51.8%, and 42.8%, respectively.
Conclusion: The enhancement methods including CLAHE with frequency Gaussian filtering and complex enhancement methods are obviously better than the original image. The t-test method is an effective way for objective evaluation of the enhancement results.
|
|---|
