Ohashi Kohei, Ishida Takayuki
Division of Health Science, Graduate School of Medicine, Osaka University.
Nihon Hoshasen Gijutsu Gakkai Zasshi. 2016;72(10):961-969. doi: 10.6009/jjrt.2016_JSRT_72.10.961.
In radiography, when a blurred image caused by patient motion was acquired, radiologists retake an image as needed. However, retaking an image leads to extra radiation exposure to patients and reducing work efficiency. This study proposes the deblurring algorithm for blurred images caused by patient motion in radiography. In the proposed algorithm, we first take a video using an optical device during radiography. Second, we calculate the optical flow between each frame, and estimate a point spread function (PSF) based on the optical flows. Finally, we restore the blurred image by deconvolution processing. In this study, blurred images with the blur width from 1.0 mm to 5.0 mm at 0.5 mm intervals were obtained by using own moving body phantom, and applied proposed algorithm for each blurred image. To evaluate the algorithm, we measured blur area and structural similarity (SSIM) of the blurred images and deblurred images, and compared the values. As a result, a significant decrease in blur area and a significant increase in SSIM were confirmed in each blur condition. These results suggest the usefulness of the proposed algorithm.
在放射成像中，当获取到由患者移动导致的模糊图像时，放射科医生会根据需要重新拍摄图像。然而，重新拍摄图像会使患者受到额外的辐射暴露，并降低工作效率。本研究提出了一种针对放射成像中由患者移动导致的模糊图像的去模糊算法。在所提出的算法中，我们首先在放射成像过程中使用光学设备拍摄一段视频。其次，我们计算每一帧之间的光流，并基于这些光流估计一个点扩散函数（PSF）。最后，我们通过反卷积处理来恢复模糊图像。在本研究中，使用自行制作的移动人体模型获得了模糊宽度从1.0毫米到5.0毫米、间隔为0.5毫米的模糊图像，并对每个模糊图像应用所提出的算法。为了评估该算法，我们测量了模糊图像和去模糊图像的模糊区域和结构相似性（SSIM），并比较了这些值。结果表明，在每种模糊条件下，模糊区域都有显著减小，SSIM有显著增加。这些结果表明了所提出算法的有效性。
