Shalchian Bahareh, Rajabi Hossein, Soltanian-Zadeh Hamid
Department of Medical Physics, Tarbiat Modares University, Tehran, Iran.
J Nucl Med Technol. 2009 Dec;37(4):223-8. doi: 10.2967/jnmt.109.067454. Epub 2009 Nov 13.
An efficient method for tomographic imaging in nuclear medicine is PET. Higher sensitivity, higher spatial resolution, and more accurate quantification are advantages of PET, in comparison to SPECT. However, a high noise level in the images limits the diagnostic utility of PET. Noise removal in nuclear medicine is traditionally based on the Fourier decomposition of the images. This method is based on frequency components, irrespective of the spatial location of the noise or signal. The wavelet transform presents a solution by providing information on frequency contents while retaining spatial information, alleviating the shortcoming of Fourier transformation. Thus, wavelet transformation has been extensively used for noise reduction, edge detection, and compression.
In this research, SimSET software was used for simulation of PET images of the nonuniform rational B-spline-based cardiac-torso phantom. The images were acquired using 250 million counts in 128 x 128 matrices. For a reference image, we acquired an image with high counts (6 billion). Then, we reconstructed these images using our own software developed in a commercially available program. After image reconstruction, a 250-million-count image (noisy image or test image) and a reference image were normalized, and then root mean square error was used to compare the images. Next, we wrote and applied denoising programs. These programs were based on using 54 different wavelets and 4 methods. Denoised images were compared with the reference image using root mean square error.
Our results indicate stationary wavelet transformation and global thresholding are more efficient at noise reduction than are other methods that we investigated.
Wavelet transformation is a useful method for denoising simulated PET images. Noise reduction using this transform and loss of high-frequency information are simultaneous with each other. It seems we should attend to mutual agreement between noise reduction and visual quality of the image.
正电子发射断层扫描（PET）是核医学中一种高效的断层成像方法。与单光子发射计算机断层扫描（SPECT）相比，PET具有更高的灵敏度、更高的空间分辨率和更准确的定量分析能力。然而，图像中的高噪声水平限制了PET的诊断效用。核医学中的噪声去除传统上基于图像的傅里叶分解。该方法基于频率成分，而不考虑噪声或信号的空间位置。小波变换通过在保留空间信息的同时提供频率内容信息来提供一种解决方案，从而缓解了傅里叶变换的缺点。因此，小波变换已广泛用于降噪、边缘检测和压缩。
在本研究中，使用SimSET软件对基于非均匀有理B样条的心脏躯干模型的PET图像进行模拟。图像在128×128矩阵中使用2.5亿计数采集。对于参考图像，我们采集了高计数（60亿）的图像。然后，我们使用在商业可用程序中开发的自己的软件重建这些图像。图像重建后，对2.5亿计数的图像（噪声图像或测试图像）和参考图像进行归一化，然后使用均方根误差比较图像。接下来，我们编写并应用了去噪程序。这些程序基于使用54种不同的小波和4种方法。使用均方根误差将去噪后的图像与参考图像进行比较。
我们的结果表明，平稳小波变换和全局阈值处理在降噪方面比我们研究的其他方法更有效。
小波变换是一种用于去噪模拟PET图像的有用方法。使用这种变换进行降噪与高频信息的损失是同时发生的。似乎我们应该关注降噪与图像视觉质量之间的相互协调。
