Itagaki Koji, Miyake Kanae K, Tanoue Minori, Oishi Tae, Kataoka Masako, Kawashima Masahiro, Toi Masakazu, Nakamoto Yuji
Division of Clinical Radiology Service, Kyoto University Hospital, Kyoto, Japan.
Department of Advanced Medical Imaging Research, Graduate School of Medicine Kyoto University, Kyoto , Japan.
Asia Ocean J Nucl Med Biol. 2023;11(2):145-157. doi: 10.22038/AOJNMB.2023.71598.1501.
This study aimed to create a deep learning (DL)-based denoising model using a residual neural network (Res-Net) trained to reduce noise in ring-type dedicated breast positron emission tomography (dbPET) images acquired in about half the emission time, and to evaluate the feasibility and the effectiveness of the model in terms of its noise reduction performance and preservation of quantitative values compared to conventional post-image filtering techniques.
Low-count (LC) and full-count (FC) PET images with acquisition durations of 3 and 7 minutes, respectively, were reconstructed. A Res-Net was trained to create a noise reduction model using fifteen patients' data. The inputs to the network were LC images and its outputs were denoised PET (LC + DL) images, which should resemble FC images. To evaluate the LC + DL images, Gaussian and non-local mean (NLM) filters were applied to the LC images (LC + Gaussian and LC + NLM, respectively). To create reference images, a Gaussian filter was applied to the FC images (FC + Gaussian). The usefulness of our denoising model was objectively and visually evaluated using test data set of thirteen patients. The coefficient of variation (CV) of background fibroglandular tissue or fat tissue were measured to evaluate the performance of the noise reduction. The SUV and SUV of lesions were also measured. The agreement of the SUV measurements was evaluated by Bland-Altman plots.
The CV of background fibroglandular tissue in the LC + DL images was significantly lower (9.102.76) than the CVs in the LC (13.60 3.66) and LC + Gaussian images (11.51 3.56). No significant difference was observed in both SUV and SUV of lesions between LC + DL and reference images. For the visual assessment, the smoothness rating for the LC + DL images was significantly better than that for the other images except for the reference images.
Our model reduced the noise in dbPET images acquired in about half the emission time while preserving quantitative values of lesions. This study demonstrates that machine learning is feasible and potentially performs better than conventional post-image filtering in dbPET denoising.
本研究旨在使用残差神经网络（Res-Net）创建基于深度学习（DL）的去噪模型，该模型经过训练以减少在大约一半发射时间内采集的环形专用乳腺正电子发射断层扫描（dbPET）图像中的噪声，并与传统的图像后滤波技术相比，从降噪性能和定量值保留方面评估该模型的可行性和有效性。
分别重建采集持续时间为3分钟和7分钟的低计数（LC）和全计数（FC）PET图像。使用15名患者的数据训练Res-Net以创建降噪模型。网络的输入是LC图像，其输出是去噪后的PET（LC + DL）图像，该图像应类似于FC图像。为了评估LC + DL图像，将高斯滤波器和非局部均值（NLM）滤波器分别应用于LC图像（分别为LC +高斯和LC + NLM）。为了创建参考图像，将高斯滤波器应用于FC图像（FC +高斯）。使用13名患者的测试数据集对我们的去噪模型的有用性进行客观和视觉评估。测量背景纤维腺组织或脂肪组织的变异系数（CV）以评估降噪性能。还测量了病变的SUV和SUV。通过Bland-Altman图评估SUV测量值的一致性。
LC + DL图像中背景纤维腺组织的CV显著低于LC（13.60±3.66）和LC +高斯图像（11.51±3.56）中的CV（9.10±2.76）。LC + DL与参考图像之间在病变的SUV和SUV方面均未观察到显著差异。对于视觉评估，LC + DL图像的平滑度评分明显优于除参考图像之外的其他图像。
我们的模型在大约一半发射时间内采集的dbPET图像中降低了噪声，同时保留了病变的定量值。本研究表明，机器学习在dbPET去噪中是可行的，并且可能比传统的图像后滤波表现更好。
Zotero 插件
