A comparative analysis of image harmonization techniques in mitigating differences in CT acquisition and reconstruction

作     者：Yadav, Anil Welland, Spencer Hoffman, John M. Kim, Grace Brown, Matthew S. Prosper, Ashley E. Aberle, Denise R. McNitt-Gray, Michael F. Hsu, William 

作者机构：Univ Calif Los Angeles Samueli Sch Engn Dept Bioengn Los Angeles CA 90095 USA UCLA Dept Radiol Sci Med & Imaging Informat Grp David Geffen Sch Med Los Angeles CA 90095 USA UCLA Ctr Comp Vis & Imaging Biomarkers Dept Radiol Sci David Geffen Sch Med Los Angeles CA 90095 USA 

出 版 物：《PHYSICS IN MEDICINE AND BIOLOGY》 (Phys. Med. Biol.)

年 卷 期：2025年第70卷第5期

页      面：055015-055015页

核心收录：

学科分类：0831[工学-生物医学工程（可授工学、理学、医学学位）] 1001[医学-基础医学(可授医学、理学学位)] 1009[医学-特种医学] 10[医学] 

基　　金：Division of Cancer Prevention, National Cancer Institutehttp://dx.doi.org/10.13039/100007316 [R01 EB031993] NIH/National Institute for Biomedical Imaging and Bioengineering [U2C CA271898] NIH/National Cancer Institute 

主　　题：Computed Tomography Image Reconstruction Harmonization Convolutional Neural Network Generative Adversarial Network 

摘      要：Objective. The study aims to systematically characterize the effect of CT parameter variations on images and lung radiomic and deep features, and to evaluate the ability of different image harmonization methods to mitigate the observed variations. Approach. A retrospective in-house sinogram dataset of 100 low-dose chest CT scans was reconstructed by varying radiation dose (100%, 25%, 10%) and reconstruction kernels (smooth, medium, sharp). A set of image processing, convolutional neural network (CNNs), and generative adversarial network-based (GANs) methods were trained to harmonize all image conditions to a reference condition (100% dose, medium kernel). Harmonized scans were evaluated for image similarity using peak signal-to-noise ratio (PSNR), structural similarity index measure (SSIM), and learned perceptual image patch similarity (LPIPS), and for the reproducibility of radiomic and deep features using concordance correlation coefficient (CCC). Main Results. CNNs consistently yielded higher image similarity metrics amongst others;for Sharp/10%, which exhibited the poorest visual similarity, PSNR increased from a mean +/- CI of 17.763 +/- 0.492 to 31.925 +/- 0.571, SSIM from 0.219 +/- 0.009 to 0.754 +/- 0.017, and LPIPS decreased from 0.490 +/- 0.005 to 0.275 +/- 0.016. Texture-based radiomic features exhibited a greater degree of variability across conditions, i.e. a CCC of 0.500 +/- 0.332, compared to intensity-based features (0.972 +/- 0.045). GANs achieved the highest CCC (0.969 +/- 0.009 for radiomic and 0.841 +/- 0.070 for deep features) amongst others. CNNs are suitable if downstream applications necessitate visual interpretation of images, whereas GANs are better alternatives for generating reproducible quantitative image features needed for machine learning applications. Significance. Understanding the efficacy of harmonization in addressing multi-parameter variability is crucial for optimizing diagnostic accuracy and a critical step toward building gen