Search

Purpose: Deep learning (DL) is rapidly finding applications in low-dose CT image denoising. While having the potential to improve the image quality (IQ) over the filtered back projection method (FBP) and produce images quickly, performance generalizability of the data-driven DL methods is not fully understood yet. The main purpose of this work is to investigate the performance generalizability of a low-dose CT image denoising neural network in data acquired under different scan conditions, particularly relating to these three parameters: reconstruction kernel, slice thickness, and dose (noise) level. A secondary goal is to identify any underlying data property associated with the CT scan settings that might help predict the generalizability of the denoising network., Methods: We select the residual encoder-decoder convolutional neural network (REDCNN) as an example of a low-dose CT image denoising technique in this work. To study how the network generalizes on the three imaging parameters, we grouped the CT volumes in the Low-Dose Grand Challenge (LDGC) data into three pairs of training datasets according to their imaging parameters, changing only one parameter in each pair. We trained REDCNN with them to obtain six denoising models. We test each denoising model on datasets of matching and mismatching parameters with respect to its training sets regarding dose, reconstruction kernel, and slice thickness, respectively, to evaluate the denoising performance changes. Denoising performances are evaluated on patient scans, simulated phantom scans, and physical phantom scans using IQ metrics including mean-squared error (MSE), contrast-dependent modulation transfer function (MTF), pixel-level noise power spectrum (pNPS), and low-contrast lesion detectability (LCD)., Results: REDCNN had larger MSE when the testing data were different from the training data in reconstruction kernel, but no significant MSE difference when varying slice thickness in the testing data. REDCNN trained with quarter-dose data had slightly worse MSE in denoising higher-dose images than that trained with mixed-dose data (17%-80%). The MTF tests showed that REDCNN trained with the two reconstruction kernels and slice thicknesses yielded images of similar image resolution. However, REDCNN trained with mixed-dose data preserved the low-contrast resolution better compared to REDCNN trained with quarter-dose data. In the pNPS test, it was found that REDCNN trained with smooth-kernel data could not remove high-frequency noise in the test data of sharp kernel, possibly because the lack of high-frequency noise in the smooth-kernel data limited the ability of the trained model in removing high-frequency noise. Finally, in the LCD test, REDCNN improved the lesion detectability over the original FBP images regardless of whether the training and testing data had matching reconstruction kernels., Conclusions: REDCNN is observed to be poorly generalizable between reconstruction kernels, more robust in denoising data of arbitrary dose levels when trained with mixed-dose data, and not highly sensitive to slice thickness. It is known that reconstruction kernel affects the in-plane pNPS shape of a CT image, whereas slice thickness and dose level do not, so it is possible that the generalizability performance of this CT image denoising network highly correlates to the pNPS similarity between the testing and training data., (© 2021 American Association of Physicists in Medicine. This article has been contributed to by US Government employees and their work is in the public domain in the USA.)

Purpose: To measure radiation absorbed dose and its distribution in an anthropomorphic head phantom under clinically representative scan conditions in three widely used computed tomography (CT) scanners, and to relate those dose values to metrics such as high-contrast resolution, noise, and contrast-to-noise ratio (CNR) in the American College of Radiology CT accreditation phantom., Methods: By inserting optically stimulated luminescence dosimeters (OSLDs) in the head of an anthropomorphic phantom specially developed for CT dosimetry (University of Florida, Gainesville), we measured dose with three commonly used scanners (GE Discovery CT750 HD, Siemens Definition, Philips Brilliance 64) at two different clinical sites (Walter Reed National Military Medical Center, National Institutes of Health). The scanners were set to operate with the same data-acquisition and image-reconstruction protocols as used clinically for typical head scans, respective of the practices of each facility for each scanner. We also analyzed images of the ACR CT accreditation phantom with the corresponding protocols. While the Siemens Definition and the Philips Brilliance protocols utilized only conventional, filtered back-projection (FBP) image-reconstruction methods, the GE Discovery also employed its particular version of an adaptive statistical iterative reconstruction (ASIR) algorithm that can be blended in desired proportions with the FBP algorithm. We did an objective image-metrics analysis evaluating the modulation transfer function (MTF), noise power spectrum (NPS), and CNR for images reconstructed with FBP. For images reconstructed with ASIR, we only analyzed the CNR, since MTF and NPS results are expected to depend on the object for iterative reconstruction algorithms., Results: The OSLD measurements showed that the Siemens Definition and the Philips Brilliance scanners (located at two different clinical facilities) yield average absorbed doses in tissue of 42.6 and 43.1 mGy, respectively. The GE Discovery delivers about the same amount of dose (43.7 mGy) when run under similar operating and image-reconstruction conditions, i.e., without tube current modulation and ASIR. The image-metrics analysis likewise showed that the MTF, NPS, and CNR associated with the reconstructed images are mutually comparable when the three scanners are run with similar settings, and differences can be attributed to different edge-enhancement properties of the applied reconstruction filters. Moreover, when the GE scanner was operated with the facility's scanner settings for routine head exams, which apply 50% ASIR and use only approximately half of the 100%-FBP dose, the CNR of the images showed no significant change. Even though the CNR alone is not sufficient to characterize the image quality and justify any dose reduction claims, it can be useful as a constancy test metric., Conclusions: This work presents a straightforward method to connect direct measurements of CT dose with objective image metrics such as high-contrast resolution, noise, and CNR. It demonstrates that OSLD measurements in an anthropomorphic head phantom allow a realistic and locally precise estimation of magnitude and spatial distribution of dose in tissue delivered during a typical CT head scan. Additional objective analysis of the images of the ACR accreditation phantom can be used to relate the measured doses to high contrast resolution, noise, and CNR.