Search

OBJECTIVE: The aim of this study was to evaluate a narrowly trained convolutional neural network (CNN) denoising algorithm when applied to images reconstructed differently than training data set. METHODS: A residual CNN was trained using 10 noise inserted examinations. Training images were reconstructed with 275 mm of field of view (FOV), medium smooth kernel (D30), and 3 mm of thickness. Six examinations were reserved for testing; these were reconstructed with 100 to 450 mm of FOV, smooth to sharp kernels, and 1 to 5 mm of thickness. RESULTS: When test and training reconstruction settings were not matched, there was either reduced denoising efficiency or resolution degradation. Denoising efficiency was reduced when FOV was decreased or a smoother kernel was used. Resolution loss occurred when the network was applied to an increased FOV, sharper kernel, or decreased image thickness. CONCLUSIONS: The CNN denoising performance was degraded with variations in FOV, kernel, or decreased thickness. Denoising performance was not affected by increased thickness.

This study evaluated the clinical utility of a phantom-based convolutional neural network noise reduction framework for whole-body-low-dose CT skeletal surveys. The CT exams of ten patients with multiple myeloma were retrospectively analyzed. Exams were acquired with routine whole-body-low-dose CT protocol and projection noise insertion was used to simulate 25% dose exams. Images were reconstructed with either iterative reconstruction or filtered back projection with convolutional neural network post-processing. Diagnostic quality and structure visualization were blindly rated (subjective scale ranging from 0 [poor] to 100 [excellent]) by three musculoskeletal radiologists for iterative reconstruction and convolutional neural network images at routine whole-body-low-dose and 25% dose CT. For the diagnostic quality rating, the convolutional neural network outscored iterative reconstruction at routine whole-body-low-dose CT (convolutional neural network: 95 ± 5, iterative reconstruction: 85 ± 8) and at the 25% dose level (convolutional neural network: 79 ± 10, iterative reconstruction: 22 ± 13). Convolutional neural network applied to 25% dose was rated inferior to iterative reconstruction applied to routine dose. Similar trends were observed in rating experiments focusing on structure visualization. Results indicate that the phantom-based convolutional neural network noise reduction framework can improve visualization of critical structures within CT skeletal surveys. At matched dose level, the convolutional neural network outscored iterative reconstruction for all conditions studied. The image quality improvement of convolutional neural network applied to 25% dose indicates a potential for dose reduction; however, the 75% dose reduction condition studied is not currently recommended for clinical implementation.

PURPOSE: The aims of this study were to develop a single-scan dual-contrast protocol for biphasic liver imaging with 2 intravenous contrast agents (iodine and gadolinium) and to evaluate its effectiveness in an exploratory swine study using a photon-counting detector computed tomography (PCD-CT) system. MATERIALS AND METHODS: A dual-contrast CT protocol was developed for PCD-CT to simultaneously acquire 2 phases of liver contrast enhancement, with the late arterial phase enhanced by 1 contrast agent (iodine-based) and the portal venous phase enhanced by the other (gadolinium-based). A gadolinium contrast bolus (gadobutrol: 64 mL, 8 mL/s) and an iodine contrast bolus (iohexol: 40 mL, 5 mL/s) were intravenously injected in the femoral vein of a healthy domestic swine, with the second injection initiated after 17 seconds from the beginning of the first injection; PCD-CT image acquisition was performed 12 seconds after the beginning of the iodine contrast injection. A convolutional neural network (CNN)–based denoising technique was applied to PCD-CT images to overcome the inherent noise magnification issue in iodine/gadolinium decomposition task. Iodine and gadolinium material maps were generated using a 3-material decomposition method in image space. A set of contrast samples (mixed iodine and gadolinium) was attached to the swine belly; quantitative accuracy of material decomposition in these inserts between measured and true concentrations was calculated using root mean square error. An abdominal radiologist qualitatively evaluated the delineation of arterial and venous vasculatures in the swine liver using iodine and gadolinium maps obtained using the dual-contrast PCD-CT protocol. RESULTS: The iodine and gadolinium samples attached to the swine were quantified with root mean square error values of 0.75 mg/mL for iodine and 0.45 mg/mL for gadolinium from the contrast material maps derived from the denoised PCD-CT images. Hepatic arteries containing iodine and veins containing gadolinium in the swine liver could be clearly visualized. Compared with the original images, better distinctions between 2 liver phases were achieved using CNN denoising, with approximately 60% to 80% noise reduction in contrast material maps acquired with the denoised PCD-CT images compared with the original images. CONCLUSIONS: Simultaneous biphasic liver imaging in a single multienergy PCD-CT acquisition using a dual-contrast (iodine and gadolinium) injection protocol and CNN denoising was demonstrated in a swine study, where the enhanced hepatic arteries (containing iodine) and the enhanced hepatic veins (containing gadolinium) could be clearly visualized and delineated in the swine liver.

In this study, we describe a systematic approach to optimize deep-learning-based image processing algorithms using random search. The optimization technique is demonstrated on a phantom-based noise reduction training framework; however, the techniques described can be applied generally for other deep learning image processing applications. The parameter space explored included number of convolutional layers, number of filters, kernel size, loss function, and network architecture (either U-Net or ResNet). A total of 100 network models were examined (50 random search, 50 ablation experiments). Following the random search, ablation experiments resulted in a very minor performance improvement indicating near optimal settings were found during the random search. The top performing network architecture was a U-Net with 4 pooling layers, 64 filters, 3×3 kernel size, ELU activation, and a weighted feature reconstruction loss (0.2×VGG + 0.8×MSE). Relative to the low-dose input image, the CNN reduced noise by 90%, reduced RMSE by 34%, and increased SSIM by 76% on six patient exams reserved for testing. The visualization of hepatic and bone lesions was greatly improved following noise reduction.

Purpose: We developed a deep learning method to reduce noise and beam-hardening artifact in virtual monoenergetic image (VMI) at low x-ray energy levels. Approach: An encoder–decoder type convolutional neural network was implemented with customized inception modules and in-house-designed training loss (denoted as Incept-net), to directly estimate VMI from multi-energy CT images. Images of an abdomen-sized water phantom with varying insert materials were acquired from a research photon-counting-detector CT. The Incept-net was trained with image patches ([Formula: see text]) extracted from the phantom data, as well as synthesized, random-shaped numerical insert materials. The whole CT images ([Formula: see text]) with the remaining real insert materials that were unseen in network training were used for testing. Seven contrast-enhanced abdominal CT exams were used for preliminary evaluation of Incept-net generalizability over anatomical background. Mean absolute percentage error (MAPE) was used to evaluate CT number accuracy. Results: Compared to commercial VMI software, Incept-net largely suppressed beam-hardening artifact and reduced noise (53%) in phantom study. Incept-net presented comparable CT number accuracy at higher-density ([Formula: see text]-value [0.0625, 0.999]) and improved it at lower-density inserts ([Formula: see text]) with overall MAPE: Incept-net [2.9%, 4.6%]; commercial-VMI [6.7%, 10.9%]. In patient images, Incept-net suppressed beam-hardening artifact and reduced noise (up to 50%, [Formula: see text]). Conclusion: In this preliminary study, Incept-net presented the potential to improve low-energy VMI quality.

This article shows how to train a convolutional neural network to reduce noise in CT images, although the principles apply to medical and nonmedical images; authors also explore mathematical and visually weighted loss functions to adjust the appearance.