Iterative reconstruction for limited-angle CT using implicit neural representation.

Objective.Limited-angle computed tomography (CT) presents a challenge due to its ill-posed nature. In such scenarios, analytical reconstruction methods often exhibit severe artifacts. To tackle this inverse problem, several supervised deep learning-based approaches have been proposed. However, they are constrained by limitations such as generalization issue and the difficulty of acquiring a large amount of paired CT images.Approach.In this work, we propose an iterative neural reconstruction framework designed for limited-angle CT. By leveraging a coordinate-based neural representation, we formulate tomographic reconstruction as a convex optimization problem involving a deep neural network. We then employ differentiable projection layer to optimize this network by minimizing the discrepancy between the predicted and measured projection data. In addition, we introduce a prior-based weight initialization method to ensure the network starts optimization with an informed initial guess. This strategic initialization significantly improves the quality of iterative reconstruction by stabilizing the divergent behavior in ill-posed neural fields. Our method operates in a self-supervised manner, thereby eliminating the need for extensive data.Main results.The proposed method outperforms other iterative and learning-based methods. Experimental results on XCAT and Mayo Clinic datasets demonstrate the effectiveness of our approach in restoring anatomical features as well as structures. This finding was substantiated by visual inspections and quantitative evaluations using NRMSE, PSNR, and SSIM. Moreover, we conduct a comprehensive investigation into the divergent behavior of iterative neural reconstruction, thus revealing its suboptimal convergence when starting from scratch. In contrast, our method consistently produced accurate images by incorporating an initial estimate as informed initialization.Significance.This work showcases the feasibility to reconstruct high-fidelity CT images from limited-angle x-ray projections. The proposed methodology introduces a novel data-free approach to enhance medical imaging, holding promise across various clinical applications.

Adapting low-dose CT denoisers for texture preservation using zero-shot local noise-level matching.

BACKGROUND: On enhancing the image quality of low-dose computed tomography (LDCT), various denoising methods have achieved meaningful improvements. However, they commonly produce over-smoothed results; the denoised images tend to be more blurred than the normal-dose targets (NDCTs). Furthermore, many recent denoising methods employ deep learning(DL)-based models, which require a vast amount of CT images (or image pairs). PURPOSE: Our goal is to address the problem of over-smoothed results and design an algorithm that works regardless of the need for a large amount of training dataset to achieve plausible denoising results. Over-smoothed images negatively affect the diagnosis and treatment since radiologists had developed clinical experiences with NDCT. Besides, a large-scale training dataset is often not available in clinical situations. To overcome these limitations, we propose locally-adaptive noise-level matching (LANCH), emphasizing the output should retain the same noise-level and characteristics to that of the NDCT without additional training. METHODS: We represent the NDCT image as the pixel-wisely weighted sum of an over-smoothed output from off-the-shelf denoiser (OSD) and the difference between the LDCT image and the OSD output. Herein, LANCH determines a 2D ratio map (i.e., pixel-wise weight matrix) by locally matching the noise-level of output and NDCT, where the LDCT-to-NDCT device flux (mAs) ratio reveals the NDCT noise-level. Thereby, LANCH can preserve important details in LDCT, and enhance the sharpness of the noise-free regions. Note that LANCH can enhance any LDCT denoisers without additional training data (i.e., zero-shot). RESULTS: The proposed method is applicable to any OSD denoisers, reporting significant texture plausibility development over the baseline denoisers in quantitative and qualitative manners. It is surprising that the denoising accuracy achieved by our method with zero-shot denoiser was comparable or superior to that of the best training-based denoisers; our result showed 1% and 33% gains in terms of SSIM and DISTS, respectively. Reader study with experienced radiologists shows significant image quality improvements, a gain of + 1.18 on a five-point mean opinion score scale. CONCLUSIONS: In this paper, we propose a technique to enhance any low-dose CT denoiser by leveraging the fundamental physical relationship between the x-ray flux and noise variance. Our method is capable of operating in a zero-shot condition, which means that only a single low-dose CT image is required for the enhancement process. We demonstrate that our approach is comparable or even superior to supervised DL-based denoisers that are trained using numerous CT images. Extensive experiments illustrate that our method consistently improves the performance of all tested LDCT denoisers.

Direct estimation of the noise power spectrum from patient data to generate synthesized CT noise for denoising network training.

BACKGROUND: Developing a deep-learning network for denoising low-dose CT (LDCT) images necessitates paired computed tomography (CT) images acquired at different dose levels. However, it is challenging to obtain these images from the same patient. PURPOSE: In this study, we introduce a novel approach to generate CT images at different dose levels. METHODS: Our method involves the direct estimation of the quantum noise power spectrum (NPS) from patient CT images without the need for prior information. By modeling the anatomical NPS using a power-law function and estimating the quantum NPS from the measured NPS after removing the anatomical NPS, we create synthesized quantum noise by applying the estimated quantum NPS as a filter to random noise. By adding synthesized noise to CT images, synthesized CT images can be generated as if these are obtained at a lower dose. This leads to the generation of paired images at different dose levels for training denoising networks. RESULTS: The proposed method accurately estimates the reference quantum NPS. The denoising network trained with paired data generated using synthesized quantum noise achieves denoising performance comparable to networks trained using Mayo Clinic data, as justified by the mean-squared-error (MSE), structural similarity index (SSIM)and peak signal-to-noise ratio (PSNR) scores. CONCLUSIONS: This approach offers a promising solution for LDCT image denoising network development without the need for multiple scans of the same patient at different doses.

Impact of using sinogram domain data in the super-resolution of CT images on diagnostic information.

BACKGROUND: In recent times, deep-learning-based super-resolution (DL-SR) techniques for computed tomography (CT) images have shown outstanding results in terms of full-reference image quality (FR-IQ) metrics (e.g., root mean square error and structural similarity index metric), which assesses IQ by measuring its similarity to the high-resolution (HR) image. In addition, IQ can be evaluated via task-based IQ (Task-IQ) metrics that evaluate the ability to perform specific tasks. Ironically, most proposed image domain-based SR techniques are not possible to improve a Task-IQ metric, which assesses the amount of information related to diagnosis. PURPOSE: In the case of CT imaging systems, sinogram domain data can be utilized for SR techniques. Therefore, this study aims to investigate the impact of utilizing sinogram domain data on diagnostic information restoration ability. METHODS: We evaluated three DL-SR techniques: using image domain data (Image-SR), using sinogram domain data (Sinogram-SR), and using sinogram as well as image domain data (Dual-SR). For Task-IQ evaluation, the Rayleigh discrimination task was used to evaluate diagnostic ability by focusing on the resolving power aspect, and an ideal observer (IO) can be used to perform the task. In this study, we used a convolutional neural network (CNN)-based IO that approximates the IO performance. We compared the IO performances of the SR techniques according to the data domain to evaluate the discriminative information restoration ability. RESULTS: Overall, the low-resolution (LR) and SR exhibit lower IO performances compared with that of HR owing to their degraded discriminative information when detector binning is used. Next, between the SR techniques, Image-SR does not show superior IO performances compared to the LR image, but Sinogram-SR and Dual-SR show superior IO performances than the LR image. Furthermore, in Sinogram-SR, we confirm that FR-IQ and IO performance are positively correlated. These observations demonstrate that sinogram domain upsampling improves the representation ability for discriminative information in the image domain compared to the LR and Image-SR. CONCLUSIONS: Unlike Image-SR, Sinogram-SR can improve the amount of discriminative information present in the image domain. This demonstrates that to improve the amount of discriminative information on the resolving power aspect, it is necessary to employ sinogram domain processing.

Full 3-D modulation transfer function estimation of tomosynthesis system using modified Richardson-Lucy deconvolution.

BACKGROUND: Tomosynthesis systems are three-dimensional (3-D) medical imaging devices that operate over limited acquisition angles using low radiation dosages. To measure the spatial resolution performance of a tomosynthesis system, the modulation transfer function (MTF) is widely used as a quantitative evaluation metric. PURPOSE: We previously introduced a method to estimate the full 3-D MTF of a cone-beam computed tomography system using two-dimensional (2-D) Richardson-Lucy (RL) deconvolution with Tikhonov-Miller regularization. However, this method can not be applied directly to estimate the 3-D MTF of a tomosynthesis system, since the unique artifacts (i.e., shadow artifacts, spreading tails, directional blurring, and high-level noise) of the system produce several errors that lower the estimation performance. Varying positions of the negative pixels due to shadow artifacts and spreading tails cause inconsistent deconvolution performances at each of the directional projections, and the severe noise in the reconstructed images cause noise amplification during estimation. This work proposes several modifications to the previous method to resolve the inconsistent performance and noise amplification errors to increase the full 3-D MTF estimation accuracy. METHODS: Three modifications were introduced to the 2-D RL deconvolution to prevent estimation errors and improve MTF estimation performance: non-negativity relaxation function, cost function to terminate the iterative process of RL deconvolution, and regularization strength for noise control. To validate the effectiveness of the proposed modifications, we reconstructed sphere phantoms from simulation and experimental tomosynthesis studies in the iso-center and offset-center positions as well as estimated the full 3-D MTFs using the previous and proposed methods. We compared the 3-D render images, central plane images, and center profiles of the estimated 3-D MTFs and calculated the full widths at half and tenth maximum for quantitative evaluation. RESULTS: The previous method cannot estimate the full 3-D MTF of a tomosynthesis system; its inaccurate negative pixel relaxation produces circular-shaped errors, and the mean squared error based simple cost function for termination causes inconsistent estimation at each directional projection to diminish the clear edges of the low-frequency drop and missing sample regions. Noise amplification from lack of noise regularization is also seen in the previous method results. Compared to the previous method, the proposed method shows superior estimation performance at reducing errors in both the simulation and experimental studies regardless of object position. The proposed method preserves the low-frequency drop, missing sample regions from the limited acquisition angles, and missing cone region from the offset-center position; the estimated MTFs also show FWHM and FWTM values close to those of the ideal MTFs than with the previous method. CONCLUSIONS: This work presents a method to estimate the full 3-D MTF of a tomosynthesis system. The proposed modifications prevent circular-shaped errors and noise amplification due to the geometry for limited acquisition angles and high noise levels. Compared to our previous method, the proposed scheme show better performance for estimating the 3-D MTF of the tomosynthesis system.

Deep learning-based multimodal fusion network for segmentation and classification of breast cancers using B-mode and elastography ultrasound images.

Ultrasonography is one of the key medical imaging modalities for evaluating breast lesions. For differentiating benign from malignant lesions, computer-aided diagnosis (CAD) systems have greatly assisted radiologists by automatically segmenting and identifying features of lesions. Here, we present deep learning (DL)-based methods to segment the lesions and then classify benign from malignant, utilizing both B-mode and strain elastography (SE-mode) images. We propose a weighted multimodal U-Net (W-MM-U-Net) model for segmenting lesions where optimum weight is assigned on different imaging modalities using a weighted-skip connection method to emphasize its importance. We design a multimodal fusion framework (MFF) on cropped B-mode and SE-mode ultrasound (US) lesion images to classify benign and malignant lesions. The MFF consists of an integrated feature network (IFN) and a decision network (DN). Unlike other recent fusion methods, the proposed MFF method can simultaneously learn complementary information from convolutional neural networks (CNNs) trained using B-mode and SE-mode US images. The features from the CNNs are ensembled using the multimodal EmbraceNet model and DN classifies the images using those features. The experimental results (sensitivity of 100 ± 0.00% and specificity of 94.28 ± 7.00%) on the real-world clinical data showed that the proposed method outperforms the existing single- and multimodal methods. The proposed method predicts seven benign patients as benign three times out of five trials and six malignant patients as malignant five out of five trials. The proposed method would potentially enhance the classification accuracy of radiologists for breast cancer detection in US images.

Power-law spectrum-based objective function to train a generative adversarial network with transfer learning for the synthetic breast CT image.

Objective.This paper proposes a new objective function to improve the quality of synthesized breast CT images generated by the GAN and compares the GAN performances on transfer learning datasets from different image domains.Approach.The proposed objective function, named beta loss function, is based on the fact that x-ray-based breast images follow the power-law spectrum. Accordingly, the exponent of the power-law spectrum (beta value) for breast CT images is approximately two. The beta loss function is defined in terms of L1 distance between the beta value of synthetic images and validation samples. To compare the GAN performances for transfer learning datasets from different image domains, ImageNet and anatomical noise images are used in the transfer learning dataset. We employ styleGAN2 as the backbone network and add the proposed beta loss function. The patient-derived breast CT dataset is used as the training and validation dataset; 7355 and 212 images are used for network training and validation, respectively. We use the beta value evaluation and Fréchet inception distance (FID) score for quantitative evaluation.Main results.For qualitative assessment, we attempt to replicate the images from the validation dataset using the trained GAN. Our results show that the proposed beta loss function achieves a more similar beta value to real images and a lower FID score. Moreover, we observe that the GAN pretrained with anatomical noise images achieves better equality than ImageNet for beta value evaluation and FID score. Finally, the beta loss function with anatomical noise as the transfer learning dataset achieves the lowest FID score.Significance.Overall, the GAN using the proposed beta loss function with anatomical noise images as the transfer learning dataset provides the lowest FID score among all tested cases. Hence, this work has implications for developing GAN-based breast image synthesis methods for medical imaging applications.

Chest tomosynthesis deblurring using CNN with deconvolution layer for vertebrae segmentation.

BACKGROUND: Limited scan angles cause severe distortions and artifacts in reconstructed tomosynthesis images when the Feldkamp-Davis-Kress (FDK) algorithm is used for the purpose, which degrades clinical diagnostic performance. These blurring artifacts are fatal in chest tomosynthesis images because precise vertebrae segmentation is crucial for various diagnostic analyses, such as early diagnosis, surgical planning, and injury detection. Moreover, because most spinal pathologies are related to vertebral conditions, the development of methods for accurate and objective vertebrae segmentation in medical images is an important and challenging research area. PURPOSE: The existing point-spread-function-(PSF)-based deblurring methods use the same PSF in all sub-volumes without considering the spatially varying property of tomosynthesis images. This increases the PSF estimation error, thus further degrading the deblurring performance. However, the proposed method estimates the PSF more accurately by using sub-CNNs that contain a deconvolution layer for each sub-system, which improves the deblurring performance. METHODS: To minimize the effect of the spatially varying property, the proposed deblurring network architecture comprises four modules: (1) block division module, (2) partial PSF module, (3) deblurring block module, and (4) assembling block module. We compared the proposed DL-based method with the FDK algorithm, total-variation iterative reconstruction with GP-BB (TV-IR), 3D U-Net, FBPConvNet, and two-phase deblurring method. To investigate the deblurring performance of the proposed method, we evaluated its vertebrae segmentation performance by comparing the pixel accuracy (PA), intersection-over-union (IoU), and F-score values of reference images to those of the deblurred images. Also, pixel-based evaluations of the reference and deblurred images were performed by comparing their root mean squared error (RMSE) and visual information fidelity (VIF) values. In addition, 2D analysis of the deblurred images were performed by artifact spread function (ASF) and full width half maximum (FWHM) of the ASF curve. RESULTS: The proposed method was able to recover the original structure significantly, thereby further improving the image quality. The proposed method yielded the best deblurring performance in terms of vertebrae segmentation and similarity. The IoU, F-score, and VIF values of the chest tomosynthesis images reconstructed using the proposed SV method were 53.5%, 28.7%, and 63.2% higher, respectively, than those of the images reconstructed using the FDK method, and the RMSE value was 80.3% lower. These quantitative results indicate that the proposed method can effectively restore both the vertebrae and the surrounding soft tissue. CONCLUSIONS: We proposed a chest tomosynthesis deblurring technique for vertebrae segmentation by considering the spatially varying property of tomosynthesis systems. The results of quantitative evaluations indicated that the vertebrae segmentation performance of the proposed method was better than those of the existing deblurring methods.

Sparsier2Sparse: Self-supervised convolutional neural network-based streak artifacts reduction in sparse-view CT images.

BACKGROUND: Sparse-view computed tomography (CT) has attracted a lot of attention for reducing both scanning time and radiation dose. However, sparsely-sampled projection data generate severe streak artifacts in the reconstructed images. In recent decades, many sparse-view CT reconstruction techniques based on fully-supervised learning have been proposed and have shown promising results. However, it is not feasible to acquire pairs of full-view and sparse-view CT images in real clinical practice. PURPOSE: In this study, we propose a novel self-supervised convolutional neural network (CNN) method to reduce streak artifacts in sparse-view CT images. METHODS: We generate the training dataset using only sparse-view CT data and train CNN based on self-supervised learning. Since the streak artifacts can be estimated using prior images under the same CT geometry system, we acquire prior images by iteratively applying the trained network to given sparse-view CT images. We then subtract the estimated steak artifacts from given sparse-view CT images to produce the final results. RESULTS: We validated the imaging performance of the proposed method using extended cardiac-torso (XCAT) and the 2016 AAPM Low-Dose CT Grand Challenge dataset from Mayo Clinic. From the results of visual inspection and modulation transfer function (MTF), the proposed method preserved the anatomical structures effectively and showed higher image resolution compared to the various streak artifacts reduction methods for all projection views. CONCLUSIONS: We propose a new framework for streak artifacts reduction when only the sparse-view CT data are given. Although we do not use any information of full-view CT data for CNN training, the proposed method achieved the highest performance in preserving fine details. By overcoming the limitation of dataset requirements on fully-supervised-based methods, we expect that our framework can be utilized in the medical imaging field.

Strategy to implement a convolutional neural network based ideal model observer via transfer learning for multi-slice simulated breast CT images.

Objective.In this work, we propose a convolutional neural network (CNN)-based multi-slice ideal model observer using transfer learning (TL-CNN) to reduce the required number of training samples.Approach.To train model observers, we generate simulated breast CT image volumes that are reconstructed using the FeldkampDavisKress algorithm with a ramp and Hanning-weighted ramp filter. The observer performance is evaluated on the background-known-statistically (BKS)/signal-known-exactly task with a spherical signal, and the BKS/signal-known-statistically task with random signal generated by the stochastic grown method. We compare the detectability of the CNN-based model observer with that of conventional linear model observers for multi-slice images (i.e. a multi-slice channelized Hotelling observer (CHO) and volumetric CHO). We also analyze the detectability of the TL-CNN for different numbers of training samples to examine its performance robustness to a limited number of training samples. To further analyze the effectiveness of transfer learning, we calculate the correlation coefficients of filter weights in the CNN-based multi-slice model observer.Main results.When using transfer learning for the CNN-based multi-slice ideal model observer, the TL-CNN provides the same performance with a 91.7% reduction in the number of training samples compared to that when transfer learning is not used. Moreover, compared to the conventional linear model observer, the proposed CNN-based multi-slice model observers achieve 45% higher detectability in the signal-known-statistically detection tasks and 13% higher detectability in the SKE detection tasks. In correlation coefficient analysis, it is observed that the filters in most of the layers are highly correlated, demonstrating the effectiveness of the transfer learning for multi-slice model observer training.Significance.Deep learning-based model observers require large numbers of training samples, and the required number of training samples increases as the dimensions of the image (i.e. the number of slices) increase. With applying transfer learning, the required number of training samples is significantly reduced without performance drop.

Convolutional neural network-based model observer for signal known statistically task in breast tomosynthesis images.

BACKGROUND: Since human observer studies are resource-intensive, mathematical model observers are frequently used to assess task-based image quality. The most common implementation of these model observers assume that the signal information is exactly known. However, these tasks cannot thoroughly represent situations where the signal information is not exactly known in terms of size and shape. PURPOSE: Considering the limitations of the tasks for which signal information is exactly known, we proposed a convolutional neural network (CNN)-based model observer for signal known statistically (SKS) and background known statistically (BKS) detection tasks in breast tomosynthesis images. METHODS: A wide parameter search was conducted from six different acquisition angles (i.e., 10°, 20°, 30°, 40°, 50°, and 60°) within the same dose level (i.e., 2.3 mGy) under two separate acquisition schemes: (1) constant total number of projections, and (2) constant angular separation between projections. Two different types of signals: spherical (i.e., SKE tasks) and spiculated (i.e., SKS tasks) were used. The detection performance of the CNN-based model observer was compared with that of the Hotelling observer (HO) instead of the IO. Pixel-wise gradient-weighted class activation mapping (pGrad-CAM) map was extracted from each reconstructed tomosynthesis image to provide an intuitive understanding of the trained CNN-based model observer. RESULTS: The CNN-based model observer achieved a higher detection performance compared to that of the HO for all tasks. Moreover, the improvement in its detection performance was greater for SKS tasks compared to that for SKE tasks. These results demonstrated that the addition of nonlinearity improved the detection performance owing to the variation of the background and signal. Interestingly, the pGrad-CAM results effectively localized the class-specific discriminative region, further supporting the quantitative evaluation results of the CNN-based model observer. In addition, we verified that the CNN-based model observer required fewer images to achieve the detection performance of the HO. CONCLUSIONS: In this work, we proposed a CNN-based model observer for SKS and BKS detection tasks in breast tomosynthesis images. Throughout the study, we demonstrated that the detection performance of the proposed CNN-based model observer was superior to that of the HO.

Utilization of an attentive map to preserve anatomical features for training convolutional neural-network-based low-dose CT denoiser.

BACKGROUND: The purpose of a convolutional neural network (CNN)-based denoiser is to increase the diagnostic accuracy of low-dose computed tomography (LDCT) imaging. To increase diagnostic accuracy, there is a need for a method that reflects the features related to diagnosis during the denoising process. PURPOSE: To provide a training strategy for LDCT denoisers that relies more on diagnostic task-related features to improve diagnostic accuracy. METHODS: An attentive map derived from a lesion classifier (i.e., determining lesion-present or not) is created to represent the extent to which each pixel influences the decision by the lesion classifier. This is used as a weight to emphasize important parts of the image. The proposed training method consists of two steps. In the first one, the initial parameters of the CNN denoiser are trained using LDCT and normal-dose CT image pairs via supervised learning. In the second one, the learned parameters are readjusted using the attentive map to restore the fine details of the image. RESULTS: Structural details and the contrast are better preserved in images generated by using the denoiser trained via the proposed method than in those generated by conventional denoisers. The proposed denoiser also yields higher lesion detectability and localization accuracy than conventional denoisers. CONCLUSIONS: A denoiser trained using the proposed method preserves the small structures and the contrast in the denoised images better than without it. Specifically, using the attentive map improves the lesion detectability and localization accuracy of the denoiser.

Carbon nanotube-based multiple source C-arm CT system: feasibility study with prototype system.

To extend the field of view while reducing dimensions of the C-arm, we propose a carbon nanotube (CNT)-based C-arm computed tomography (CT) system with multiple X-ray sources. A prototype system was developed using three CNT X-ray sources, enabling a feasibility study. Geometry calibration and image reconstruction were performed to improve the quality of image acquisition. However, the geometry of the prototype system led to projection truncation for each source and an overlap region of object area covered by each source in the two-dimensional Radon space, necessitating specific corrective measures. We addressed these problems by implementing truncation correction and applying weighting techniques to the overlap region during the image reconstruction phase. Furthermore, to enable image reconstruction with a scan angle less than 360°, we designed a weighting function to solve data redundancy caused by the short scan angle. The accuracy of the geometry calibration method was evaluated via computer simulations. We also quantified the improvements in reconstructed image quality using mean-squared error and structural similarity. Moreover, detector lag correction was applied to address the afterglow observed in the experimental data obtained from the prototype system. Our evaluation of image quality involved comparing reconstructed images obtained with and without incorporating the geometry calibration results and images with and without lag correction. The outcomes of our simulation study and experimental investigation demonstrated the efficacy of our proposed geometry calibration, image reconstruction method, and lag correction in reducing image artifacts.

A streak artifact reduction algorithm in sparse-view CT using a self-supervised neural representation.

PURPOSE: Sparse-view computed tomography (CT) has been attracting attention for its reduced radiation dose and scanning time. However, analytical image reconstruction methods suffer from streak artifacts due to insufficient projection views. Recently, various deep learning-based methods have been developed to solve this ill-posed inverse problem. Despite their promising results, they are easily overfitted to the training data, showing limited generalizability to unseen systems and patients. In this work, we propose a novel streak artifact reduction algorithm that provides a system- and patient-specific solution. METHODS: Motivated by the fact that streak artifacts are deterministic errors, we regenerate the same artifacts from a prior CT image under the same system geometry. This prior image need not be perfect but should contain patient-specific information and be consistent with full-view projection data for accurate regeneration of the artifacts. To this end, we use a coordinate-based neural representation that often causes image blur but can greatly suppress the streak artifacts while having multiview consistency. By employing techniques in neural radiance fields originally proposed for scene representations, the neural representation is optimized to the measured sparse-view projection data via self-supervised learning. Then, we subtract the regenerated artifacts from the analytically reconstructed original image to obtain the final corrected image. RESULTS: To validate the proposed method, we used simulated data of extended cardiac-torso phantoms and the 2016 NIH-AAPM-Mayo Clinic Low-Dose CT Grand Challenge and experimental data of physical pediatric and head phantoms. The performance of the proposed method was compared with a total variation-based iterative reconstruction method, naive application of the neural representation, and a convolutional neural network-based method. In visual inspection, it was observed that the small anatomical features were best preserved by the proposed method. The proposed method also achieved the best scores in the visual information fidelity, modulation transfer function, and lung nodule segmentation. CONCLUSIONS: The results on both simulated and experimental data suggest that the proposed method can effectively reduce the streak artifacts while preserving small anatomical structures that are easily blurred or replaced with misleading features by the existing methods. Since the proposed method does not require any additional training datasets, it would be useful in clinical practice where the large datasets cannot be collected.

Convolutional neural network-based metal and streak artifacts reduction in dental CT images with sparse-view sampling scheme.

PURPOSE: Sparse-view sampling has attracted attention for reducing the scan time and radiation dose of dental cone-beam computed tomography (CBCT). Recently, various deep learning-based image reconstruction techniques for sparse-view CT have been employed to produce high-quality image while effectively reducing streak artifacts caused by the lack of projection views. However, most of these methods do not fully consider the effects of metal implants. As sparse-view sampling strengthens the artifacts caused by metal objects, simultaneously reducing both metal and streak artifacts in sparse-view CT images has been challenging. To solve this problem, in this study, we propose a novel framework. METHODS: The proposed method was based on the normalized metal artifact reduction (NMAR) method, and its performance was enhanced using two convolutional neural networks (CNNs). The first network reduced the initial artifacts while preserving the fine details to generate high-quality priors for NMAR processing. Subsequently, the second network was employed to reduce the streak artifacts after NMAR processing of sparse-view CT data. To validate the proposed method, we generated training and test data by computer simulations using both extended cardiac-torso (XCAT) and clinical data sets. RESULTS: Visual inspection and quantitative evaluations demonstrated that the proposed method effectively reduced both metal and streak artifacts while preserving the details of anatomical structures compared with the conventional metal artifact reduction methods. CONCLUSIONS: We propose a framework for reconstructing accurate CT images in metal-inserted sparse-view CT. The proposed method reduces streak artifacts from both metal objects and sparse-view sampling while recovering the anatomical details, indicating the feasibility of fast-scan dental CBCT imaging.

Effect of optical blurring of X-ray source on breast tomosynthesis image quality: Modulation transfer function, anatomical noise power spectrum, and signal detectability perspectives.

We investigated the effect of the optical blurring of X-ray source on digital breast tomosynthesis (DBT) image quality using well-designed DBT simulator and table-top experimental systems. To measure the in-plane modulation transfer function (MTF), we used simulated sphere phantom and Teflon sphere phantom and generated their projection data using two acquisition modes (i.e., step-and-shoot mode and continuous mode). After reconstruction, we measured the in-plane MTF using reconstructed sphere phantom images. In addition, we measured the anatomical noise power spectrum (aNPS) and signal detectability. We constructed simulated breast phantoms with a 50% volume glandular fraction (VGF) of breast anatomy using the power law spectrum and inserted spherical objects with 1 mm, 2 mm, and 5 mm diameters as breast masses. Projection data were acquired using two acquisition modes, and in-plane breast images were reconstructed using the Feldkamp-Davis-Kress (FDK) algorithm. For the experimental study, we used BR3D breast phantom with 50% VGF and obtained projection data using a table-top experimental system. To compare the detection performance of the two acquisition modes, we calculated the signal detectability using the channelized Hotelling observer (CHO) with Laguerre-Gauss (LG) channels. Our results show that spatial resolution of in-plane image in continuous mode was degraded due to the optical blurring of X-ray source. This blurring effect was reflected in aNPS, resulting in large ß values. From a signal detectability perspective, the signal detectability in step-and-shoot mode is higher than that in continuous mode for small spherical signals but not large spherical signals. Although the step-and-shoot mode has disadvantage in terms of scan time compared to the continuous mode, scanning in step-and-shoot mode is better for detecting small signals, indicating that there is a tradeoff between scan time and image quality.

Method to estimate fan-beam CT noise power spectrum using two basis functions with a limited number of noise realizations.

PURPOSE: The noise power spectrum (NPS) plays a key role in image quality (IQ) evaluation as it can be used for predicting detection performance or calculating detective quantum efficiency (DQE). Traditionally, the NPS is estimated by ensemble averaging multiple realizations of noise-only images. However, the estimation error increases when there are a limited number of images. Since the estimation error directly affects the image quality (IQ) index, an accurate NPS estimation method is required. Recent works have proposed NPS estimation methods using the radial one-dimensional (1D) NPS as the basis; however, when sharp kernels are used during image reconstruction, these methods cannot accurately estimate the amplitude of each angular spoke of the 2D NPS composed of different cutoff frequencies determined from the complementary projection magnification factors for different spatial regions. In this work, we propose a 2D NPS estimation method that reflects the accurate amplitude of each angular spoke for fan-beam CT images. METHODS: An angular spoke of the 2D NPS is composed of two basis functions with different cutoff frequencies determined from the complementary projection magnification factors. The proposed method estimates these two weighting factors for each basis function by minimizing the mean-squared error (MSE) between the 2D NPS estimated from 10 noise realizations. Two noise profiles and two types of apodization filters (i.e., rectangular and Hanning) were used to reconstruct the noise-only images. To examine the nonstationary noise property of fan-beam CT images, the 2D NPS was estimated at three different local regions. The estimation accuracy of the proposed method was further improved by estimating the approximate weighting factors with sinusoidal functions, considering that the weighting factors vary slowly throughout the view angles. Regression orders of 1 to 4 were used during these estimations. The approximate weighting factors were then multiplied with each of the basis functions to estimate the 2D NPS. The normalized mean-squared error (NMSE) was used as an index to compare the performance of each NPS estimation method, with the analytical 2D NPS as the reference. Further validation was performed using XCAT phantom data. RESULTS: We observed that the 2D NPS estimated using two basis functions reflected the accurate amplitude of each angular spoke, whereas the 2D NPS estimated using the radial 1D NPS as the basis could not. The 2D NPS estimated by applying the approximate weighting factors showed improved performance compared with that estimated using two basis functions. In addition, unlike the view-independent noise cases, where a lower regression order showed higher estimation performance, a higher regression order showed higher estimation performance in the view-dependent noise cases. CONCLUSIONS: In this work, we propose a 2D NPS estimation method that reflects the accurate amplitude of each angular spoke for fan-beam CT images using two basis functions. We observed that the proposed 2D NPS estimation method using two basis functions achieved better estimation performance compared with the method using the radial 1D NPS as the basis.

Two-phase learning-based 3D deblurring method for digital breast tomosynthesis images.

In digital breast tomosynthesis (DBT) systems, projection data are acquired from a limited number of angles. Consequently, the reconstructed images contain severe blurring artifacts that might heavily degrade the DBT image quality and cause difficulties in detecting lesions. In this study, we propose a two-phase learning approach for artifact compensation in a coarse-to-fine manner to mitigate blurring artifacts effectively along all viewing directions of the DBT image volume (i.e., along the axial, coronal, and sagittal planes) to improve the detection performance of lesions. The proposed method employs a convolutional neural network model comprising two submodels/phases, with Phase 1 performing three-dimensional (3D) deblurring and Phase 2 performing additional 2D deblurring. To investigate the effects of loss functions on the proposed model's deblurring performance, we evaluated several loss functions, such as the pixel-based loss function, adversarial-based loss function, and perception-based loss function. Compared with the DBT image, the mean squared error of the image and the root mean squared errors of the gradient of the image decreased by 82.8% and 44.9%, respectively, and the contrast-to-noise ratio increased by 183.4% in the in-focus plane. We verified that the proposed method sequentially restored the missing frequency components as the DBT images were processed through the Phase 1 and Phase 2 steps. These results indicate that the proposed method performs effective 3D deblurring, significantly reducing the blurring artifacts in the in-focus plane and other planes of the DBT image, thus improving the detection performance of lesions.

Low-dose CT denoising via convolutional neural network with an observer loss function.

PURPOSE: Convolutional neural network (CNN)-based denoising is an effective method for reducing complex computed tomography (CT) noise. However, the image blur induced by denoising processes is a major concern. The main source of image blur is the pixel-level loss (e.g., mean squared error [MSE] and mean absolute error [MAE]) used to train a CNN denoiser. To reduce the image blur, feature-level loss is utilized to train a CNN denoiser. A CNN denoiser trained using visual geometry group (VGG) loss can preserve the small structures, edges, and texture of the image.However, VGG loss, derived from an ImageNet-pretrained image classifier, is not optimal for training a CNN denoiser for CT images. ImageNet contains natural RGB images, so the features extracted by the ImageNet-pretrained model cannot represent the characteristics of CT images that are highly correlated with diagnosis. Furthermore, a CNN denoiser trained with VGG loss causes bias in CT number. Therefore, we propose to use a binary classification network trained using CT images as a feature extractor and newly define the feature-level loss as observer loss. METHODS: As obtaining labeled CT images for training classification network is difficult, we create labels by inserting simulated lesions. We conduct two separate classification tasks, signal-known-exactly (SKE) and signal-known-statistically (SKS), and define the corresponding feature-level losses as SKE loss and SKS loss, respectively. We use SKE loss and SKS loss to train CNN denoiser. RESULTS: Compared to pixel-level losses, a CNN denoiser trained using observer loss (i.e., SKE loss and SKS loss) is effective in preserving structure, edge, and texture. Observer loss also resolves the bias in CT number, which is a problem of VGG loss. Comparing observer losses using SKE and SKS tasks, SKS yields images having a more similar noise structure to reference images. CONCLUSIONS: Using observer loss for training CNN denoiser is effective to preserve structure, edge, and texture in denoised images and prevent the CT number bias. In particular, when using SKS loss, denoised images having a similar noise structure to reference images are generated.

Geometry calibration and image reconstruction for carbon-nanotube-based multisource and multidetector CT.

Conventional intraoperative computed tomography (CT) has a long scan time, degrading the image quality. Its large size limits the position of a surgeon during surgery. Therefore, this study proposes a CT system comprising of eight carbon-nanotube (CNT)-based x-ray sources and 16 detector modules to solve these limitations. Gantry only requires 45° of rotation to acquire the whole projection, reducing the scan time to 1/8 compared to the full rotation. Moreover, the volume and scan time of the system can be significantly reduced using CNT sources with a small volume and short pulse width and placing a heavy and large high-voltage generator outside the gantry. We divided the proposed system into eight subsystems and sequentially devised a geometry calibration method for each subsystem. Accordingly, a calibration phantom consisting of four polytetrafluoroethylene beads, each with 15 mm diameter, was designed. The geometry calibration parameters were estimated by minimizing the difference between the measured bead projection and the forward projection of the formulated subsystem. By reflecting the estimated geometry calibration parameters, the projection data were obtained via rebinning to be used in the filtered-backprojection reconstruction. The proposed calibration and reconstruction methods were validated by computer simulations and real experiments. Additionally, the accuracy of the geometry calibration method was examined by computer simulation. Furthermore, we validated the improved quality of the reconstructed image through the mean-squared error (MSE), structure similarity (SSIM), and visual inspections for both the simulated and experimental data. The results show that the calibrated images, reconstructed by reflecting the calibration results, have smaller MSE and higher SSIM values than the uncalibrated images. The calibrated images were observed to have fewer artifacts than the uncalibrated images in visual inspection, demonstrating that the proposed calibration and reconstruction methods effectively reduce artifacts caused by geometry misalignments.