An adaptive weighted ensemble learning network for diabetic retinopathy classification.

Diabetic retinopathy (DR) is one of the leading causes of blindness. However, because the data distribution of classes is not always balanced, it is challenging for automated early DR detection using deep learning techniques. In this paper, we propose an adaptive weighted ensemble learning method for DR detection based on optical coherence tomography (OCT) images. Specifically, we develop an ensemble learning model based on three advanced deep learning models for higher performance. To better utilize the cues implied in these base models, a novel decision fusion scheme is proposed based on the Bayesian theory in terms of the key evaluation indicators, to dynamically adjust the weighting distribution of base models to alleviate the negative effects potentially caused by the problem of unbalanced data size. Extensive experiments are performed on two public datasets to verify the effectiveness of the proposed method. A quadratic weighted kappa of 0.8487 and an accuracy of 0.9343 on the DRAC2022 dataset, and a quadratic weighted kappa of 0.9007 and an accuracy of 0.8956 on the APTOS2019 dataset are obtained, respectively. The results demonstrate that our method has the ability to enhance the ovearall performance of DR detection on OCT images.

Using HVS Dual-Pathway and Contrast Sensitivity to Blindly Assess Image Quality.

Blind image quality assessment (BIQA) aims to evaluate image quality in a way that closely matches human perception. To achieve this goal, the strengths of deep learning and the characteristics of the human visual system (HVS) can be combined. In this paper, inspired by the ventral pathway and the dorsal pathway of the HVS, a dual-pathway convolutional neural network is proposed for BIQA tasks. The proposed method consists of two pathways: the "what" pathway, which mimics the ventral pathway of the HVS to extract the content features of distorted images, and the "where" pathway, which mimics the dorsal pathway of the HVS to extract the global shape features of distorted images. Then, the features from the two pathways are fused and mapped to an image quality score. Additionally, gradient images weighted by contrast sensitivity are used as the input to the "where" pathway, allowing it to extract global shape features that are more sensitive to human perception. Moreover, a dual-pathway multi-scale feature fusion module is designed to fuse the multi-scale features of the two pathways, enabling the model to capture both global features and local details, thus improving the overall performance of the model. Experiments conducted on six databases show that the proposed method achieves state-of-the-art performance.

HCformer: Hybrid CNN-Transformer for LDCT Image Denoising.

Low-dose computed tomography (LDCT) is an effective way to reduce radiation exposure for patients. However, it will increase the noise of reconstructed CT images and affect the precision of clinical diagnosis. The majority of the current deep learning-based denoising methods are built on convolutional neural networks (CNNs), which concentrate on local information and have little capacity for multiple structures modeling. Transformer structures are capable of computing each pixel's response on a global scale, but their extensive computation requirements prevent them from being widely used in medical image processing. To reduce the impact of LDCT scans on patients, this paper aims to develop an image post-processing method by combining CNN and Transformer structures. This method can obtain a high-quality images from LDCT. A hybrid CNN-Transformer (HCformer) codec network model is proposed for LDCT image denoising. A neighborhood feature enhancement (NEF) module is designed to introduce the local information into the Transformer's operation, and the representation of adjacent pixel information in the LDCT image denoising task is increased. The shifting window method is utilized to lower the computational complexity of the network model and overcome the problems that come with computing the MSA (Multi-head self-attention) process in a fixed window. Meanwhile, W/SW-MSA (Windows/Shifted window Multi-head self-attention) is alternately used in two layers of the Transformer to gain the information interaction between various Transformer layers. This approach can successfully decrease the Transformer's overall computational cost. The AAPM 2016 LDCT grand challenge dataset is employed for ablation and comparison experiments to demonstrate the viability of the proposed LDCT denoising method. Per the experimental findings, HCformer can increase the image quality metrics SSIM, HuRMSE and FSIM from 0.8017, 34.1898, and 0.6885 to 0.8507, 17.7213, and 0.7247, respectively. Additionally, the proposed HCformer algorithm will preserves image details while it reduces noise. In this paper, an HCformer structure is proposed based on deep learning and evaluated by using the AAPM LDCT dataset. Both the qualitative and quantitative comparison results confirm that the proposed HCformer outperforms other methods. The contribution of each component of the HCformer is also confirmed by the ablation experiments. HCformer can combine the advantages of CNN and Transformer, and it has great potential for LDCT image denoising and other tasks.

BiRPN-YOLOvX: A weighted bidirectional recursive feature pyramid algorithm for lung nodule detection.

BACKGROUND: Lung cancer has the second highest cancer mortality rate in the world today. Although lung cancer screening using CT images is a common way for early lung cancer detection, accurately detecting lung nodules remains a challenged issue in clinical practice. OBJECTIVE: This study aims to develop a new weighted bidirectional recursive pyramid algorithm to address the problems of small size of lung nodules, large proportion of background region, and complex lung structures in lung nodule detection of CT images. METHODS: First, the weighted bidirectional recursive feature pyramid network (BiPRN) is proposed, which can increase the ability of network model to extract feature information and achieve multi-scale fusion information. Second, a CBAM_CSPDarknet53 structure is developed to incorporate an attention mechanism as a feature extraction module, which can aggregate both spatial information and channel information of the feature map. Third, the weighted BiRPN and CBAM_CSPDarknet53 are applied to the YOLOvX model for lung nodule detection experiments, named BiRPN-YOLOvX, where YOLOvX represents different versions of YOLO. To verify the effectiveness of our weighted BiRPN and CBAM_ CSPDarknet53 algorithm, they are fused with different models of YOLOv3, YOLOv4 and YOLOv5, and extensive experiments are carried out using the publicly available lung nodule datasets LUNA16 and LIDC-IDRI. The training set of LUNA16 contains 949 images, and the validation and testing sets each contain 118 images. There are 1987, 248 and 248 images in LIDC-IDRI's training, validation and testing sets, respectively. RESULTS: The sensitivity of lung nodule detection using BiRPN-YOLOv5 reaches 98.7% on LUNA16 and 96.2% on LIDC-IDRI, respectively. CONCLUSION: This study demonstrates that the proposed new method has potential to help improve the sensitivity of lung nodule detection in future clinical practice.

Cardiac CT motion artifact grading via semi-automatic labeling and vessel tracking using synthetic image-augmented training data.

Cardiac CT provides critical information for the evaluation of cardiovascular diseases. However, involuntary patient motion and physiological movement of the organs during CT scanning cause motion blur in the reconstructed CT images, degrading both cardiac CT image quality and its diagnostic value. In this paper, we propose and demonstrate an effective and efficient method for CT coronary angiography image quality grading via semi-automatic labeling and vessel tracking. These algorithms produce scores that accord with those of expert readers to within 0.85 points on a 5-point scale. We also train a neural network model to perform fully-automatic motion artifact grading. We demonstrate, using XCAT simulation tools to generate realistic phantom CT data, that supplementing clinical data with synthetic data improves the scoring performance of this network. With respect to ground truth scores assigned by expert operators, the mean square error of grading motion of the right coronary artery is reduced by 36% by synthetic data supplementation. This demonstrates that augmentation of clinical training data with realistically synthesized images can potentially reduce the number of clinical studies needed to train the network.

CT imaging of gold nanoparticles in a human-sized phantom.

INTRODUCTION: Gold nanoparticles (AuNPs) are visualized and quantified in a human-sized phantom with a clinical MDCT scanner. METHODS: Experiments were conducted with AuNPs between 0.00171 and 200 mgAu/mL. CT images were acquired at 80, 100, 120, and 140 kVp in a 33-cm phantom. Image contrast due to AuNPs was experimentally determined from regions of interest (ROIs) and effective linear attenuation coefficients were calculated from CT x-ray spectra with consideration of tissue attenuation. RESULTS: The typical 12-bit dynamic range of CT images was exceeded for AuNPs at 150 mgAu/mL. A threshold concentration of 0.3-1.4 mgAu/mL was determined for human visualization in 1-mm images at a typical diagnostic CTDIvol of 23.6 mGy. Optimal image contrast was also achieved at 120 kVp and verified by calculation. CONCLUSIONS: We have shown that scanners capable of reconstructing images with extended Hounsfield scales are required for distinguishing any contrast differences above 150 mgAu/mL. We have also shown that AuNPs result in optimal image contrast at 120 kVp in a human-sized phantom due to gold's 80.7 keV k-edge and the attenuation of x-rays by tissue. Typical CT contrast agents, like iodine, require the use of lower kVps for optimal visualization, but lower kVps are more difficult to implement in the clinic because of elevated noise levels, elongated scan times, and/or beam-hardening artifacts. This indicates another significant advantage of AuNPs over iodine not yet discussed in the literature.

Tensor Gradient L0-Norm Minimization-Based Low-Dose CT and Its Application to COVID-19.

Methods to recover high-quality computed tomography (CT) images in low-dose cases will be of great benefit. To reach this goal, sparse-data subsampling is one of the common strategies to reduce radiation dose, which is attracting interest among the researchers in the CT community. Since analytic image reconstruction algorithms may lead to severe image artifacts, the iterative algorithms have been developed for reconstructing images from sparsely sampled projection data. In this study, we first develop a tensor gradient L0-norm minimization (TGLM) for low-dose CT imaging. Then, the TGLM model is optimized by using the split-Bregman method. The Coronavirus Disease 2019 (COVID-19) has been sweeping the globe, and CT imaging has been deployed for detection and assessing the severity of the disease. Finally, we first apply our proposed TGLM method for COVID-19 to achieve low-dose scan by incorporating the 3-D spatial information. Two COVID-19 patients (64 years old female and 56 years old man) were scanned by the [Formula: see text]CT 528 system, and the acquired projections were retrieved to validate and evaluate the performance of the TGLM.

Compton-camera-based SPECT for thyroid cancer imaging.

Thyroid cancer is the most common type of endocrine-related cancer and the most common cancer in young women. Currently, single photon emission computed tomography (SPECT) and computed tomography (CT) are used with radioiodine scintigraphy to evaluate patients with thyroid cancer. The gamma camera for SPECT contains a mechanical collimator that greatly compromises dose efficiency and limits diagnostic sensitivity. Fortunately, the Compton camera is emerging as an ideal approach for mapping the distribution of radiopharmaceuticals inside the thyroid. In this preliminary study, based on the state-of-the-art readout chip Timepix3, we investigate the feasibility of using Compton camera for radiotracer SPECT imaging in thyroid cancer. A thyroid phantom is designed to mimic human neck, the mechanism of Compton camera-based event detection is simulated to generate realistic list-mode data, and a weighted back-projection method is developed to reconstruct the original distribution of the emission source. Study results show that the Compton camera can improve the detection efficiency for two or higher orders of magnitude comparing with the conventional gamma cameras. The thyroid gland regions can be reconstructed from the Compton camera measurements in terms of radiotracer distribution. This makes the Compton-camera-based SPECT imaging a promising modality for future clinical applications with significant benefits for dose reduction, scattering artifact reduction, temporal resolution enhancement, scan throughput increment, and others.

Review of CT image reconstruction open source toolkits.

Computed tomography (CT) has been widely applied in medical diagnosis, nondestructive evaluation, homeland security, and other science and engineering applications. Image reconstruction is one of the core CT imaging technologies. In this review paper, we systematically reviewed the currently publicly available CT image reconstruction open source toolkits in the aspects of their environments, object models, imaging geometries, and algorithms. In addition to analytic and iterative algorithms, deep learning reconstruction networks and open codes are also reviewed as the third category of reconstruction algorithms. This systematic summary of the publicly available software platforms will help facilitate CT research and development.

Image gradient L0-norm based PICCS for swinging multi-source CT reconstruction.

Dynamic computed tomography (CT) is usually employed to image motion objects, such as beating heart, coronary artery and cerebral perfusion, etc. Recently, to further improve the temporal resolution for aperiodic industrial process imaging, the swinging multi-source CT (SMCT) systems and the corresponding swinging multi-source prior image constrained compressed sensing (SM-PICCS) method were developed. Since the SM-PICCS uses the L1-norm of image gradient, the edge structures in the reconstructed images are blurred and motion artifacts are still present. Inspired by the advantages in terms of image edge preservation and fine structure recovering, the L0-norm of image gradient is incorporated into the prior image constrained compressed sensing, leading to an L0-PICCS Algorithm 1Table 1The parameters of L0-PICCS (δ1,δ2,λ1*,λ2*) for numerical simulation.Sourceswδ1(10-2)δ2(10-2)λ1*(10-2)λ2*(10-8)Noise-free510522.001.525522.001.55035002.00471014.33332.00500025522.00500050222.005000Noise51062002.505002554502.501.55054502.901.571027.385.91.5810000258.285.91.5850050522.001.5. The experimental results confirm that the L0-PICCS outperforms the SM-PICCS in both visual inspection and quantitative analysis.

Tensor decomposition and non-local means based spectral CT image denoising.

BACKGROUNDAs one type of the state-of-the-art detectors, photon counting detectors are used in spectral computed tomography (CT) to classify the received photons into several energy channels and generate multichannel projections simultaneously. However, FBP reconstructed images contain severe noise due to the low photon counts in each energy channel.OBJECTIVEA spectral CT image denoising method based on tensor-decomposition and non-local means (TDNLM) is proposed.METHODSIn a CT image, it is widely accepted that there exists self-similarity over the spatial domain. In addition, because a multichannel CT image is obtained from the same object at different energies, images among different channels are highly correlated. Motivated by these two characteristics of the spectral CT images, tensor decomposition and non-local means are employed to recover fine structures in spectral CT images. Moreover, images in all energy channels are added together to form a high signal-to-noise ratio image, which is applied to encourage the signal preservation of the TDNLM. The combination of TD, NLM and the guidance of a high-quality image enhances the low-dose spectral CT, and a parameter selection strategy is designed to achieve the optimal image quality.RESULTSThe effectiveness of the developed algorithm is validated on both numerical simulations and realistic preclinical applications. The root mean square error (RMSE) and the structural similarity (SSIM) are used to quantitatively assess the image quality. The proposed method successfully restored high-quality images (average RMSE=0.0217 cm-1 and SSIM=0.987) from noisy spectral CT images (average RMSE=0.225 cm-1 and SSIM=0.633). In addition, RMSE of each decomposed material component is also remarkably reduced. Compared to the state-of-the-art iterative spectral CT reconstruction algorithms, the proposed method achieves comparable performance with dramatically reduced computational cost, resulting in a speedup of >50.CONCLUSIONSThe outstanding denoising performance, the high computational efficiency and the adaptive parameter selection strategy make the proposed method practical for spectral CT applications.

FBP-type CT reconstruction algorithms for triple-source circular trajectory with different scanning radii.

Multi-source computed tomography (CT) imaging has unique technical advantages not only for dynamic objects, but also for large-size objects by designing its imaging scan mode. Using the triple-source fan-beam imaging scan mode under three circular trajectories with two different radii, we in this study developed and analyzed theoretically several exact reconstruction algorithms in terms of full-scan and short-scan for three sets of truncated projection data. This triple-source scan configuration in different radii cases is easier to be simulated by a single-source scan configuration in an industrial CT system. The proposed algorithms are based on the idea of filtering-back-projection (FBP) algorithm, and can reconstruct the large-size objects under the same CT devices. The developed algorithms avoid data rebinning and can provide exact and fast image reconstruction. The results of the numerical simulation based data analysis verified that new algorithms were accurate and effective.

Adaptive Nonlocal Means Method for Denoising Basis Material Images From Dual-Energy Computed Tomography.

We propose an adaptive nonlocal means approach for image-domain material decomposition in low-dose dual-energy micro-computed tomography. The key idea is to create a distribution map for decomposition error and assign a smooth weight for a given pixel. This method is applied to the decomposed images of 3 basis materials: bone, soft tissue, and gold in our applications. We assume that bone and gold cannot coexist in the same pixel and regroup these basis materials into 2 categories. For soft tissue, the proposed algorithm is implemented in a noniterative mode. For bone and gold, an iterative mode is used and followed by a postiteration process. Both our numerical simulation and in vivo preclinical experiment results show that the proposed adaptive nonlocal means outperforms other state-of-the-art denoising algorithms, such as the original nonlocal means and total variation minimization methods.

Spatial-Spectral Cube Matching Frame for Spectral CT Reconstruction.

Spectral computed tomography (CT) reconstructs the same scanned object from projections of multiple narrow energy windows, and it can be used for material identification and decomposition. However, the multi-energy projection dataset has a lower signal-noise-ratio (SNR), resulting in poor reconstructed image quality. To address this thorny problem, we develop a spectral CT reconstruction method, namely spatial-spectral cube matching frame (SSCMF). This method is inspired by the following three facts: i) human body usually consists of two or three basic materials implying that the reconstructed spectral images have a strong sparsity; ii) the same basic material component in a single channel image has similar intensity and structures in local regions. Different material components within the same energy channel share similar structural information; iii) multi-energy projection datasets are collected from the subject by using different narrow energy windows, which means images reconstructed from different energy-channels share similar structures. To explore those information, we first establish a tensor cube matching frame (CMF) for a BM4D denoising procedure. Then, as a new regularizer, the CMF is introduced into a basic spectral CT reconstruction model, generating the SSCMF method. Because the SSCMF model contains an L0-norm minimization of 4D transform coefficients, an effective strategy is employed for optimization. Both numerical simulations and realistic preclinical mouse studies are performed. The results show that the SSCMF method outperforms the state-of-the-art algorithms, including the simultaneous algebraic reconstruction technique, total variation minimization, total variation plus low rank, and tensor dictionary learning.

Geometry and energy constrained projection extension.

BACKGROUND: In clinical computed tomography (CT) applications, when a patient is obese or improperly positioned, the final tomographic scan is often partially truncated. Images directly reconstructed by the conventional reconstruction algorithms suffer from severe cupping and direct current bias artifacts. Moreover, the current methods for projection extension have limitations that preclude incorporation from clinical workflows, such as prohibitive computational time for iterative reconstruction, extra radiation dose, hardware modification, etc.METHOD:In this study, we first established a geometrical constraint and estimated the patient habitus using a modified scout configuration. Then, we established an energy constraint using the integral invariance of fan-beam projections. Two constraints were extracted from the existing CT scan process with minimal modification to the clinical workflows. Finally, we developed a novel dual-constraint based optimization model that can be rapidly solved for projection extrapolation and accurate local reconstruction. RESULTS: Both numerical phantom and realistic patient image simulations were performed, and the results confirmed the effectiveness of our proposed approach. CONCLUSION: We establish a dual-constraint-based optimization model and correspondingly develop an accurate extrapolation method for partially truncated projections. The proposed method can be readily integrated into the clinical workflow and efficiently solved by using a one-dimensional optimization algorithm. Moreover, it is robust for noisy cases with various truncations and can be further accelerated by GPU based parallel computing.

Wavelet-based joint CT-MRI reconstruction.

Since their inceptions, the multimodal imaging techniques have received a great deal of attention for achieving enhanced imaging performance. In this work, a novel joint reconstruction framework using sparse computed tomography (CT) and magnetic resonance imaging (MRI) data is developed and evaluated. CT and MRI images are synchronously acquired and registered from a hybrid CT-MRI platform. Because image data are highly undersampled, analytic methods are unable to generate decent image quality. To overcome this drawback, we resort to the compressed sensing (CS) techniques, which employ sparse priors that result from an application of a wavelet transform. To utilize multimodal information, projection distance is introduced and is tuned to tailor the texture and pattern of the final images. Specifically, CT and MRI images are alternately reconstructed using the updated multimodal results that are calculated at the latest step of the iterative optimization algorithm. The method exploits the structural similarities shared by the CT and MRI images to achieve better reconstruction quality. The good performance of the proposed approach is demonstrated on a pair of undersampled CT and MRI body images. Clinical CT and MRI images are tested with the joint reconstruction, the analytic reconstruction, and the independent reconstruction which does not uses multimodal imaging information. Results show that the proposed method improves about 5dB in signal to noise ratio (SNR) and nearly 10% in structural similarity measure comparing to independent reconstruction. It offers similar quality with fully sampled analytic reconstruction with only 20% sampling rate for CT and 40% for MRI. Structural similarities and correlations residing in images from different modalities are useful to mutually promote the quality of image reconstruction.

Low-dose spectral CT reconstruction using image gradient ℓ 0-norm and tensor dictionary.

Spectral computed tomography (CT) has a great superiority in lesion detection, tissue characterization and material decomposition. To further extend its potential clinical applications, in this work, we propose an improved tensor dictionary learning method for low-dose spectral CT reconstruction with a constraint of image gradient ℓ 0-norm, which is named as ℓ 0TDL. The ℓ 0TDL method inherits the advantages of tensor dictionary learning (TDL) by employing the similarity of spectral CT images. On the other hand, by introducing the ℓ 0-norm constraint in gradient image domain, the proposed method emphasizes the spatial sparsity to overcome the weakness of TDL on preserving edge information. The split-bregman method is employed to solve the proposed method. Both numerical simulations and real mouse studies are perform to evaluate the proposed method. The results show that the proposed ℓ 0TDL method outperforms other competing methods, such as total variation (TV) minimization, TV with low rank (TV+LR), and TDL methods.

Swinging multi-source industrial CT systems for aperiodic dynamic imaging.

The goal of this paper is to develop a new architecture for industrial computed tomography (ICT) aiming at dynamically imaging an aperiodic changing object. We propose a data acquisition approach with multiple x-ray source/detector pairs targeting a continuously changeable object with corresponding timeframes. In this named swinging multi-source CT (SMCT) structure, each source and its associated detector swing forth and back within a certain angle for CT scanning. In the SMCT system design, we utilize a circular journal bearing based setup to replace the normal CT slip ring by weakening the scanning speed requirement. Inspired by the prior image constrained compressed sensing (PICCS) algorithm, we apply a modified PICCS algorithm for the SMCT (SM-PICCS). Our numerical simulation and realistic specimen experiment studies demonstrate the feasibility of the proposed approach.

Interior tomography with curvelet-based regularization.

The interior problem, i.e. reconstruction from local truncated projections in computed tomography (CT), is common in practical applications. However, its solution is non-unique in a general unconstrained setting. To solve the interior problem uniquely and stably, in recent years both the prior knowledge- and compressive sensing (CS)-based methods have been developed. Those theoretically exact solutions for the interior problem are called interior tomography. Along this direction, we propose here a new CS-based method for the interior problem based on the curvelet transform. A curvelet is localized in both radial and angular directions in the frequency domain. A two-dimensional (2D) image can be represented in a curvelet frame. We employ the curvelet transform coefficients to regularize the interior problem and obtain a curvelet frame based regularization method (CFRM) for interior tomography. The curvelet coefficients of the reconstructed image are split into two sets according to their visibility from the interior data, and different regularization parameters are used for these two sets. We also presents the results of numerical experiments, which demonstrate the feasibility of the proposed approach.

Initial analysis of the middle problem in CT image reconstruction.

The interior and exterior problems have been extensively studied in the field of reconstruction of computed tomography (CT) images, which lead to important theoretical and practical results. In this study, we formulate a middle problem of CT image reconstruction, which is more challenging than either the interior or exterior problems. In the middle problem of CT image reconstruction, projection data are measured through and only through the middle dough-like region, so that each projection profile misses data not only internally but also on both sides. For an object with a radially symmetric exterior, we proved that the middle problem could be uniquely solved if the middle ring-shaped zone is piecewise constant or there is a known sub-region inside this middle region. Then, we designed and evaluated a POCS-based algorithm for middle tomography, which is to reconstruct a middle image only from the available data. Finally, the remaining issues are also discussed for further research.