Mean Structural Similarity Research Articles

Efficient and low-cost approximate multipliers for image processing applications

The image processing can be implemented in an error-tolerant mode with acceptable accuracy. In digital image processing, the adders and multipliers are the most important components. In this paper, two approximate multipliers based on an approximate 4:2 compressor and two 4-bit full adders with minimizing computational costs are proposed. The first and second proposed approximate multipliers are designed by using the proposed imprecise 4-bit full adders and 4:2 compressor, respectively. In the structure of the proposed multipliers, we have adders for the addition of two multiple-bit numbers. In the adder structure, the full adders are designed based on two modified half adders and one NAND gate. The proposed approximate 4:2 compressor and two approximate 4-bit full adders are compared from hardware and accuracy point of view such as gate count, area-delay product (ADP), error rate (ER), mean error distance (MED), mean relative error distance (MRED), and normalized error distance (NED). The efficacy of proposed multipliers in image processing applications such as image multiplication is evaluated using MATLAB. The results show the proposed approximate multipliers are comparable in terms of area, delay, and mean structural similarity index metric (MSSIM) parameter with other works.

A graphics-accelerated deep neural network approach for turbomachinery flows based on large eddy simulation

In turbomachinery, strongly unsteady rotor–stator interaction triggers complex three-dimensional turbulent flow phenomena such as flow separation and vortex dynamics. Large eddy simulation (LES) is an advanced numerical method that has recently been used to resolve large-scale turbulent motions and model subgrid-scale turbulence in turbomachinery. To largely reduce the computing cost of LES for turbomachinery flow, a graphics processing unit (GPU)-accelerated deep neural network-based flow field prediction approach is explored, which combines convolutional neural network autoencoder (CNN-AE) with long short-term memory (LSTM). CNN-AE extracts spatial features of turbomachinery flow by mapping high-dimensional flow fields into low-dimensional space, while LSTM is used to predict the temporal evolution of fluid dynamics. Automatic mixed precision (AMP) is employed to achieve rapid neural network training using Nvidia GTX 1080 Ti GPU, which shows a significant speedup compared with that without AMP. We evaluated the proposed CNN-AE-LSTM (CAL) method against gated recurrent units (GRU) and simple recurrent network (SRN) on two types of turbomachinery, i.e., centrifugal and axial flow pumps. The results show that the proposed CAL shows better capability of capturing the vortex structure details of turbomachinery. When predicting the temporal vorticity field, the mean square error of CAL results is 0.105%–0.124% for centrifugal pumps and 0.071%–0.072% for axial flow pumps. Meanwhile, the structural similarity index measure of the CAL results is 92.51%–92.77% for centrifugal pumps and 93.81%–94.61% for axial flow pumps. The proposed CAL is noticeably better than GRU and SRN in terms of both mean square error and structural similarity index measure.

A Second-Order Method for Removing Mixed Noise from Remote Sensing Images.

Remote sensing image denoising is of great significance for the subsequent use and research of images. Gaussian noise and salt-and-pepper noise are prevalent noises in images. Contemporary denoising algorithms often exhibit limitations when addressing such mixed noise scenarios, manifesting in suboptimal denoising outcomes and the potential blurring of image edges subsequent to the denoising process. To address the above problems, a second-order removal method for mixed noise in remote sensing images was proposed. In the first stage of the method, dilated convolution was introduced into the DnCNN (denoising convolutional neural network) network framework to increase the receptive field of the network, so that more feature information could be extracted from remote sensing images. Meanwhile, a DropoutLayer was introduced after the deep convolution layer to build the noise reduction model to prevent the network from overfitting and to simplify the training difficulty, and then the model was used to perform the preliminary noise reduction on the images. To further improve the image quality of the preliminary denoising results, effectively remove the salt-and-pepper noise in the mixed noise, and preserve more image edge details and texture features, the proposed method employed a second stage on the basis of adaptive median filtering. In this second stage, the median value in the original filter window median was replaced by the nearest neighbor pixel weighted median, so that the preliminary noise reduction result was subjected to secondary processing, and the final denoising result of the mixed noise of the remote sensing image was obtained. In order to verify the feasibility and effectiveness of the algorithm, the remote sensing image denoising experiments and denoised image edge detection experiments were carried out in this paper. When the experimental results are analyzed through subjective visual assessment, images denoised using the proposed method exhibit clearer and more natural details, and they effectively retain edge and texture features. In terms of objective evaluation, the performance of different denoising algorithms is compared using metrics such as mean square error (MSE), peak signal-to-noise ratio (PSNR), and mean structural similarity index (MSSIM). The experimental outcomes indicate that the proposed method for denoising mixed noise in remote sensing images outperforms traditional denoising techniques, achieving a clearer image restoration effect.

Design and realization of area-efficient approximate multiplier structures for image processing applications

The multiplier is a vital element in real-time applications. This work intends to build area-efficient multipliers based on the approximation concept. The approximate computing scheme receives considerable attention generally in error-resilient applications like image processing which reduce the design parameters at the expense of slightly sacrificing the precision of the performance. This paper proposes the two approximate multipliers (PAM-1 & PAM-2) using the modified 4:2 compressor, full-adder, and sum generation component. The main essence of the proposed approximate multiplier structures is to divide the partial product generation rows into (n/2) stages. The proposed approximate compressors are employed to compress partial products on the LSB side of the multiplier, while accurate compressors are utilized on the MSB side. By effectively combining approximate and exact compressors, the developed approximate multipliers achieve better results in terms of circuit parameters and acceptable error metrics compared to existing imprecise designs. These improvements are reflected in the implementation tables reported in the paper. Compared with the accurate multiplier, the proposed multiplier (PAM-2) reduce the area requirement, delay and power consumption by 38%, 46%, and 50%, respectively. In addition, while extending to image processing applications like multiplication, smoothing, and sharpening, the proposed approximate multipliers attain a better image quality than existing designs, which measure in terms of the mean structural similarity index metric.

Convergence of compressed sensing encryption and deep network recovery in RoF system

In this paper, a security scheme based on the convergence of compressed sensing (CS) encryption and deep network recovery is proposed in radio-over-fiber (RoF) system. In this scheme, CS encryption and deep network recovery are performed at the transmitter and receiver respectively. At the transmitter, the original data are selected based on the compression ratio (CR) and then protected by a three-stage encryption process after CS compression, and the dynamic key (DK) is generated from the original data. In the encryption process, the compressed information is scrambled by 3D rotation, then singular value decomposition (SVD) and DK embedding is implemented, followed by constellation symbol masking. At the receiver, a deep network is used to recover the original data and improve recovery performance. The proposed scheme is experimentally demonstrated and the encrypted OFDM signals are successfully transmitted over 50 km standard single-mode fiber (SSMF) and 5 m wireless channel. The experimental results show that the scheme has a key space of 1.34×10233, which can resist brute intrusion by eavesdroppers, and the bit error rate (BER) is below 10−3 when the received optical power (ROP) is greater than 4 dBm. Additionally, the mean structural similarity (MSSIM) is higher than 0.97, which provides a gain of 0.2 compared to traditional methods. So, the proposed scheme has good security performance and recovery performance, which has a significant potential for application in future optical networks.

Synthesis of diffusion-weighted MRI scalar maps from FLAIR volumes using generative adversarial networks.

Acquisition and pre-processing pipelines for diffusion-weighted imaging (DWI) volumes are resource- and time-consuming. Generating synthetic DWI scalar maps from commonly acquired brain MRI sequences such as fluid-attenuated inversion recovery (FLAIR) could be useful for supplementing datasets. In this work we design and compare GAN-based image translation models for generating DWI scalar maps from FLAIR MRI for the first time. We evaluate a pix2pix model, two modified CycleGANs using paired and unpaired data, and a convolutional autoencoder in synthesizing DWI fractional anisotropy (FA) and mean diffusivity (MD) from whole FLAIR volumes. In total, 420 FLAIR and DWI volumes (11,957 images) from multi-center dementia and vascular disease cohorts were used for training/testing. Generated images were evaluated using two groups of metrics: (1) human perception metrics including peak signal-to-noise ratio (PSNR) and structural similarity (SSIM), (2) structural metrics including a newly proposed histogram similarity (Hist-KL) metric and mean squared error (MSE). Pix2pix demonstrated the best performance both quantitatively and qualitatively with mean PSNR, SSIM, and MSE metrics of 23.41 dB, 0.8, 0.004, respectively for MD generation, and 24.05 dB, 0.78, 0.004, respectively for FA generation. The new histogram similarity metric demonstrated sensitivity to differences in fine details between generated and real images with mean pix2pix MD and FA Hist-KL metrics of 11.73 and 3.74, respectively. Detailed analysis of clinically relevant regions of white matter (WM) and gray matter (GM) in the pix2pix images also showed strong significant (p < 0.001) correlations between real and synthetic FA values in both tissue types (R = 0.714 for GM, R = 0.877 for WM). Our results show that pix2pix's FA and MD models had significantly better structural similarity of tissue structures and fine details than other models, including WM tracts and CSF spaces, between real and generated images. Regional analysis of synthetic volumes showed that synthetic DWI images can not only be used to supplement clinical datasets, but demonstrates potential utility in bypassing or correcting registration in data pre-processing.

Three-dimensional ultrasound image reconstruction based on 3D-ResNet in the musculoskeletal system using a 1D probe: ex vivo and in vivo feasibility studies

Objective. Three-dimensional (3D) ultrasound (US) is needed to provide sonographers with a more intuitive panoramic view of the complex anatomical structure, especially the musculoskeletal system. In actual scanning, sonographers may perform fast scanning using a one-dimensional (1D) array probe .at random angles to gain rapid feedback, which leads to a large US image interval and missing regions in the reconstructed volume. Approach. In this study, a 3D residual network (3D-ResNet) modified by a 3D global residual branch (3D-GRB) and two 3D local residual branches (3D-LRBs) was proposed to retain detail and reconstruct high-quality 3D US volumes with high efficiency using only sparse two-dimensional (2D) US images. The feasibility and performance of the proposed algorithm were evaluated on ex vivo and in vivo sets. Main r esults. High-quality 3D US volumes in the fingers, radial and ulnar bones, and metacarpophalangeal joints were obtained by the 3D-ResNet, respectively. Their axial, coronal, and sagittal slices exhibited rich texture and speckle details. Compared with kernel regression, voxel nearest-neighborhood, squared distance weighted methods, and a 3D convolution neural network in the ablation study, the mean peak-signal-to-noise ratio and mean structure similarity of the 3D-ResNet were up to 28.53 ± 1.29 dB and 0.98 ± 0.01, respectively, and the corresponding mean absolute error dropped to 0.023 ± 0.003 with a better resolution gain of 1.22 ± 0.19 and shorter reconstruction time. Significance. These results illustrate that the proposed algorithm can rapidly reconstruct high-quality 3D US volumes in the musculoskeletal system in cases of a large amount of data loss. This suggests that the proposed algorithm has the potential to provide rapid feedback and precise analysis of stereoscopic details in complex and meticulous musculoskeletal system scanning with a less limited scanning speed and pose variations for the 1D array probe.

Advanced framework for enhancing ultrasound images through an optimized hybrid search algorithm and a novel motion compounding processing chain

Ultrasound imaging is a fast, widespread, and essential diagnostic technique for examining the body’s internal anatomy to find abnormal tissues or diseases. However, speckle noise in ultrasound imaging corrupts fine details and edges and degrades the image’s resolution and contrast, making diagnosing more difficult. In this work, a speckle reduction method using motion compounding is proposed. The objective of this work is ultrasound image enhancement keeping all the diagnostics details and edges. The proposed method uses the pre-locations frames that the sonographer generates before locating the required diagnostic frame. These frames are applied to the proposed optimized Modified Three-Step Search (O-MTSS) algorithm to enhance the final processed frame. The O-MTSS algorithm is a hybrid between the Three Step Search algorithm (TSS) and the New Diamond Search Algorithm (NDS). The method requires a scoring layer for storing additional frames to select optimal frames that will be used for enhancement. The proposed methodology is tested on synthetic and real ultrasound images. The outcomes produce considerable speckle reduction while maintaining the image edges with good computational time for real-time scanning. The result is evaluated by subjective physicians, radiologists, and image evaluation metrics. According to the percentages of the subjective analysis, there is a remarkable improvement in the processed image. The proposed algorithm result is compared with existing algorithms; It is observed that the improvement in terms of signal to noise ratio, peak signal to noise ratio, mean square error, root mean square error and structural similarity index values of the proposed method are 4.6%, 3.32%, 12.02%, 6.2% and 1.57% respectively over Non-Local Low-Rank (NLLR) method. According to qualitative and quantitative analysis, the suggested method outperforms existing speckle reduction techniques regarding edge and fine detail preservation.

Fast optical coherence tomography angiography image acquisition and reconstruction pipeline for skin application.

Traditional high-quality OCTA images require multi-repeated scans (e.g., 4-8 repeats) in the same position, which may cause the patient to be uncomfortable. We propose a deep-learning-based pipeline that can extract high-quality OCTA images from only two-repeat OCT scans. The performance of the proposed image reconstruction U-Net (IRU-Net) outperforms the state-of-the-art UNet vision transformer and UNet in OCTA image reconstruction from a two-repeat OCT signal. The results demonstrated a mean peak-signal-to-noise ratio increased from 15.7 to 24.2; the mean structural similarity index measure improved from 0.28 to 0.59, while the OCT data acquisition time was reduced from 21 seconds to 3.5 seconds (reduced by 83%).

On the effect of training database size for MR-based synthetic CT generation in the head.

To investigate the performance of unsupervised training using a large number of unpaired data sets as well as the potential gain in performance after fine-tuning with supervised training using spatially registered data sets in generation of synthetic computed tomography (sCT) from magnetic resonance (MR) images. A cycleGAN method consisting of two generators (residual U-Net) and two discriminators (patchGAN) was used for unsupervised training. Unsupervised training utilized unpaired T1-weighted MR and CT images (2061 sets for each modality). Five supervised models were then fine-tuned starting with the generator of the unsupervised model for 1, 10, 25, 50, and 100 pairs of spatially registered MR and CT images. Four supervised training models were also trained from scratch for 10, 25, 50, and 100 pairs of spatially registered MR and CT images using only the residual U-Net generator. All models were evaluated on a holdout test set of spatially registered images from 253 patients, including 30 with significant pathology. sCT images were compared against the acquired CT images using mean absolute error (MAE), Dice coefficient, and structural similarity index (SSIM). sCT images from 60 test subjects generated by the unsupervised, and most accurate of the fine-tuned and supervised models were qualitatively evaluated by a radiologist. While unsupervised training produced realistic-appearing sCT images, addition of even one set of registered images improved quantitative metrics. Addition of more paired data sets to the training further improved image quality, with the best results obtained using the highest number of paired data sets (n=100). Supervised training was found to be superior to unsupervised training, while fine-tuned training showed no clear benefit over supervised learning, regardless of the training sample size. Supervised learning (using either fine tuning or full supervision) leads to significantly higher quantitative accuracy in the generation of sCT from MR images. However, fine-tuned training using both a large number of unpaired image sets was generally no better than supervised learning using registered image sets alone, suggesting the importance of well registered paired data set for training compared to a large set of unpaired data.

Deep-learning based fast and accurate 3D CT deformable image registration in lung cancer.

Deformable Image Registration (DIR) is an essential technique required in many applications of radiation oncology. However, conventional DIR approaches typically take several minutes to register one pair of 3D CT images and the resulting deformable vector fields (DVFs) are only specific to the pair of images used, making it less appealing for clinical application. A deep-learning-based DIR method using CT images is proposed for lung cancer patients to address the common drawbacks of the conventional DIR approaches and in turn can accelerate the speed of related applications, such as contour propagation, dose deformation, adaptive radiotherapy (ART), etc. METHODS: A deep neural network based on VoxelMorph was developed to generate DVFs using CT images collected from 114 lung cancer patients. Two models were trained with the weighted mean absolute error (wMAE) loss and structural similarity index matrix (SSIM) loss (optional) (i.e., the MAE model and the M+S model). In total, 192 pairs of initial CT (iCT) and verification CT (vCT) were included as a training dataset and the other independent 10 pairs of CTs were included as a testing dataset. The vCTs usually were taken 2weeks after the iCTs. The synthetic CTs (sCTs) were generated by warping the vCTs according to the DVFs generated by the pre-trained model. The image quality of the synthetic CTs was evaluated by measuring the similarity between the iCTs and the sCTs generated by the proposed methods and the conventional DIR approaches, respectively. Per-voxel absolute CT-number-difference volume histogram (CDVH) and MAE were used as the evaluation metrics. The time to generate the sCTs was also recorded and compared quantitatively. Contours were propagated using the derived DVFs and evaluated with SSIM. Forward dose calculations were done on the sCTs and the corresponding iCTs. Dose volume histograms (DVHs) were generated based on dose distributions on both iCTs and sCTs generated by two models, respectively. The clinically relevant DVH indices were derived for comparison. The resulted dose distributions were also compared using 3D Gamma analysis with thresholds of 3mm/3%/10% and 2mm/2%/10%, respectively. The two models (wMAE and M+S) achieved a speed of 263.7±163 / 265.8±190ms and a MAE of 13.15±3.8 / 17.52±5.8 HU for the testing dataset, respectively. The average SSIM scores of 0.987±0.006 and 0.988±0.004 were achieved by the two proposed models, respectively. For both models, CDVH of a typical patient showed that less than 5% of the voxels had a per-voxel absolute CT-number-difference larger than 55 HU. The dose distribution calculated based on a typical sCT showed differences of ≤2cGy[RBE] for clinical target volume (CTV) D95 and D5 , within ±0.06% for total lung V5 , ≤1.5cGy[RBE] for heart and esophagus Dmean , and ≤6cGy[RBE] for cord Dmax compared to the dose distribution calculated based on the iCT. The good average 3D Gamma passing rates (>96% for 3mm/3%/10% and>94% for 2mm/2%/10%, respectively) were also observed. A deep neural network-based DIR approach was proposed and has been shown to be reasonably accurate and efficient to register the initial CTs and verification CTs in lung cancer.

Directed searching optimized texture based adaptive gamma correction (DSOTAGC) technique for medical image enhancement.

Because of complexity and low contrast in medical images, few enhancement techniques result unwanted artifacts and information loss by affecting the structure similarity and peak signal to noise ratio. To meet these challenges, a Directed searching optimized texture-based adaptive gamma correction technique is proposed in this article. This proposed technique utilizes the textured regions of the image and suppresses the effect of non-textured regions for eliminating the artifacts. An adaptive clipping threshold is used in the textured image to control the enhancement rate. For improving the contrast, the transfer function of the enhanced image is evaluated using the modified weighted probability density function and adaptive gamma parameter. To make the algorithm more adaptive, parameters like clipped threshold, gamma parameter, and textural threshold are to be optimized using directed searching optimization algorithm. For improving the information contents and noise suppression capability, the proposed technique incorporated a fitness function which is a combination of entropy and peak signal to noise ratio. Equal weightage has been given to each parameter in the fitness function for obtaining a balanced optimal result. Then, the performance of the proposed technique is evaluated in terms of visual quality, information contents, average mean brightness error, noise suppression, and structural similarity. Experimental results show the proposed technique results in better visual effects without information loss. It effectively suppresses the effect of artifacts and significantly improves the contrast by making edges clearer and textures richer over other algorithms.

Generation of fruit’s spectra with hundreds of wavelengths from obtained multi-spectra and spectral application using deep learning

Although multi-spectral technology has been widely used to develop low-cost and low-power fruit internal quality detector, the detection precision was poor because of few wavelengths used. To improve the detection precision, we proposed a novel generation method on fruit’s false spectra with hundreds of wavelengths (named false full-spectra) from obtained multi-spectra with 12 wavelengths for kiwifruit and 11 wavelengths for pear using deep convolutional generative adversarial network (DCGAN), and the method was evaluated by traditional partial least squares regression (PLSR) and one dimensional convolutional neural network models (1D-CNN). The results showed that the generated false full-spectra obtained by DCGAN had high similarity with the ‘real’ full-spectra no matter for kiwifruit and pear. The mean structural similarity between the false full-spectra and the ‘real’ full-spectra of prediction set of kiwifruit and pear was 0.9490 and 0.9877, respectively, and their mean peak signal-to-noise ratio was 35.82 dB and 44.59 dB, respectively. The prediction performances of the built PLSR and 1D-CNN models based on false full-spectra for internal qualities (soluble solids content and firmness) of kiwifruit and pear were much better than those based on multi-spectra and were very close to those based on ‘real’ full-spectra. This study provides a novel idea to improve the detection performance of the multi-spectra-based detector without increasing the hardware cost.

AFSFusion: An Adjacent Feature Shuffle Combination Network for Infrared and Visible Image Fusion

To obtain fused images with excellent contrast, distinct target edges, and well-preserved details, we propose an adaptive image fusion network called the adjacent feature shuffle-fusion network (AFSFusion). The proposed network adopts a UNet-like architecture and incorporates key refinements to enhance network architecture and loss functions. Regarding the network architecture, the proposed two-branch adjacent feature fusion module, called AFSF, expands the number of channels to fuse the feature channels of several adjacent convolutional layers in the first half of the AFSFusion, enhancing its ability to extract, transmit, and modulate feature information. We replace the original rectified linear unit (ReLU) with leaky ReLU to alleviate the problem of gradient disappearance and add a channel shuffling operation at the end of AFSF to facilitate information interaction capability between features. Concerning loss functions, we propose an adaptive weight adjustment (AWA) strategy to assign weight values to the corresponding pixels of the infrared (IR) and visible images in the fused images, according to the VGG16 gradient feature response of the IR and visible images. This strategy efficiently handles different scene contents. After normalization, the weight values are used as weighting coefficients for the two sets of images. The weighting coefficients are applied to three loss items simultaneously: mean square error (MSE), structural similarity (SSIM), and total variation (TV), resulting in clearer objects and richer texture detail in the fused images. We conducted a series of experiments on several benchmark databases, and the results demonstrate the effectiveness of the proposed network architecture and the superiority of the proposed network compared to other state-of-the-art fusion methods. It ranks first in several objective metrics, showing the best performance and exhibiting sharper and richer edges of specific targets, which is more in line with human visual perception. The remarkable enhancement in performance is ascribed to the proposed AFSF module and AWA strategy, enabling balanced feature extraction, fusion, and modulation of image features throughout the process.

QUANTITATIVE BIOMARKERS REPRODUCIBILITY USING GENERATIVE ADVERSARIAL APPROACHES IN REDUCED TO CONVENTIONAL DOSE CT.

In recent years, several techniques for image-to-image translation by means of generative adversarial neural networks (GAN) have been proposed to learn mapping characteristics between a source and a target domain. In particular, in the medical imaging field conditional GAN frameworks with paired samples (cGAN) and unconditional cycle-consistent GANs with unpaired data (CycleGAN) have been demonstrated as a powerful scheme to model non-linear mappings that produce realistic target images from different modality sources. When proposing the usage and adaptation of these frameworks for medical image synthesis, quantitative and qualitative validation are usually performed by assessing the similarity between synthetic and target images in terms of metrics such as mean absolute error (MAE) or structural similarity (SSIM) index. However, an evaluation of clinically relevant markers showing that diagnostic information is not overlooked in the translation process is often missing. In this work, we aim at demonstrating the importance of validating medical image-to-image translation techniques by assessing their effect on the measurement of clinically relevant metrics and biomarkers. We implemented both a conditional and an unconditional approach to synthesize conventional dose chest CT scans from reduced dose CT and show that while both visually and in terms of traditional metrics the network appears to successfully minimize perceptual discrepancies, these methods are not reliable to systematically reproduce quantitative measurements of various chest biomarkers.

High-turbidity underwater active single-pixel imaging based on generative adversarial networks with double Attention U-Net under low sampling rate

Single-pixel imaging (SPI) is widely used in underwater optical imaging because it has the feature of turbulence-free, little backscattering, and simple system structure. However, the image reconstruction quality of SPI is severely degraded as the increasing intensity of scattering medium in underwater environments. In this paper, an underwater active SPI system based on generative adversarial network with double Attention U-Net is proposed to reconstruct the underwater target image with high turbidity under low sampling rates. The generator of the proposed network use two modified Attention U-Net with double skip connections between different layers to improve the fidelity of reconstructed images. The atrous spatial pyramid pooling, and spatial squeeze and channel excitation modules are also integrated to obtain multi-scale information and remove redundant information. What is more, the least squares loss, content loss and mean structural similarity loss are also incorporated into the total loss function to stabilize the training process and avoid the gradient disappearance. Numerical simulations and physical experiments have shown that the proposed method can reconstruct underwater target images at a low sampling rate of 3.52% under the turbidity of 128NTU. The proposed method shows relatively strong reconstruction ability and generalization, and the PSNR and SSIM can be respectively improved by 0.75 times and 3 times compared with compressed sensing SPI. Our work provides a new insight into underwater SPI with high turbidity and can greatly improve the capability of underwater SPI.

Direct estimation of regional lung volume change from paired and single CT images using residual regression neural network.

Chest computed tomography (CT) enables characterization of pulmonary diseases by producing high-resolution and high-contrast images of the intricate lung structures. Deformable image registration is used to align chest CT scans at different lung volumes, yielding estimates of local tissue expansion and contraction. We investigated the utility of deep generative models for directly predicting local tissue volume change from lung CT images, bypassing computationally expensive iterative image registration and providing a method that can be utilized in scenarios where either one or two CT scans areavailable. A residual regression convolutional neural network, called Reg3DNet+, is proposed for directly regressing high-resolution images of local tissue volume change (i.e., Jacobian) from CT images. Image registration was performed between lung volumes at total lung capacity (TLC) and functional residual capacity (FRC) using a tissue mass- and structure-preserving registration algorithm. The Jacobian image was calculated from the registration-derived displacement field and used as the ground truth for local tissue volume change. Four separate Reg3DNet+ models were trained to predict Jacobian images using a multifactorial study design to compare the effects of network input (i.e., single image vs. paired images) and output space (i.e., FRC vs. TLC). The models were trained and evaluated on image datasets from the COPDGene study. Models were evaluated against the registration-derived Jacobian images using local, regional, and global evaluationmetrics. Statistical analysis revealed that both factors - network input and output space - were significant determinants for change in evaluation metrics. Paired-input models performed better than single-input models, and model performance was better in the output space of FRC rather than TLC. Mean structural similarity index for paired-input models was 0.959 and 0.956 for FRC and TLC output spaces, respectively, and for single-input models was 0.951 and 0.937. Global evaluation metrics demonstrated correlation between registration-derived Jacobian mean and predicted Jacobian mean: coefficient of determination (r2 ) for paired-input models was 0.974 and 0.938 for FRC and TLC output spaces, respectively, and for single-input models was 0.598 and 0.346. After correcting for effort, registration-derived lobar volume change was strongly correlated with the predicted lobar volume change: for paired-input models r2 was 0.899 for both FRC and TLC output spaces, and for single-input models r2 was 0.803 and 0.862, respectively. Convolutional neural networks can be used to directly predict local tissue mechanics, eliminating the need for computationally expensive image registration. Networks that use paired CT images acquired at TLC and FRC allow for more accurate prediction of local tissue expansion compared to networks that use a single image. Networks that only require a single input image still show promising results, particularly after correcting for effort, and allow for local tissue expansion estimation in cases where multiple CT scans are not available. For single-input networks, the FRC image is more predictive of local tissue volume change compared to the TLCimage.

Fiber Bundle Image Reconstruction Using Convolutional Neural Networks and Bundle Rotation in Endomicroscopy.

Fiber-bundle endomicroscopy has several recognized drawbacks, the most prominent being the honeycomb effect. We developed a multi-frame super-resolution algorithm exploiting bundle rotation to extract features and reconstruct underlying tissue. Simulated data was used with rotated fiber-bundle masks to create multi-frame stacks to train the model. Super-resolved images are numerically analyzed, which demonstrates that the algorithm can restore images with high quality. The mean structural similarity index measurement (SSIM) improved by a factor of 1.97 compared with linear interpolation. The model was trained using images taken from a single prostate slide, 1343 images were used for training, 336 for validation, and 420 for testing. The model had no prior information about the test images, adding to the robustness of the system. Image reconstruction was completed in 0.03 s for 256 × 256 images indicating future real-time performance is within reach. The combination of fiber bundle rotation and multi-frame image enhancement through machine learning has not been utilized before in an experimental setting but could provide a much-needed improvement to image resolution in practice.

Adaptive tetrahedral interpolation for reconstruction of uneven freehand 3D ultrasound

Objective. Freehand 3D ultrasound volume reconstruction has received considerable attention in medical research because it can freely perform spatial imaging at a low cost. However, the uneven distribution of the original ultrasound images in space reduces the reconstruction effect of the traditional method. Approach. An adaptive tetrahedral interpolation algorithm is proposed to reconstruct 3D ultrasound volume data. The algorithm adaptively divides the unevenly distributed images into numerous tetrahedrons and interpolates the voxel value in each tetrahedron to reconstruct 3D ultrasound volume data. Main results. Extensive experiments on simulated and clinical data confirm that the proposed method can achieve more accurate reconstruction than six benchmark methods. Specifically, the averaged interpolation error at the gray level can be reduced by 0.22–0.82, and the peak signal-to-noise ratio and the mean structure similarity can be improved by 0.32–1.83 dB and 0.01–0.05, respectively. Significance. With the parallel implementation of the algorithm, one 3D ultrasound volume data with size 279 × 279 × 276 can be reconstructed from 100 slices 2D ultrasound images with size 200 × 200 at 1.04 s. Such a quick and accurate approach has practical value in medical research.

An optimized EBRSA-Bi LSTM model for highly undersampled rapid CT image reconstruction

COVID-19 has spread all over the world, causing serious panic around the globe. Chest computed tomography (CT) images are integral in confirming COVID positive patients. Several investigations were conducted to improve or maintain the image reconstruction quality for the sample image reconstruction. Deep learning (DL) methods have recently been proposed to achieve fast reconstruction, but many have focused on a single domain, such as the image domain of k-space. In this research, the highly under-sampled enhanced battle royale self-attention based bi-directional long short-term (EBRSA-bi LSTM) CT image reconstruction model is proposed to reconstruct the image from the under-sampled data. The research is adapted with two phases, namely, pre-processing and reconstruction. The extended cascaded filter (ECF) is proposed for image pre-processing and tends to suppress the noise and enhance the reconstruction accuracy. In the reconstruction model, the battle royale optimization (BrO) is intended to diminish the loss function of the reconstruction network model and weight updation. The proposed model is tested with two datasets, COVID-CT- and SARS-CoV-2 CT. The reconstruction accuracy of the proposed model with two datasets is 93.5 % and 97.7 %, respectively. Also, the image quality assessment parameters such as Peak-Signal to Noise Ratio (PSNR), Root Mean Square Error (RMSE) and Structural Similarity Index metric (SSIM) are evaluated, and it yields an outcome of (45 and 46 dB), (0.0026 and 0.0022) and (0.992, 0.996) with two datasets.