Imaging Scenarios Research Articles

Spatial Domain Indirect Holography for a Full Phaseless Approach to Biomedical Microwave Imaging

Microwave imaging approaches generally require the knowledge of the complex incident field distribution inside the imaging domain. This task can be accomplished by probing the observation domain, thus resulting in a time-consuming process, or assuming a numerical/simulated model. Alternatively, the incident field can be collected on the measurement domain, and the field distribution can be successfully retrieved on the whole imaging space. In this paper, a full phaseless, two-step strategy is described, where the phase of the incident field is retrieved by exploiting a spatial domain indirect holography technique, applied to the acquisition points only. The retrieved complex incident field can be so used for the reconstruction of the incident field inside the domain of interest. More specifically, a modal expansion technique and a source reconstruction method are adopted and compared to perform the above task. Once performing the characterization of the incident field distribution, the proposed strategy is included in the framework of a phaseless inverse scattering problem, where intensity-only data of the total field are involved. The effectiveness of the combined approach is demonstrated by considering an imaging scenario which involves a series of breast models as inhomogeneous targets.

Underwater Acoustic Imaging via Online Bayesian Compressive Beamforming

Beamforming is extensively used in underwater acoustic imaging systems. As a high-resolution variety, a Bayesian compressive beamformer treats the acoustic echo of each snapshot independently, and achieves enhanced recovery performance. However, its associated computational cost is extremely high compared with conventional beamformers, and its iterative implementation makes it difficult to be used online. To overcome these obstacles, an online Bayesian compressive beamformer based on Kalman filtering (online-KSBL) is proposed in this work, which is non-iterative and computationally efficient for a continual working scenario, where a long-term imaging task is conducted underwater. Imposing independent assumptions on snapshots, the online-SBL approach can be derived from online-KSBL. The hyperparameters of the model are estimated recursively, following efficient procedures that are implemented approximately with a sawtooth lag scheme. For the underwater imaging scenario, the correlation among snapshots is exploited and inferred from online-KSBL, and it is found that online-SBL performs competitively with online-KSBL, and it is sufficient to employ online-SBL without considering the snapshot correlation, retaining low complexity, as demonstrated by experimental results.

New transform to project axisymmetric deflection fields along arbitrary rays

Axisymmetric tomography is used to extract quantitative information from line-of-sight measurements of gas flow and combustion fields. For instance, background-oriented schlieren (BOS) measurements are typically inverted by tomographic reconstruction to estimate the density field of a high-speed or high-temperature flow. Conventional reconstruction algorithms are based on the inverse Abel transform, which assumes that rays are parallel throughout the target object. However, camera rays are not parallel, and this discrepancy can result in significant errors in many practical imaging scenarios. We present a generalization of the Abel transform for use in tomographic reconstruction of light-ray deflections through an axisymmetric target. The new transform models the exact path of camera rays instead of assuming parallel paths, thereby improving the accuracy of estimates. We demonstrate our approach with a simulated BOS scenario in which we reconstruct noisy synthetic deflection data across a range of camera positions. Results are compared to state-of-the-art Abel-based algorithms. Reconstructions computed using the new transform are consistently more stable and accurate than conventional reconstructions.

Hybrid Segmentation Method for Boundary Delineation of Agricultural Fields in Multitemporal Satellite Image using HS-PSO-FCNN

Demonstrating the ability to map and track the geographic distribution of agricultural resources is crucial for enhancing agricultural production and guaranteeing food security in numerous parts of the world. A satellite-based technique can significantly cut expenses and boost productivity when compared to traditional field surveys, allowing for the comprehensive mapping of agricultural resources across broad geographic areas. This work focuses on delineating agricultural fields from multispectral satellite images, a field is a piece of land used for agriculture that contains specific land, water, and other features. In this paper, morphological leveling and morphological continuity are utilized to formalize a morphological characteristic. In Multiscale Segmentation (MS), this attribute is defined by taking the derivative of the morphological profile. MS is very beneficial for complicated image scenarios such as aerial or fine-resolution satellite imagery. Based on Hybrid Segmentation of Particle-Swarm-Optimization with a Fully Convolution Network (HS-PSO-FCN), the efficiency of the work has been increased. PSNR processing time has been taken as output metrics used to assess the quality.

A review of explainable and interpretable AI with applications in COVID-19 imaging.

The development of medical imaging artificial intelligence (AI) systems for evaluating COVID‐19 patients has demonstrated potential for improving clinical decision making and assessing patient outcomes during the recent COVID‐19 pandemic. These have been applied to many medical imaging tasks, including disease diagnosis and patient prognosis, as well as augmented other clinical measurements to better inform treatment decisions. Because these systems are used in life‐or‐death decisions, clinical implementation relies on user trust in the AI output. This has caused many developers to utilize explainability techniques in an attempt to help a user understand when an AI algorithm is likely to succeed as well as which cases may be problematic for automatic assessment, thus increasing the potential for rapid clinical translation. AI application to COVID‐19 has been marred with controversy recently. This review discusses several aspects of explainable and interpretable AI as it pertains to the evaluation of COVID‐19 disease and it can restore trust in AI application to this disease. This includes the identification of common tasks that are relevant to explainable medical imaging AI, an overview of several modern approaches for producing explainable output as appropriate for a given imaging scenario, a discussion of how to evaluate explainable AI, and recommendations for best practices in explainable/interpretable AI implementation. This review will allow developers of AI systems for COVID‐19 to quickly understand the basics of several explainable AI techniques and assist in the selection of an approach that is both appropriate and effective for a given scenario.

Differentiable time-gated rendering

The continued advancements of time-of-flight imaging devices have enabled new imaging pipelines with numerous applications. Consequently, several forward rendering techniques capable of accurately and efficiently simulating these devices have been introduced. However, general-purpose differentiable rendering techniques that estimate derivatives of time-of-flight images are still lacking. In this paper, we introduce a new theory of differentiable time-gated rendering that enjoys the generality of differentiating with respect to arbitrary scene parameters. Our theory also allows the design of advanced Monte Carlo estimators capable of handling cameras with near-delta or discontinuous time gates. We validate our theory by comparing derivatives generated with our technique and finite differences. Further, we demonstrate the usefulness of our technique using a few proof-of-concept inverse-rendering examples that simulate several time-of-flight imaging scenarios.

Computer-Assisted Reporting and Decision Support in Standardized Radiology Reporting for Cancer Imaging.

Recent advances in structured reporting are providing an opportunity to enhance cancer imaging assessment to drive value-based care and improve patient safety. The computer-assisted reporting and decision support (CAR/DS) framework has been developed to enable systematic ingestion of guidelines as clinical decision structured reporting tools embedded within the radiologist's workflow. CAR/DS tools can reduce the radiology reporting variability and increase compliance with clinical guidelines. The lung cancer use-case is used to describe various scenarios of a cancer imaging structured reporting pathway, including incidental findings, screening, staging, and restaging or continued care. Various aspects of these tools are also described using cancer-related examples for different imaging modalities and applications such as calculators. Such systems can leverage artificial intelligence (AI) algorithms to assist with the generation of structured reports and there are opportunities for new AI applications to be created using the structured data associated with CAR/DS tools. These AI-enabled systems are starting to allow information from multiple sources to be integrated and inserted into structured reports to drive improvements in clinical decision support and patient care.

MLMT-CNN for object detection and segmentation in multi-layer and multi-spectral images

Precisely localising solar Active Regions (AR) from multi-spectral images is a challenging but important task in understanding solar activity and its influence on space weather. A main challenge comes from each modality capturing a different location of the 3D objects, as opposed to typical multi-spectral imaging scenarios where all image bands observe the same scene. Thus, we refer to this special multi-spectral scenario as multi-layer. We present a multi-task deep learning framework that exploits the dependencies between image bands to produce 3D AR localisation (segmentation and detection) where different image bands (and physical locations) have their own set of results. Furthermore, to address the difficulty of producing dense AR annotations for training supervised machine learning (ML) algorithms, we adapt a training strategy based on weak labels (i.e. bounding boxes) in a recursive manner. We compare our detection and segmentation stages against baseline approaches for solar image analysis (multi-channel coronal hole detection, SPOCA for ARs) and state-of-the-art deep learning methods (Faster RCNN, U-Net). Additionally, both detection and segmentation stages are quantitatively validated on artificially created data of similar spatial configurations made from annotated multi-modal magnetic resonance images. Our framework achieves an average of 0.72 IoU (segmentation) and 0.90 F1 score (detection) across all modalities, comparing to the best performing baseline methods with scores of 0.53 and 0.58, respectively, on the artificial dataset, and 0.84 F1 score in the AR detection task comparing to baseline of 0.82 F1 score. Our segmentation results are qualitatively validated by an expert on real ARs.

Model reduction in acoustic inversion by artificial neural network.

In ultrasound tomography, the speed of sound inside an object is estimated based on acoustic measurements carried out by sensors surrounding the object. An accurate forward model is a prominent factor for high-quality image reconstruction, but it can make computations far too time-consuming in many applications. Using approximate forward models, it is possible to speed up the computations, but the quality of the reconstruction may have to be compromised. In this paper, a neural network-based approach is proposed that can compensate for modelling errors caused by the approximate forward models. The approach is tested with various different imaging scenarios in a simulated two-dimensional domain. The results show that with fairly small training datasets, the proposed approach can be utilized to approximate the modelling errors, and to significantly improve the image reconstruction quality in ultrasound tomography, compared to commonly used inversion algorithms.

A Cascaded Model Based on EfficientDet and YOLACT++ for Instance Segmentation of Cow Collar ID Tag in an Image.

In recent years, many imaging systems have been developed to monitor the physiological and behavioral status of dairy cows. However, most of these systems do not have the ability to identify individual cows because the systems need to cooperate with radio frequency identification (RFID) to collect information about individual animals. The distance at which RFID can identify a target is limited, and matching the identified targets in a scenario of multitarget images is difficult. To solve the above problems, we constructed a cascaded method based on cascaded deep learning models, to detect and segment a cow collar ID tag in an image. First, EfficientDet-D4 was used to detect the ID tag area of the image, and then, YOLACT++ was used to segment the area of the tag to realize the accurate segmentation of the ID tag when the collar area accounts for a small proportion of the image. In total, 938 and 406 images of cows with collar ID tags, which were collected at Coldstream Research Dairy Farm, University of Kentucky, USA, in August 2016, were used to train and test the two models, respectively. The results showed that the average precision of the EfficientDet-D4 model reached 96.5% when the intersection over union (IoU) was set to 0.5, and the average precision of the YOLACT++ model reached 100% when the IoU was set to 0.75. The overall accuracy of the cascaded model was 96.5%, and the processing time of a single frame image was 1.92 s. The performance of the cascaded model proposed in this paper is better than that of the common instance segmentation models, and it is robust to changes in brightness, deformation, and interference around the tag.

Deepening into the suitability of using pre-trained models of ImageNet against a lightweight convolutional neural network in medical imaging: an experimental study.

Transfer learning (TL) has been widely utilized to address the lack of training data for deep learning models. Specifically, one of the most popular uses of TL has been for the pre-trained models of the ImageNet dataset. Nevertheless, although these pre-trained models have shown an effective performance in several domains of application, those models may not offer significant benefits in all instances when dealing with medical imaging scenarios. Such models were designed to classify a thousand classes of natural images. There are fundamental differences between these models and those dealing with medical imaging tasks regarding learned features. Most medical imaging applications range from two to ten different classes, where we suspect that it would not be necessary to employ deeper learning models. This paper investigates such a hypothesis and develops an experimental study to examine the corresponding conclusions about this issue. The lightweight convolutional neural network (CNN) model and the pre-trained models have been evaluated using three different medical imaging datasets. We have trained the lightweight CNN model and the pre-trained models with two scenarios which are with a small number of images once and a large number of images once again. Surprisingly, it has been found that the lightweight model trained from scratch achieved a more competitive performance when compared to the pre-trained model. More importantly, the lightweight CNN model can be successfully trained and tested using basic computational tools and provide high-quality results, specifically when using medical imaging datasets.

Highly coherent illumination for imaging through opacity

Imaging through opacity has captured extensive attention in recent years for its promising applications in bioimaging systems and information extraction from scrambled light. Many optimized image reconstruction algorithms and advanced imaging scenarios are introduced and intensively studied in previous researches. However, intensity-based image reconstruction under the highly coherent light illumination is still a major challenge due to the strong interference effect. Here, we demonstrate that imaging through opacity can be implemented efficiently even under the highly coherent illumination with the aid of an intensity-based/incoherent deconvolution algorithm. Based on the experimental and theoretical analysis, we find the small phase correlation length of the diffuser plays a pivotal role in enabling the incoherent deconvolution operation for highly coherent illumination. Furthermore, the deconvolution algorithm shows insensitivity to the light coherence since the spatial frequencies of the object are well preserved in the spatial frequency spectrum of the speckle pattern. Our results provide a phase-information-free scheme for imaging through opacity under the highly coherent illumination. We anticipate our work would motivate the deep understanding of imaging using highly coherent illumination, increase the structural robustness of imaging systems and facilitate wavelength-sensitive imaging applications.

Planar Near-Field Electric Field Sensor Array Applications Facilitated by Neural Networks

Electrical capacitance tomography is employed for various process tomography applications, typically with circular imaging regions (e.g., to estimate fluid levels in plastic pipes). Typical state-of-the-art implementations focus on circular or cylindrical sensor arrays. In contrast, this research explores using a planar 2D array of electric-field sensors to image volumes composed of various dielectric materials. The array is designed to be used with very-low-frequency electric fields, which are desirable due to their ability to differentiate between various non-conducting objects. D-dot sensors (i.e., charge induction sensors) are used as the electric-field sensing element. In this research, imaging regions of interest are modeled as a composition of (25 cm) <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3</sup> voxels of dielectric material with randomized relative permittivities. Neural networks are utilized as the inversion algorithm to map measured E-field distortions to the voxels’ relative permittivities. Three applications are explored in a simulated environment: 1) predicting relative permittivities of the entire (pseudo-3D) imaging region from one measurement of electric-field distortions (modeled in free space), 2) imaging regions arbitrarily large (in two dimensions) using the planar array as an imaging kernel, and 3) repeating application (1) in a model of a practical, real-world imaging scenario both with and without interfering material. Application (3) is performed with a real-world experimental setup using a room-sized “E-field Cage” meant to generate a uniform electric field. This work showcases a new electric-field imaging modality using a planar 2D D-dot sensor array paired with a DNN-based inversion algorithm.

Analysis of absorbed dose in radioimmunotherapy with 177Lu-trastuzumab using two different imaging scenarios: a pilot study.

Internal organ dosimetry is an important procedure to demonstrate the reliable application of 177Lu-trastuzumab radioimmunotherapy for human epidermal growth factor receptor-positive metastatic breast cancers. We are reporting the first human dosimetry study for 177Lu-trastuzumab. Another objective of our study was to calculate and compare the absorbed doses for normal organs and tumor lesions in patients before radioimmunotherapy with 177Lu-trastuzumab using two different imaging scenarios. Eleven patients (48.27 ± 8.95 years) with a history of metastatic breast cancer were included in the study. Postadministration of 177Lu-trastuzumab (351.09 ± 23.89 MBq/2 mg), acquisition was performed using planar and hybrid imaging scenarios at 4, 24, 72 and 168 h. Single-photon emission computed tomography/computed tomography imaging was performed at 72 h postinjection. Acquired images were processed using Dosimetry Toolkit software for the estimation of normalized cumulated activity in organs and tumor lesions. OLINDA/EXM 2.0 software was used for absorbed dose calculation in both scenarios. Significant difference in normalized cumulated activity and the absorbed dose is noted between two imaging scenarios for the organs and tumor lesions (P < 0.05). Mean absorbed dose (mGy/MBq) estimated from heart, lungs, liver, spleen, kidney, adrenal, pancreas and colon using planar and hybrid scenarios were 0.81 ± 0.19 and 0.63 ± 0.17; 0.75 ± 0.13 and 0.32 ± 0.06; 1.26 ± 0.25 and 1.01 ± 0.17; 0.68 ± 0.22 and 0.53 ± 0.16; 0.91 ± 0.3 and 0.69 ± 0.24; 0.18 ± 0.04 and 0.11 ± 0.02; 0.25 ± 0.22 and 0.09 ± 0.02 and 0.75 ± 0.61 and 0.44 ± 0.28, respectively. On the basis of our dosimetric evaluation, we concluded that radioimmunotherapy with 177Lu-trastuzumab is well tolerated to be implemented in routine clinical practice against HER2 positive metastatic breast cancer. Liver is the main critical organ at risk. Hybrid scenario demonstrated significantly lower absorbed doses in organs and tumors compared to the multiplanar method.

Deconvolution-based partial volume correction of PET images with parallel level set regularization

The partial volume effect (PVE), caused by the limited spatial resolution of positron emission tomography (PET), degrades images both qualitatively and quantitatively. Anatomical information provided by magnetic resonance (MR) images has the potential to play an important role in partial volume correction (PVC) methods. Post-reconstruction MR-guided PVC methods typically use segmented MR tissue maps, and further, assume that PET activity distribution is uniform in each region, imposing considerable constraints through anatomical guidance. In this work, we present a post-reconstruction PVC method based on deconvolution with parallel level set (PLS) regularization. We frame the problem as an iterative deconvolution task with PLS regularization that incorporates anatomical information without requiring MR segmentation or assuming uniformity of PET distributions within regions. An efficient algorithm for non-smooth optimization of the objective function (invoking split Bregman framework) is developed so that the proposed method can be feasibly applied to 3D images and produces sharper images compared to PLS method with smooth optimization. The proposed method was evaluated together with several other PVC methods using both realistic simulation experiments based on the BrainWeb phantom as well as in vivo human data. Our proposed method showed enhanced quantitative performance when realistic MR guidance was provided. Further, the proposed method is able to reduce image noise while preserving structure details on in vivo human data, and shows the potential to better differentiate amyloid positive and amyloid negative scans. Overall, our results demonstrate promise to provide superior performance in clinical imaging scenarios.

Memory effect assisted imaging through multimode optical fibres

When light propagates through opaque material, the spatial information it holds becomes scrambled, but not necessarily lost. Two classes of techniques have emerged to recover this information: methods relying on optical memory effects, and transmission matrix (TM) approaches. Here we develop a general framework describing the nature of memory effects in structures of arbitrary geometry. We show how this framework, when combined with wavefront shaping driven by feedback from a guide-star, enables estimation of the TM of any such system. This highlights that guide-star assisted imaging is possible regardless of the type of memory effect a scatterer exhibits. We apply this concept to multimode fibres (MMFs) and identify a ‘quasi-radial’ memory effect. This allows the TM of an MMF to be approximated from only one end - an important step for micro-endoscopy. Our work broadens the applications of memory effects to a range of novel imaging and optical communication scenarios.

Gaussian Mixture Model Parameter Optimization in Range Points Migration-Based Three-Dimensional Radar Imaging

Millimeter-wave (MMW) imaging radar is one of the major environmental sensors in recent self-driving or driver assistance systems and provides environmentally robust sensing even in optically challenging conditions. As a promising three-dimensional (3-D) radar imaging technique, this study focuses on the range points migration (RPM) method, which has advantages in terms of accuracy and low complexity. In the original RPM algorithm, the parameters are manually or empirically determined by considering the sensor configuration and target shape or distance. To address this limitation, we introduced a novel parameter optimization scheme based on the Gaussian mixture model (GMM) and the expectation maximization (EM) algorithm. In addition, we used the k-space decomposition-based parameter determination scheme to determine target shape-dependent parameter selection. The results, assuming the human body imaging scenario, showed that our proposed method retains a highly accurate target image without requiring empirical parameter selection.

Simulation of anisoplanatic lucky look imaging and statistics through optical turbulence using numerical wave propagation.

This paper investigates anisoplanatic numerical wave simulation in the context of lucky look imaging. We demonstrate that numerical wave propagation can produce root mean square (RMS) wavefront distributions and probability of lucky look (PLL) statistics that are consistent with Kolmogorov theory. However, the simulated RMS statistics are sensitive to the sampling parameters used in the propagation window. To address this, we propose and validate a new sample spacing rule based on the point source bandwidth used in the propagation and the level of atmospheric turbulence. We use the tuned simulator to parameterize the wavefront RMS probability density function as a function of turbulence strength. The fully parameterized RMS distribution model is used to provide a way to accurately predict the PLL for a range of turbulence strengths. We also propose and validate a new parametric average lucky look optical transfer function (OTF) model that could be used to aid in image restoration. Our OTF model blends the theoretical diffraction-limited OTF and the average turbulence short exposure OTF. Finally, we show simulated images for several anisoplanatic imaging scenarios that reveal the spatially varying nature of the RMS values impacting local image quality.

On the potential of ROI imaging in x-ray CT - A comparison of novel dynamic beam attenuators with current technology.

In this work, we explore the potential of region-of-interest (ROI) imaging in x-ray computed tomography (CT). Using two dynamic beam attenuator (DBA) concepts for fluence field modulation (FFM) previously developed, we investigate and evaluate the potential dose savings in comparison with current FFM technology. ROI imaging is a special application of FFM where the bulk of x-ray radiation is propagated toward a certain anatomical target (ROI), specified by the imaging task, while the surrounding tissue is spared from radiation. We introduce a criterion suitable to quantitatively describe the balance between image quality inside an ROI and total radiation dose with respect to a given ROI imaging task. It accounts for the mean image variance at the ROI and the effective patient dose calculated from Monte Carlosimulations. The criterion is further used to compile task-specific DBA trajectories determining the primary x-ray fluence, and eventually used for comparing different FFM techniques, namely the sheet-based dynamic beam attenuator(sbDBA), the z-aligned sbDBA(z-sbDBA), and an adjustable static operation mode of the z-sbDBA. Furthermore, two static bowtie filters and the influence of tube current modulation (TCM) are included in the comparison. Our findings demonstrate by simulations that the presented trajectory optimization method determines reasonable DBA trajectories. The influence of TCM is strongly depending on the imaging task. The narrow bowtie filter allows for dose reductions of about 10% compared to the regular bowtie filter in the considered ROI imaging tasks. The DBAs are shown to realize substantially larger dose reductions. In our cardiac imaging scenario, the DBAs can reduce the effective dose by about 30% (z-sbDBA) or 60% (sbDBA). We can further verify that the noise characteristics are not adversely affected by the DBAs. Our research demonstrates that ROI imaging using the presented DBA concepts is a promising technique toward a more patient- and task-specific CT imaging requiring lower radiation dose. Both the sbDBA and the z-sbDBA are potential technical solutions for realizing ROI imaging in x-ray CT.

Flexible Electromagnetic Cap for Three-Dimensional Electromagnetic Head Imaging.

The timely treatment is the crucial element for the survival of patients with brain stroke. Thus, a fast, cost-effective, and portable device is needed for the early and on-the-spot diagnosis of stroke patients. A 3D electromagnetic head imaging system for rapid brain stroke diagnosis with a wearable and lightweight platform is presented. The platform comprises a custom-built flexible cap with a 24-element planar antenna array, and a flexible matching medium layer. The custom-built cap is made out of an engineered polymer-ceramic composite substrate of RTV silicone rubber and aluminum oxide (Al2O3) for enhanced dielectric properties and mechanical flexibility and robustness. The array is arranged into two elliptical rings that are entirely incorporated into the flexible cap. The employed antenna elements within the system are compact with low SAR values over the utilized frequency range of 0.9-2.5 GHz. Moreover, a flexible matching medium layer is introduced on the front of the apertures of the antenna array to enhance the impedance matching with the skin. The detection capability of the system is experimentally verified on 3D realistic head phantoms at multiple imaging scenarios and different types of strokes. The reconstructed 3D and 2D multi-slice images using the beamforming and polar sensitivity encoding (PSE) image processing algorithms indicate the applicability and potential of the system for onsite brain imaging.