Imaging Scenarios Research Articles

Self-Supervised Learning for Few-Shot Medical Image Segmentation.

Fully-supervised deep learning segmentation models are inflexible when encountering new unseen semantic classes and their fine-tuning often requires significant amounts of annotated data. Few-shot semantic segmentation (FSS) aims to solve this inflexibility by learning to segment an arbitrary unseen semantically meaningful class by referring to only a few labeled examples, without involving fine-tuning. State-of-the-art FSS methods are typically designed for segmenting natural images and rely on abundant annotated data of training classes to learn image representations that generalize well to unseen testing classes. However, such a training mechanism is impractical in annotation-scarce medical imaging scenarios. To address this challenge, in this work, we propose a novel self-supervised FSS framework for medical images, named SSL-ALPNet, in order to bypass the requirement for annotations during training. The proposed method exploits superpixel-based pseudo-labels to provide supervision signals. In addition, we propose a simple yet effective adaptive local prototype pooling module which is plugged into the prototype networks to further boost segmentation accuracy. We demonstrate the general applicability of the proposed approach using three different tasks: organ segmentation of abdominal CT and MRI images respectively, and cardiac segmentation of MRI images. The proposed method yields higher Dice scores than conventional FSS methods which require manual annotations for training in our experiments.

Low‐Light Shadow Imaging Using Quadrature‐Noise Detection with a Camera

AbstractThis paper experimentally demonstrates an imaging technique based on quadrature noise modification after interaction with an opaque object. By using a homodyne‐like detection scheme, the detrimental effect of the camera dark noise is eliminated, making this approach particularly attractive for imaging scenarios which require weak illumination. As an example, the image of an object illuminated with a squeezed vacuum is reconstructed using a total of 800 probing photons, utilizing less than one photon per frame on average.

A perception and imaging framework for far-range frequency-division-multiplexed computational imaging

ABSTRACT So far, implementations of frequency-diverse computational framework with metamaterial element have used a frequency sweep across the operating frequency band, where the driving waveguide mode is coupled to subsets of the resonant elements at a corresponding excitation frequency. Scene information interrogating and processing efficiency are relatively confined in current imaging system especially for far-range imaging scenarios. Here, we implement frequency-division methods to parallelize computational imaging (CI). This concept multiplexes the modes in the frequency domain where different subsets of the resonant metamaterial elements are simultaneously excited by wideband multicarrier waveforms, allowing the scene to be encoded with different spatially diverse radiation patterns. Besides, we present a unified fast perception and processing framework for far-range imaging purpose. The proposed technique marries communications concepts to computational imaging, achieving a key development due to the limitations imposed by massive amount of data and burdensome computation. Numerical simulations with the measured radiation field pattern data are conducted to verify the effectiveness the technique.

Real-time physical compression computational ghost imaging based on array spatial light field modulation and deep learning

Ghost imaging has great application potential in remote sensing observation and biomedical due to the use of a single pixel detector for imaging, which makes it easy to imaging in low light conditions and imperfect spectral regions of the camera. However, the higher the reconstructed image resolution (the more pixels), the larger the number of measurements are required, which makes it difficult to realize the real-time imaging required in the application. Therefore, a new real-time computational ghost imaging method based on deep learning technology and array spatial light field modulation is proposed, which can realize high-quality imaging with high magnification down sampling. Furthermore, the physical compression imaging of objects is realized by adjusting the array light field composed of several identical sub-light fields, and the original image of the object is successfully obtained from the compressed image by deep learning. The resulting compressed image is realized by the physical means of array light field regulation, and allows the far-field transmission of physical compressed image (or use the image compression algorithm for deep compression of the physical compressed image) with low storage data to reduce the transmission data, and then use deep learning to obtain the original image. We implement the proposed method in numerical simulation and real-time imaging experiments of multifold down sampling and physical compressed (eg, 4x, 9x, 16x), and demonstrate its validity and extensibility. This makes the method have a great potential application and prospect in the application scenarios of earth observation imaging and Limited space-time transmission window.

3D printed anthropomorphic left ventricular myocardial phantom for nuclear medicine imaging applications

BackgroundAnthropomorphic torso phantoms, including a cardiac insert, are frequently used to investigate the imaging performance of SPECT and PET systems. These phantom solutions are generally featuring a simple anatomical representation of the heart. 3D printing technology paves the way to create cardiac phantoms with more complex volume definition. This study aimed to describe how a fillable left ventricular myocardium (LVm) phantom can be manufactured using geometry extracted from a patient image.MethodsThe LVm of a healthy subject was segmented from 18F-FDG attenuation corrected PET image set. Two types of phantoms were created and 3D printed using polyethylene terephthalate glycol (PETG) material: one representing the original healthy LVm, and the other mimicking myocardium with a perfusion defect. The accuracy of the LVm phantom production was investigated by high-resolution CT scanning of 3 identical replicas. 99mTc SPECT acquisitions using local cardiac protocol were performed, without additional scattering media (“in air” measurements) for both phantom types. Furthermore, the healthy LVm phantom was inserted in the commercially available DataSpectrum Anthropomorphic Torso Phantom (“in torso” measurement) and measured with hot background and hot liver insert.ResultsPhantoms were easy to fill without any air-bubbles or leakage, were found to be reproducible and fully compatible with the torso phantom. Seventeen segments polar map analysis of the "in air” measurements revealed that a significant deficit in the distribution appeared where it was expected. 59% of polar map segments had less than 5% deviation for the "in torso” and "in air” measurement comparison. Excluding the deficit area, neither comparison had more than a 12.4% deviation. All the three polar maps showed similar apex and apical region values for all configurations.ConclusionsFillable anthropomorphic 3D printed phantom of LVm can be produced with high precision and reproducibility. The 3D printed LVm phantoms were found to be suitable for SPECT image quality tests during different imaging scenarios. The flexibility of the 3D printing process presented in this study provides scalable and anthropomorphic image quality phantoms in nuclear cardiology imaging.

MR Image Based Brain Tumor Classification with Deep Learning Neural Networks

The unique combination of Artificial Intelligence and Machine Learning, which helps the computer to imitate the ways and behaviour of human beings can be termed as deep learning. The field of deep learning is an emerging field that has gained a lot of interest toward past years. The Deep Learning have proven already to solve the complex problem using the powerful machine learning tools. One of the best deep learning algorithm is used to classify the brain tumor data set in this paper. The deep learning architecture is able to classify the brain tumor into 4 categories of images. The first being no tumor, the second being pituitary tumor, the third is meningioma and the last one classified as glioma. As we are well aware, the training datasets for the medical imaging scenario are very few. This is a challenging task to apply the deep learning that is obtained from a trained CNN model to dig up the small data set to attain the result. A pre trained CNN model is used here to solve the problem. The obtained results are good over all Performance is measured.

Diagnostic Accuracy of MRI in Local Staging (T Category) of Penile Cancer and the Value of Artificial Erection: A Systematic Review and Meta-Analysis.

BACKGROUND. Treatment recommendations of penile cancers are determined primarily by the local extent of the primary tumor. Clinical palpation is used for local staging. OBJECTIVE. We reviewed diagnostic performance of MRI in local staging of penile cancer in three clinical scenarios (questions [Qs] 1 through 3, Q1-Q3) and one imaging scenario (Q4). Q1 asked whether MRI reliably distinguishes ≤ T1 from ≥ T2 disease. Q2 asked whether clinical staging reliably identifies ≤ T1 versus ≥ T2 disease and how clinical staging compares to MRI. Q3 asked if MRI is accurate for diagnosis of T3 disease. Q4 asked if artificial erection (by intracavernosal injection of prostaglandin E1) improved accuracy of MRI in T categorization. EVIDENCE ACQUISITION. MEDLINE, EMBASE, and Cochrane Library databases were searched through September 13, 2021, for studies evaluating local staging of penile cancer using MRI with surgical pathology as the reference standard. Diagnostic accuracy was calculated using a bivariate random-effects model and hierarchic summary ROC mode Meta-regression was performed to test for covariate effects of MRI and artificial erection in Q3 and Q4, respectively. EVIDENCE SYNTHESIS. Eight studies and 481 patients were included. The sensitivity and specificity of MRI for Q1 were 86% (95% CI, 73-94%) and 89% (95% CI, 77-95%), respectively. AUC for MRI (0.94; 95% CI, 0.92-0.96) did not differ from clinical staging (0.87; 95% CI, 0.84-0.90; p = .83). For Q3, MRI had sensitivity and specificity of 80% (95% CI, 70-87%) and 96% (95% CI, 85-99%), respectively. For Q4, sensitivity and specificity for MRI with versus without artificial erection were 85% (95% CI, 71-92%) and 93% (95% CI, 77-98%) versus 86% (95% CI, 68-95%) and 84% (95% CI, 70-93%), respectively (p = .50). CONCLUSION. MRI staging of penile cancer may be considered for ≤ T1 versus ≥ T2 disease but did not appear more accurate than clinical staging. High specificity of MRI for diagnosis of ≥ T3 disease suggests that MRI may be useful when organ-sparing approaches are planned. MRI with and without artificial erection showed similar accuracy in local staging. CLINICAL IMPACT. MRI, with or without artificial erection, may be valuable in routine preoperative evaluation of local staging of penile cancer, particularly when organ-sparing options are considered.

Rapid and automatic burned area detection using sentinel-2 time-series images in google earth engine cloud platform: a case study over the Andika and Behbahan Regions, Iran.

For proper forest management, accurate detection and mapping of burned areas are needed, yet the practice is difficult to perform due to the lack of an appropriate method, time, and expense. It is also critical to obtain accurate information about the density and distribution of burned areas in a large forest and vegetated areas. For the most efficient and up-to-date mapping of large areas, remote sensing is one of the best technologies. However, the complex image scenario and the similar spectral behavior of classes in multispectral satellite images may lead to many false-positive mistakes, making it challenging to extract the burned areas accurately. This research aims to develop an automated framework in the Google Earth Engine (GEE) cloud computing platform for detecting burned areas in Andika and Behbahan, located in the south and southwest of Iran, using Sentinel-2 time-series images. After importing the images and applying the necessary preprocessing, the Sentinel-2 Burned Areas Index (BAIS2) was used to create a map of the Primary Burned Areas (PBA). Detection accuracy was then improved by masking out disturbing classes (vegetation and water) on the PBA map, which resulted in Final Burned Areas (FBA). The unimodal method is used to calculate the ideal thresholds of indices to make the proposed method automatic. The final results demonstrated that the proposed method performed well in both homogeneous and heterogeneous areas for detecting the burned areas. Based on a test dataset, maps of burned areas were produced in the Andika and Behbahan regions with an overall accuracy of 90.11% and 92.40% and a kappa coefficient of 0.87 and 0.88, respectively, which were highly accurate when compared to the BAIS2, Normalized Burn Ratio (NBR), Normalized Difference Vegetation Index (NDVI), Mid-Infrared Bispectral Index (MIRBI), and Normalized Difference SWIR (NDSWIR) indices. Based on the results, accurate determination of vegetation classes and water zones and eliminating them from the map of burned areas led to a considerable increase in the accuracy of the obtained final map from the BAIS2 spectral index.

Distributed Edge-Enhanced Imaging With a Fractional Spiral Phase Filter Using Random Light

A gradual edge-enhanced ghost imaging method with pseudo-thermal light is proposed in both the theory and experiment. In the experiment, a phase object and fractional spiral phase filter are placed symmetrically in the imaging plane of the pseudo-thermal light source in the distributed test and reference beams of the lensless ghost imaging system. The procedure of gradual edge-enhanced ghost imaging is carried out by modulating the fractional topological charge from 0 to 1. We observe that the brightness of the object edge increases with the increase of the fractional topological charge. It is also found that the intensity distribution is uniform and isotropic when the topological charge is an integer; otherwise, the intensity distribution is not uniform. Theoretical analysis is also provided. The proposed gradual edge-enhanced ghost imaging scenario releases the position limitation in the Fourier plane for the filter of the traditional phase filtering imaging process. The method is believed to have prospective applications in microscopic imaging and biomedical detection.

Fast CBCT Reconstruction using Convolutional Neural Networks for Arbitrary Robotic C-arm Orbits.

Cone-beam CT (CBCT) with non-circular acquisition orbits has the potential to improve image quality, increase the field-of view, and facilitate minimal interference within an interventional imaging setting. Because time is of the essence in interventional imaging scenarios, rapid reconstruction methods are advantageous. Model-Based Iterative Reconstruction (MBIR) techniques implicitly handle arbitrary geometries; however, the computational burden for these approaches is particularly high. The aim of this work is to extend a previously proposed framework for fast reconstruction of non-circular CBCT trajectories. The pipeline combines a deconvolution operation on the backprojected measurements using an approximate, shift-invariant system response prior to processing with a Convolutional Neural Network (CNN). We trained and evaluated the CNN for this approach using 1800 randomized arbitrary orbits. Noisy projection data were formed from 1000 procedurally generated tetrahedral phantoms as well as anthropomorphic data in the form of 800 CT and CBCT images from the Lung Image Database Consortium Image Collection (LIDC). Using this proposed reconstruction pipeline, computation time was reduced by 90% as compared to MBIR with only minor differences in performance. Quantitative comparisons of nRMSE, FSIM and SSIM are reported. Performance was consistent for projection data simulated with acquisition orbits the network has not previously been trained on. These results suggest the potential for fast processing of arbitrary CBCT trajectory data with reconstruction times that are clinically relevant and applicable - facilitating the application of non-circular orbits in CT image-guided interventions and intraoperative imaging.

Technical assessment of 2D and 3D imaging performance of an IGZO-based flat-panel X-ray detector.

Indirect detection flat-panel detectors (FPDs) consisting of hydrogenated amorphous silicon (a-Si:H) thin-film transistors (TFTs) are a prevalent technology for digital x-ray imaging. However, their performance is challenged in applications requiring low exposure levels, high spatial resolution, and high frame rate. Emerging FPD designs using metal oxide TFTs may offer potential performance improvements compared to FPDs based on a-Si:H TFTs. This work investigates the imaging performance of a new indium gallium zinc oxide (IGZO) TFT-based detector in 2D fluoroscopy and 3D cone-beam CT (CBCT). The new FPD consists of a sensor array combining IGZO TFTs with a-Si:H photodiodes and a 0.7-mm thick CsI:Tl scintillator. The FPD was implemented on an x-ray imaging bench with system geometry emulating intraoperative CBCT. A conventional FPD with a-Si:H TFTs and a 0.6-mm thick CsI:Tl scintillator was similarly implemented as a basis of comparison. 2D imaging performance was characterized in terms of electronic noise, sensitivity, linearity, lag, spatial resolution (modulation transfer function, MTF), image noise (noise-power spectrum, NPS), and detective quantum efficiency (DQE) with entrance air kerma (EAK) ranging from 0.3 to 1.2 μGy. 3D imaging performance was evaluated in terms of the 3D MTF and noise-equivalent quanta (NEQ), soft-tissue contrast-to-noise ratio (CNR), and image quality evident in anthropomorphic phantoms for a range of anatomical sites and dose, with weighted air kerma, , ranging from 0.8 to 4.9 mGy. The 2D imaging performance of the IGZO-based FPD exhibited up to ∼1.7× lower electronic noise than the a-Si:H FPD at matched pixel pitch. Furthermore, the IGZO FPD exhibited ∼27% increase in mid-frequency DQE (1mm-1 ) at matched pixel size and dose (EAK ≈ 1.0 μGy) and ∼11% increase after adjusting for differences in scintillator thickness. 2D spatial resolution was limited by the scintillator for each FPD. The IGZO-based FPD demonstrated improved 3D NEQ at all spatial frequencies in both head (≥25% increase for all dose levels) and body (≥10% increase for ≤2 mGy) imaging scenarios. These characteristics translated to improved low-contrast visualization in anthropomorphic phantoms, demonstrating ≥10% improvement in CNR and extension of the low-dose range for which the detector is input-quantum limited. The IGZO-based FPD demonstrated improvements in electronic noise, image lag, and NEQ that translated to measurable improvements in 2D and 3D imaging performance compared to a conventional FPD based on a-Si:H TFTs. The improvements are most beneficial for 2D or 3D imaging scenarios involving low-dose and/or high-frame rate.

Weakly supervised segmentation with cross-modality equivariant constraints.

Weakly supervised learning has emerged as an appealing alternative to alleviate the need for large labeled datasets in semantic segmentation. Most current approaches exploit class activation maps (CAMs), which can be generated from image-level annotations. Nevertheless, resulting maps have been demonstrated to be highly discriminant, failing to serve as optimal proxy pixel-level labels. We present a novel learning strategy that leverages self-supervision in a multi-modal image scenario to significantly enhance original CAMs. In particular, the proposed method is based on two observations. First, the learning of fully-supervised segmentation networks implicitly imposes equivariance by means of data augmentation, whereas this implicit constraint disappears on CAMs generated with image tags. And second, the commonalities between image modalities can be employed as an efficient self-supervisory signal, correcting the inconsistency shown by CAMs obtained across multiple modalities. To effectively train our model, we integrate a novel loss function that includes a within-modality and a cross-modality equivariant term to explicitly impose these constraints during training. In addition, we add a KL-divergence on the class prediction distributions to facilitate the information exchange between modalities which, combined with the equivariant regularizers further improves the performance of our model. Exhaustive experiments on the popular multi-modal BraTS and prostate DECATHLON segmentation challenge datasets demonstrate that our approach outperforms relevant recent literature under the same learning conditions.

MSMT-CNN for Solar Active Region Detection with Multi-Spectral Analysis

Precisely detecting solar active regions (AR) from multi-spectral images is a challenging task yet important in understanding solar activity and its influence on space weather. A main challenge comes from each modality capturing a different location of these 3D objects, as opposed to more traditional multi-spectral imaging scenarios where all image bands observe the same scene. We present a multi-task deep learning framework that exploits the dependencies between image bands to produce 3D AR detection where different image bands (and physical locations) each have their own set of results. Different feature fusion strategies are investigated in this work, where information from different image modalities is aggregated at different semantic levels throughout the network. This allows the network to benefit from the joint analysis while preserving the band-specific information. We compare our detection method against baseline approaches for solar image analysis (multi-channel coronal hole detection, SPOCA for ARs (Verbeeck et al. Astron Astrophys 561:16, 2013)) and a state-of-the-art deep learning method (Faster RCNN) and show enhanced performances in detecting ARs jointly from multiple bands. We also evaluate our proposed approach on synthetic data of similar spatial configurations obtained from annotated multi-modal magnetic resonance images.

A Three-Dimensional Microwave Sparse Imaging Approach Using Higher-Order Basis Functions

This paper presents a novel method for three-dimensional microwave imaging based on sparse processing. To enforce the sparsity of the unknown function, we take advantage of the fact that arbitrary three-dimensional electromagnetic fields can be decomposed into two components with respect to the radial direction: one with transverse-magnetic polarization and the other with transverse-electric polarization. Each component can be further expressed as a sum of spherical harmonics, which provide the dictionary exploited by the sparse processing algorithm. Our measurement model relates the data and the parameters of the spherical harmonics’ sources, which are uniformly distributed on a grid sampling the imaging domain. By relying on the theory of degrees of freedom of electromagnetic fields, it can be shown that only a few harmonics are sufficient to accurately represent the measured scattered field from objects whose diameter is of the order of the wavelength, thus allowing reducing the dimension of the adopted dictionary. We analyze several imaging scenarios to assess the algorithm’s performance, including different object shapes, sensor orientations, and signal-to-noise ratios. Moreover, we compare the obtained results with other state-of-the-art linear imaging techniques. Notably, thanks to the adopted dictionary, the proposed algorithm can yield accurate images of both convex and concave objects.

What should be the input: Investigating the environment representations in sim-to-real transfer for navigation tasks

While training an end-to-end navigation network in the real world is usually costly, simulation serves as a safe and low-cost tool in this training process. However, training neural network models in simulation brings up the problem of effectively transferring the model from simulation to the real world (sim-to-real). In this work, we regard the environment representation as a crucial element in this transfer process, and we propose a visual information pyramid (VIP) model to investigate a practical environment representation systematically. A novel representation composed of spatial and semantic information synthesis is established accordingly, where noise model embedding is particularly considered. To explore the effectiveness of the proposed representation, we compared its performance with other popularly used representations in the literature, such as RGB image, depth image, and semantic segmentation image, in both simulated and real-world scenarios. Results suggest that our environment representation stands out. Furthermore, an analysis on the feature map is implemented to investigate the effectiveness through hidden layer reaction, which could be irradiative for future researches on sim-to-real learning-based navigation.

Traveling Salesman Problem with neighborhoods on a sphere in reflectance transformation imaging scenarios

In this paper, we propose a solution to the non-Euclidean variant of the Traveling Salesman Problem with Neighborhoods on a Sphere (TSPNS). The introduced problem formulation is motivated by practical scenarios of employing unmanned aerial vehicles in the Reflectance Transformation Imaging (RTI). In the RTI, a vehicle is requested to visit a set of sites at a constant distance from the object of interest and cast light from different directions to model the object from the images captured from another fixed location. Even though the problem can be formulated as an instance of the regular traveling salesman problem, we report a significant reduction of the solution cost by exploiting a non-zero tolerance on the light direction and defining the sites as regions on a sphere. The continuous neighborhoods of the TSPNS can be sampled into discrete sets, and the problem can be transformed into the generalized traveling salesman problem. However, finding high-quality solutions requires dense sampling, which increases the computational requirements. Therefore, we propose a practical heuristic solution based on the unsupervised learning of the Growing Self-Organizing Array (GSOA) that quickly finds an initial solution with the competitive quality to the sampling-based method. Furthermore, we propose a fast post-processing optimization to improve the initial solutions of both solvers. Based on the reported results, the proposed GSOA-based solver provides solutions of a similar quality to the transformation approach while it is about two orders of magnitude less computationally demanding.

Real-Time Imaging With Simultaneous Use of Born and Rytov Approximations in Quantitative Microwave Holography

Microwave and millimeter-wave measurements acquire total-field responses from measurements, yet imaging algorithms instead require the data in the form of scattered-field responses. Two approaches exist for the extraction of the scattered-field data from the total-field responses, namely, the Born and the Rytov data approximations. It is well known that, depending on the target’s size, contrast, and structural complexity, one approximation can achieve an improved accuracy over the other. Yet, the Rytov approximation is rarely used in microwave and millimeter-wave imaging, likely due to phase-unwrapping problems occurring in the case of strongly heterogeneous electrically large targets. Here, we propose a method to utilize the Born and the Rytov approximations simultaneously in a single inversion step for real-time imaging with quantitative microwave holography (QMH). We show through examples with simulated and experimental data that in near-field imaging scenarios, including the imaging of a breast-tissue phantom, there are significant benefits in employing the new method.

Embracing the disharmony in medical imaging: A Simple and effective framework for domain adaptation.

Domain shift, the mismatch between training and testing data characteristics, causes significant degradation in the predictive performance in multi-source imaging scenarios. In medical imaging, the heterogeneity of population, scanners and acquisition protocols at different sites presents a significant domain shift challenge and has limited the widespread clinical adoption of machine learning models. Harmonization methods, which aim to learn a representation of data invariant to these differences are the prevalent tools to address domain shift, but they typically result in degradation of predictive accuracy. This paper takes a different perspective of the problem: we embrace this disharmony in data and design a simple but effective framework for tackling domain shift. The key idea, based on our theoretical arguments, is to build a pretrained classifier on the source data and adapt this model to new data. The classifier can be fine-tuned for intra-study domain adaptation. We can also tackle situations where we do not have access to ground-truth labels on target data; we show how one can use auxiliary tasks for adaptation; these tasks employ covariates such as age, gender and race which are easy to obtain but nevertheless correlated to the main task. We demonstrate substantial improvements in both intra-study domain adaptation and inter-study domain generalization on large-scale real-world 3D brain MRI datasets for classifying Alzheimer's disease and schizophrenia.

Evaluating metallic artefact of biodegradable magnesium-based implants in magnetic resonance imaging

Magnesium (Mg) implants have shown to cause image artefacts or distortions in magnetic resonance imaging (MRI). Yet, there is a lack of information on how the degradation of Mg-based implants influences the image quality of MRI examinations. In this study, Mg-based implants are analysed in vitro, ex vivo, and in the clinical setting for various magnetic field strengths with the aim to quantify metallic artefact behaviour. In vitro corroded Mg-based screws and a titanium (Ti) equivalent were imaged according to the ASTM F2119. Mg-based and Ti pins were also implanted into rat femurs for different time points and scanned to provide insights on the influence of soft and hard tissue on metallic artefact. Additionally, MRI data of patients with scaphoid fractures treated with CE-approved Mg-based compression screws (MAGNEZIX®) were analysed at various time points post-surgery. The artefact production of the Mg-based material decreased as implant material degraded in all settings. The worst-case imaging scenario was determined to be when the imaging plane was selected to be perpendicular to the implant axis. Moreover, the Mg-based implant outperformed the Ti equivalent in all experiments by producing lower metallic artefact (p < 0.05). This investigation demonstrates that Mg-based implants generate significantly lower metallic distortion in MRI when compared to Ti. Our positive findings suggest and support further research into the application of Mg-based implants including post-operative care facilitated by MRI monitoring of degradation kinetics and bone/tissue healing processes.

Accurate 3-D Microwave Near-Field Imaging of Typical Concave Objects

High-resolution microwave and millimeter-wave (MMW) detection and recognition are often plagued with artifacts when facing complex objects with concave structures due to the ignorance of high-order scattering propagation. Previous works have demonstrated the effectiveness of accurate near-field reconstruction by considering multiple reflections reconstruction in planar dihedral structures. In this letter, we extend the proposed imaging model to three-dimensional (3-D) imaging scenario. Based on the shooting and bouncing ray (SBR) concept, we first analyze the mechanism of multiple scattering and establish a 3-D forward model. Utilizing circularly polarized measurements, each number of reflection times (RTs) echo can be separated. Also, by calculating and compensating for the multiple reflection propagation paths, the accurate 3-D reconstruction of dihedrals is realized. Numerical simulations and experiments show that the proposed method can obtain the high resolution in the <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$z$ </tex-math></inline-formula> -axis and effectively restore the pattern of reflectivity distribution on the target surface.