Search

Deep learning (DL) models are capable of successfully exploiting latent representations in MR data and have become state-of-the-art for accelerated MRI reconstruction. However, undersampling the measurements in k-space as well as the over- or under-parameterized and non-transparent nature of DL make these models exposed to uncertainty. Consequently, uncertainty estimation has become a major issue in DL MRI reconstruction. To estimate uncertainty, Monte Carlo (MC) inference techniques have become a common practice where multiple reconstructions are utilized to compute the variance in reconstruction as a measurement of uncertainty. However, these methods demand high computational costs as they require multiple inferences through the DL model. To this end, we introduce a method to estimate uncertainty during MRI reconstruction using a pixel classification framework. The proposed method, PixCUE (stands for Pixel Classification Uncertainty Estimation) produces the reconstructed image along with an uncertainty map during a single forward pass through the DL model. We demonstrate that this approach generates uncertainty maps that highly correlate with the reconstruction errors with respect to various MR imaging sequences and under numerous adversarial conditions. We also show that the estimated uncertainties are correlated to that of the conventional MC method. We further provide an empirical relationship between the uncertainty estimations using PixCUE and well-established reconstruction metrics such as NMSE, PSNR, and SSIM. We conclude that PixCUE is capable of reliably estimating the uncertainty in MRI reconstruction with a minimum additional computational cost., 19 pages, 7 figures, 1 table

Purpose: We propose a novel contrastive learning latent space representation for MRI datasets with partially acquired scans. We show that this latent space can be utilized for accelerated MR image reconstruction. Theory and Methods: Our novel framework, referred to as COLADA (stands for Contrastive Learning for highly accelerated MR image reconstruction), maximizes the mutual information between differently accelerated images of an MRI scan by using self-supervised contrastive learning. In other words, it attempts to "pull" the latent representations of the same scan together and "push" the latent representations of other scans away. The generated MRI latent space is subsequently utilized for MR image reconstruction and the performance was assessed in comparison to several baseline deep learning reconstruction methods. Furthermore, the quality of the proposed latent space representation was analyzed using Alignment and Uniformity. Results: COLADA comprehensively outperformed other reconstruction methods with robustness to variations in undersampling patterns, pathological abnormalities, and noise in k-space during inference. COLADA proved the high quality of reconstruction on unseen data with minimal fine-tuning. The analysis of representation quality suggests that the contrastive features produced by COLADA are optimally distributed in latent space. Conclusion: To the best of our knowledge, this is the first attempt to utilize contrastive learning on differently accelerated images for MR image reconstruction. The proposed latent space representation has practical usage due to a large number of existing partially sampled datasets. This implies the possibility of exploring self-supervised contrastive learning further to enhance the latent space of MRI for image reconstruction.

A multi-bifurcated cantilever piezoelectric energy harvester (BCPEH) is designed and verified to achieve a wide and adjustable response frequency band. The theoretical model is derived based on the Euler-Bernoulli beam theory and continuity boundary conditions to investigate the dynamic response of the BCPEH. The displacement frequency response function and the voltage frequency response function of the BCPEH are deduced based on the Galerkin method, and the theoretical results of a typical multi-bifurcated cantilever piezoelectric energy harvester, the Y-shaped BCPEH, are verified by the finite element method (FEM) and experiments. In addition, by comparing experimental output power of the Y-shaped BCPEH with that of the traditional cantilever-based piezoelectric energy harvester with the same mass of the bifurcated part at the beam-tip, it demonstrates that the Y-shaped BCPEH has a wider operational frequency band. Moreover, it is found that the Y-shaped BCPEH can be designed with an asymmetric configuration to adjust its response frequency distribution. The number of resonant frequencies and the output power of the asymmetric Y-shaped BCPEH are higher than that of the symmetric Y-shaped BCPEH. And the Y-shaped BCPEH has even better performance than L-shaped BCPEH. This study provides a new design concept for enhanced energy harvester.

As intensities of MRI volumes are inconsistent across institutes, it is essential to extract universal features of multi-modal MRIs to precisely segment brain tumors. In this concept, we propose a volumetric vision transformer that follows two windowing strategies in attention for extracting fine features and local distributional smoothness (LDS) during model training inspired by virtual adversarial training (VAT) to make the model robust. We trained and evaluated network architecture on the FeTS Challenge 2022 dataset. Our performance on the online validation dataset is as follows: Dice Similarity Score of 81.71%, 91.38% and 85.40%; Hausdorff Distance (95%) of 14.81 mm, 3.93 mm, 11.18 mm for the enhancing tumor, whole tumor, and tumor core, respectively. Overall, the experimental results verify our method's effectiveness by yielding better performance in segmentation accuracy for each tumor sub-region. Our code implementation is publicly available : https://github.com/himashi92/vizviva_fets_2022

Global correlations are widely seen in human anatomical structures due to similarity across tissues and bones. These correlations are reflected in magnetic resonance imaging (MRI) scans as a result of close-range proton density and T1/T2 parameters. Furthermore, to achieve accelerated MRI, k-space data are undersampled which causes global aliasing artifacts. Convolutional neural network (CNN) models are widely utilized for accelerated MRI reconstruction, but those models are limited in capturing global correlations due to the intrinsic locality of the convolution operation. The self-attention-based transformer models are capable of capturing global correlations among image features, however, the current contributions of transformer models for MRI reconstruction are minute. The existing contributions mostly provide CNN-transformer hybrid solutions and rarely leverage the physics of MRI. In this paper, we propose a physics-based stand-alone (convolution free) transformer model titled, the Multi-head Cascaded Swin Transformers (McSTRA) for accelerated MRI reconstruction. McSTRA combines several interconnected MRI physics-related concepts with the transformer networks: it exploits global MR features via the shifted window self-attention mechanism; it extracts MR features belonging to different spectral components separately using a multi-head setup; it iterates between intermediate de-aliasing and k-space correction via a cascaded network with data consistency in k-space and intermediate loss computations; furthermore, we propose a novel positional embedding generation mechanism to guide self-attention utilizing the point spread function corresponding to the undersampling mask. Our model significantly outperforms state-of-the-art MRI reconstruction methods both visually and quantitatively while depicting improved resolution and removal of aliasing artifacts.

Segmentation of images is a long-standing challenge in medical AI. This is mainly due to the fact that training a neural network to perform image segmentation requires a significant number of pixel-level annotated data, which is often unavailable. To address this issue, we propose a semi-supervised image segmentation technique based on the concept of multi-view learning. In contrast to the previous art, we introduce an adversarial form of dual-view training and employ a critic to formulate the learning problem in multi-view training as a min-max problem. Thorough quantitative and qualitative evaluations on several datasets indicate that our proposed method outperforms state-of-the-art medical image segmentation algorithms consistently and comfortably. The code is publicly available at https://github.com/himashi92/Duo-SegNet

We propose a Transformer architecture for volumetric segmentation, a challenging task that requires keeping a complex balance in encoding local and global spatial cues, and preserving information along all axes of the volume. Encoder of the proposed design benefits from self-attention mechanism to simultaneously encode local and global cues, while the decoder employs a parallel self and cross attention formulation to capture fine details for boundary refinement. Empirically, we show that the proposed design choices result in a computationally efficient model, with competitive and promising results on the Medical Segmentation Decathlon (MSD) brain tumor segmentation (BraTS) Task. We further show that the representations learned by our model are robust against data corruptions. \href{https://github.com/himashi92/VT-UNet}{Our code implementation is publicly available}.

This thesis was scanned from the print manuscript for digital preservation and is copyright the author. Researchers can access this thesis by asking their local university, institution or public library to make a request on their behalf. Monash staff and postgraduate students can use the link in the References field.

Simultaneous FDG-PET/fMRI ([18F]-fluorodeoxyglucose positron emission tomography functional magnetic resonance imaging) provides the capacity to image two sources of energetic dynamics in the brain – glucose metabolism and haemodynamic response. Functional fMRI connectivity has been enormously useful for characterising interactions between distributed brain networks in humans. Metabolic connectivity based on static FDG-PET has been proposed as a biomarker for neurological disease; but static FDG-PET cannot be used to estimate subjectlevel measures of connectivity, only across-subject covariance. Here, we applied high-temporal resolution constant infusion fPET to measure subject-level metabolic connectivity simultaneously with fMRI connectivity. fPET metabolic connectivity was characterised by fronto-parietal connectivity within and between hemispheres. fPET metabolic connectivity showed moderate similarity with fMRI primarily in superior cortex and frontoparietal regions. Significantly, fPET metabolic connectivity showed little similarity with static FDG-PET metabolic covariance, indicating that metabolic brain connectivity is a non-ergodic process whereby individual brain connectivity cannot be inferred from group level metabolic covariance. Our results highlight the complementary strengths of fPET and fMRI in measuring the intrinsic connectivity of the brain, and open up the opportunity for novel fundamental studies of human brain connectivity as well as multi-modality biomarkers of neurological diseases.

In an attempt to improve the capacitive deionization (CDI) performance and develop an efficient CDI electrode material, the effect of carbon characteristics on CDI was investigated. Activated carbon (AC) and ordered mesoporous carbon (OMC) materials synthesized by template method were used for the electrode manufacture. The capacitive deionization performance of AC and OMC based electrodes was investigated and compared. We found that when the specific surface areas of AC materials increased from 712 to 1347 m2/g, the electrosorption capacities of AC electrodes increased slightly from 9.43 to 10.14 mg/g. The characteristics of OMC materials can be optimized by the adjustment of the mesoporous silica beads template proportion, resulting in a maximum average pore size of 11.35 nm and total pore volume of 2.101 cm3/g. The optimum electrosorption capacity obtained with OMC4 electrode was 10.83 mg/g, which was not only higher than that of AC electrodes, but obviously higher than the values of most carbon-based electrodes in other studies. In conclusion, these results provide important insights into the effect of carbon characteristics on capacitive deionization and suggest that OMC synthesized by template method could be a promising carbon material for the CDI electrode manufacture.

The formation of emerging nitrogenous disinfection by-products (N-DBPs) from the chlorination of aspartic acid (Asp) was investigated. The yield of dichloroacetonitrile (DCAN) was higher than other N-DBPs, such as dichloroacetamide(DCAcAm) and chloropicrin (TCNM) during the chlorination of Asp. The formation of DCAN, DCAcAm, and TCNM all showed a trend of first increasing and then decreasing during the chlorination of Asp with increasing contact time. The dosage of chlorine had an impact on the formation of DCAN, DCAcAm, and TCNM. The highest yields of DCAN and DCAcAm appeared when the Cl2/Asp molar ratio was about 20, the yield of TCNM increased with increasing the Cl2/Asp molar ratio from 5 to 30 and TCNM was not produced when the ratio was less than 5. Cyanogen chloride (CNCl) was detected when the Cl2/Asp molar ratio was lower than 5. N-DBPs formation was influenced by pH. DCAN formation increased with increasing pH from 5 to 6 and then decreased with increasing pH from 6 to 9, but DCAcAm and TCNM increased with increasing pH from 5 to 8 and then decreased. Higher temperatures reduced the formation of DCAN and DCAcAm, but increased TCNM formation. DCAN and DCAcAm formation decreased, and relatively stable TCNM formation increased, with increasing free chlorine contact time during chloramination. N-nitrosodimethylamine (NDMA) was produced during chloramination of Asp and increased with prolonged chloramination contact time. The presence of bromide ions enhanced the yields of haloacetonitriles and shifted N-DBPs to more brominated species.

This thesis studies the network performance of the Robot Operating System (ROS) 2 when used in a network of nodes similar to how a group of unmanned assets would operate. Specifically, this thesis evaluates the impact of combining varying Quality of Service (QoS) and security settings in the ROS 2. It also explores the effect that scaling to multiple nodes has on network performance. This is the first work to comprehensively study ROS 2 network performance using QoS and security classification as a function of scale and message size. Network performance metrics include latency and message drop rate between nodes. Our research uniquely integrates ROS 2 with NS-3, developing a simulation architecture that is effective for rapidly studying ROS 2 network performance. Our simulation results demonstrated the trade-offs in choosing different QoS policies as well as the trade-offs in performance when security settings were enabled. We found that enabling security resulted in a higher message drop rate across all QoS profiles. We also found that scaling the network to more nodes resulted in various consequences with the use of different QoS settings. Scaling up to more nodes in a network also resulted in an equivalent increase in the average latency of messages. This work contributes to evaluating and configuring ROS 2 parameters for different unmanned system use cases while providing a simulation framework on which tests can be run. http://archive.org/details/performanceanaly1094563441 Major, Republic of Singapore Air Force Approved for public release; distribution is unlimited.

Traditional fleet operations and technologies are not adequately suited to counter the growing threat to undersea infrastructure from autonomous undersea systems. A cost-effective unmanned and manned system of systems is required to provide defense of this seabed infrastructure. This paper proposes possible system architectures to defend against this emerging threat to include passive barriers and active defense systems. The effectiveness of those candidate systems is evaluated through multiple agent-based modeling simulations of UUV versus UUV engagements. Analysis resulted in two major findings. First, point defense of critical assets is more effective than barrier defense. Second, system design must focus on minimizing the time required to effectively engage and neutralize threats, either through improvement to defensive UUV speed or investment in more UUV docking stations and sensor arrays. Cost analysis suggests that acquisition and operations cost of the recommended defensive system is less than the projected financial impact of a successful attack. http://archive.org/details/seabedinfrastruc1094562767 Lieutenant, United States Navy Lieutenant, United States Navy Lieutenant, United States Navy Major, Israel Defence Forces Major, Republic of Singapore Air Force Major, Republic of Singapore Air Force Captain, Singapore Army Lieutenant, United States Navy Lieutenant, United States Navy Lieutenant, United States Navy Major, Republic of Singapore Air Force Captain, Singapore Army Civilian, Ministry of Defense, Singapore Lieutenant, United States Navy Lieutenant Commander, United States Navy Lieutenant Junior Grade, United States Navy Civilian, Ministry of Defense, Singapore Civilian, Ministry of Defense, Singapore Major, Republic of Singapore Air Force Major, United States Marine Corps Major, Singapore Army Approved for public release; distribution is unlimited.

Studies of task-evoked brain activity are the cornerstone of cognitive neuroscience, and unravel the spatial and temporal brain dynamics of cognition in health and disease. Blood oxygenation level dependent functional magnetic resonance imaging (BOLD-fMRI) is one of the most common methods of studying brain function in humans. BOLD-fMRI indirectly infers neuronal activity from regional changes in blood oxygenation and is not a quantitative metric of brain function. Regional variation in glucose metabolism, measured using [18-F] fluorodeoxyglucose positron emission tomography (FDG-PET), provides a more direct and interpretable measure of neuronal activity. However, while the temporal resolution of BOLD-fMRI is in the order of seconds, standard FDG-PET protocols provide a static snapshot of glucose metabolism. Here, we develop a novel experimental design for measurement of task-evoked changes in regional blood oxygenation and glucose metabolism with high temporal resolution. Over a 90-min simultaneous BOLD-fMRI/FDG-PET scan, [18F] FDG was constantly infused to 10 healthy volunteers, who viewed a flickering checkerboard presented in a hierarchical block design. Dynamic task-related changes in blood oxygenation and glucose metabolism were examined with temporal resolution of 2.5sec and 1-min, respectively. Task-related, temporally coherent brain networks of haemodynamic and metabolic connectivity were maximally related in the visual cortex, as expected. Results demonstrate that the hierarchical block design, together with the infusion FDG-PET technique, enabled both modalities to track task-related neural responses with high temporal resolution. The simultaneous MR-PET approach has the potential to provide unique insights into the dynamic haemodynamic and metabolic interactions that underlie cognition in health and disease.

Head motion is a major source of image artefacts in neuroimaging studies and can lead to degradation of the quantitative accuracy of reconstructed PET images. Simultaneous Magnetic Resonance-Positron Emission Tomography (MR-PET) makes it possible to estimate head motion information from high-resolution MR images and then correct motion artefacts in PET images. In this paper, we introduce a fully automated PET motion correction method, MR-guided MAF, based on the co-registration of multi-contrast MR images. The performance of the MR-guided MAF method was evaluated using MR-PET data acquired from a cohort of ten healthy participants who received a slow infusion of fluorodeoxyglucose ([18-F]FDG). Compared with conventional methods, MR guided PET image reconstruction can reduce head motion introduced artefacts and improve the image sharpness and quantitative accuracy of PET images acquired using simultaneous MR-PET scanners. The fully automated motion estimation method has been implemented as a publicly available web-service.

Objective To investigate the clinical effectiveness of intestinal cleaning knowledge training on intestinal cleaning rate of inpatients undergoing colonoscopy. Methods A comparative study was conducted by selection of 1166 inpatients undergoing colonoscopy in our hospital from January to November in 2016. Among them, 510 inpatients from January to June 2016 were induced into the pre-training group and 656 inpatients from July to December 2016 were induced into in the post-training group. The training contents included the indications and contraindications of colonoscopy, the dietary guidance before colonoscopy, the usage of laxatives for colonoscopy, the standards of intestinal cleanliness, and the precautions before and after colonoscopy. The effective intestinal cleansing rates of the two groups were observed and recorded. Results The average score of intestinal cleaning knowledge assessment of nurses increased from 83.5 before the training to 93.1 after the training; the effective intestinal cleansing rate from July to September of the same year was continuously improved by the training combined rectification, increasing from 73.1% to 92.3%, which was above the target value of 91.2%. Conclusion Intestinal cleaning knowledge training plays a significant role in improvement of the rate of effective intestinal cleaning, which can significantly reduce the rate of missed diagnosis and misdiagnosis.

Objective: To investigate the effect of shared decision-making on compliance with oral nutrition supplementation (ONS) and quality of life of among elderly patients (≥ 60 years) with cirrhosis. Methods: This study used an action research design that tested the intervention of shared decision-making among elderly patients with cirrhosis. It measured how this intervention influenced compliance with oral nutrition supplements and quality of life. Sixty-eight elderly patients with cirrhosis who were hospitalized in our department from May 2019 to November 2019 were randomly divided into two groups: the control group the observation group. The total number of participants in each group was 34. The control group received routine care. The observation group received shared decision-making nutrition care in addition to routine care. Results: After the intervention, patients in the observation group had higher oral nutritional compliance and quality of life scores than the control group and had a statistical significance (P < 0.05). Conclusion: After the implementation of Shared decision nutrition care for 34 elderly patients with cirrhosis, the patients' ONS compliance and quality of life were greatly improved. At present, there are relatively few researches on the application of Shared decision in digestive system diseases，Hope to play a certain role of reference or guidance.

Gliomas are the most common primary brain malignancies, with different degrees of aggressiveness, variable prognosis and various heterogeneous histologic sub-regions, i.e., peritumoral edematous/invaded tissue, necrotic core, active and non-enhancing core. This intrinsic heterogeneity is also portrayed in their radio-phenotype, as their sub-regions are depicted by varying intensity profiles disseminated across multi-parametric magnetic resonance imaging (mpMRI) scans, reflecting varying biological properties. Their heterogeneous shape, extent, and location are some of the factors that make these tumors difficult to resect, and in some cases inoperable. The amount of resected tumor is a factor also considered in longitudinal scans, when evaluating the apparent tumor for potential diagnosis of progression. Furthermore, there is mounting evidence that accurate segmentation of the various tumor sub-regions can offer the basis for quantitative image analysis towards prediction of patient overall survival. This study assesses the state-of-the-art machine learning (ML) methods used for brain tumor image analysis in mpMRI scans, during the last seven instances of the International Brain Tumor Segmentation (BraTS) challenge, i.e., 2012-2018. Specifically, we focus on i) evaluating segmentations of the various glioma sub-regions in pre-operative mpMRI scans, ii) assessing potential tumor progression by virtue of longitudinal growth of tumor sub-regions, beyond use of the RECIST/RANO criteria, and iii) predicting the overall survival from pre-operative mpMRI scans of patients that underwent gross total resection. Finally, we investigate the challenge of identifying the best ML algorithms for each of these tasks, considering that apart from being diverse on each instance of the challenge, the multi-institutional mpMRI BraTS dataset has also been a continuously evolving/growing dataset., Comment: The International Multimodal Brain Tumor Segmentation (BraTS) Challenge

Purpose: The suppression of motion artefacts from MR images is a challenging task. The purpose of this paper is to develop a standalone novel technique to suppress motion artefacts from MR images using a data-driven deep learning approach. Methods: A deep learning convolutional neural network (CNN) was developed to remove motion artefacts in brain MR images. A CNN was trained on simulated motion corrupted images to identify and suppress artefacts due to the motion. The network was an encoder-decoder CNN architecture where the encoder decomposed the motion corrupted images into a set of feature maps. The feature maps were then combined by the decoder network to generate a motion-corrected image. The network was tested on an unseen simulated dataset and an experimental, motion corrupted in vivo brain dataset. Results: The trained network was able to suppress the motion artefacts in the simulated motion corrupted images, and the mean percentage error in the motion corrected images was 2.69 % with a standard deviation of 0.95 %. The network was able to effectively suppress the motion artefacts from the experimental dataset, demonstrating the generalisation capability of the trained network. Conclusion: A novel and generic motion correction technique has been developed that can suppress motion artefacts from motion corrupted MR images. The proposed technique is a standalone post-processing method that does not interfere with data acquisition or reconstruction parameters, thus making it suitable for a multitude of MR sequences.

The capacitive deionization (CDI) performance of activated carbon electrodes with various anions and competitive electrosorption between two anions in mixed solution were studied. It was found that the electrosorption capacity was strongly dependent on the ionic charge and hydrated radius. With the same concentration, smaller monovalent anions showed their size-affinity to access the pores. The order of normalized equivalent capacity was as follows: trivalent anion > divalent anion > monovalent anion. In mixed Cl− and NO3− solution, Cl− is preferably electrosorped over NO3− in early period of the electrosorption process. However, during the end of the electrosorption process, part of Cl− would be exchanged by NO3− due to the stronger competitiveness of NO3− than Cl−. There was no obvious selectivity between Cl− and SO42 −.

The effects of ageing and incorporation of ion-exchange membrane on the electrosorption performance of activated carbon (AC) electrodes-based electrosorption modules were investigated. It was found that the surface area of aged electrode decreased while the average pore size increased. In addition, more surface functional groups, especially –COOH groups were created on the electrode surface due to ageing. All these lead to a decrease in the electrosorption capacity and electrosorption rate. Incorporation of ion-exchange membranes can improve the desalination performance of the electrosorption module, especially for the aged electrode.

In this study, effects of ionic radius, charge, and mass on the electro sorption desalination performance of activated carbon electrodes were investigated. The sorption capacity of carbon electrodes was approximately 0.08-0.18 mmol/g carbon. The electro sorption capacity of monovalent ions was higher than multivalent ions. Covering area of cation on electrode is smaller than anion, when considering Covering area of hydrated ions, the cation and anion is almost equal. Because of the relatively small pore diameter (2 nm) of the carbon electrode, only 15.62 to 37.90 m2/g carbon surface area was available to ions eletro sorption.

The state controllability for generalised dynamical systems with constrained control is discussed in this paper. The main results of the paper are the following:(1) a necessary and sufficient condition of the state controllability in the sense of control energy or amplitude constrained for generalised dynamical systems is obtained;(2) a control function u(t) is constructed such thata) u(t) satisfies constrained energy or amplitude condition,b) the state driven by u(t) moves from an arbitrary x(0−) = x0 to x(T(x0)) = 0,c) the trajectory driven by u(t) has no impulsive behaviour within (0, T(x0)].