In this paper, we propose perceptual adversarial networks (PANs) for image-to-image transformations. Different from existing application driven algorithms, PAN provides a generic framework of ...learning to map from input images to desired images (Fig. 1), such as a rainy image to its de-rained counterpart, object edges to photos, and semantic labels to a scenes image. The proposed PAN consists of two feed-forward convolutional neural networks: the image transformation network T and the discriminative network D. Besides the generative adversarial loss widely used in GANs, we propose the perceptual adversarial loss, which undergoes an adversarial training process between the image transformation network T and the hidden layers of the discriminative network D. The hidden layers and the output of the discriminative network D are upgraded to constantly and automatically discover the discrepancy between the transformed image and the corresponding ground truth, while the image transformation network T is trained to minimize the discrepancy explored by the discriminative network D. Through integrating the generative adversarial loss and the perceptual adversarial loss, D and T can be trained alternately to solve image-to-image transformation tasks. Experiments evaluated on several image-to-image transformation tasks (e.g., image deraining and image inpainting) demonstrate the effectiveness of the proposed PAN and its advantages over many existing works.
In the past few years, intelligent structural damage identification algorithms based on machine learning techniques have been developed and obtained considerable attentions worldwide, due to the ...advantages of reliable analysis and high efficiency. However, the performances of existing machine learning–based damage identification methods are heavily dependent on the selected signatures from raw signals. This will cause the fact that the damage identification method, which is the optimal solution for a specific application, may fail to provide the similar performance on other cases. Besides, the feature extraction is a time-consuming task, which may affect the real-time performance in practical applications. To address these problems, this article proposes a novel method based on deep convolutional neural networks to identify and localise damages of building structures equipped with smart control devices. The proposed deep convolutional neural network is capable of automatically extracting high-level features from raw signals or low-level features and optimally selecting the combination of extracted features via a multi-layer fusion to satisfy any damage identification objective. To evaluate the performance of the proposed deep convolutional neural network method, a five-level benchmark building equipped with adaptive smart isolators subjected to the seismic loading is investigated. The result shows that the proposed method has outstanding generalisation capacity and higher identification accuracy than other commonly used machine learning methods. Accordingly, it is deemed as an ideal and effective method for damage identification of smart structures.
•A novel multi-modal multi-instance knowledge distillation scheme.•Improve the performance of MRI-based Alzheimer’s disease prediction.•Use multi-modal data for model training and only MRI data for ...testing.•Generate heat-maps on the basis of image patches for visual interpretation.
Mild cognitive impairment (MCI) conversion prediction, i.e., identifying MCI patients of high risks converting to Alzheimer’s disease (AD), is essential for preventing or slowing the progression of AD. Although previous studies have shown that the fusion of multi-modal data can effectively improve the prediction accuracy, their applications are largely restricted by the limited availability or high cost of multi-modal data. Building an effective prediction model using only magnetic resonance imaging (MRI) remains a challenging research topic. In this work, we propose a multi-modal multi-instance distillation scheme, which aims to distill the knowledge learned from multi-modal data to an MRI-based network for MCI conversion prediction. In contrast to existing distillation algorithms, the proposed multi-instance probabilities demonstrate a superior capability of representing the complicated atrophy distributions, and can guide the MRI-based network to better explore the input MRI. To our best knowledge, this is the first study that attempts to improve an MRI-based prediction model by leveraging extra supervision distilled from multi-modal information. Experiments demonstrate the advantage of our framework, suggesting its potentials in the data-limited clinical settings.
The combined effects of different severities of Wooden Breast (WB), White Striping (WS), and Spaghetti Meat (SM) were examined in 300 chicken breast fillets from 10 flocks. Severity (0 = absent, ...1 = mild, noticeable upon close inspection, 2 = severe), noticeably altered from normal breast fillet (NB). Results showed that any combination of myopathies and severity resulted in significantly elevated compression force, pH and peak counts measured by the shear force test. With the exception of mild WB + mild WS, all combinations resulted in significantly higher drip loss, cooking loss and lightness value. Overall, the quality of fillets was affected the least by WS, while negatively affected the most by SM. There were limited effects on fillet quality from mild WB but major deleterious effects from severe WB.
Polyvinylidene fluoride (PVDF) has been widely studied and applied in separation membranes due to its high thermal and chemical stability and mechanical strength. However, PVDF has strong ...hydrophobicity, resulting in easy contamination of the membrane surface and fast flux attenuation, so it is necessary to modify the membrane surface to improve its separation selectivity and service life. In this paper, PVDF microporous membrane was used as the matrix material and graphene oxide (GO) as the separation layer material. The GO/Mn3O4/PVDF composite membrane was prepared by layer self-assembly of GO nanosheets, and the functional layer spacing was adjusted by nanometer Mn3O4 intercalation. The prepared composite membrane showed high flux and separation selectivity in the filtration of organic compounds. The results showed that the rejection of methylene blue increased from 34% to 99.5%, and the flux decreased from 3000 L m−2 h−1 to 95 L m−2 h−1 when GO nanosheets covered the PVDF supporting membrane. After the introduction of Mn3O4 nanowires in the GO interlayer, the dye rejection reached 99.9% and the flux reached 612 L m−2 h−1. Compared with the unintercalated composite membranes, the flux of the prepared composite membranes showed good stability in the treatment of methylene blue, and the rejection remained unchanged.
Structural magnetic resonance imaging (sMRI) can capture the spatial patterns of brain atrophy in Alzheimer's disease (AD) and incipient dementia. Recently, many sMRI‐based deep learning methods have ...been developed for AD diagnosis. Some of these methods utilize neural networks to extract high‐level representations on the basis of handcrafted features, while others attempt to learn useful features from brain regions proposed by a separate module. However, these methods require considerable manual engineering. Their stepwise training procedures would introduce cascading errors. Here, we propose the parallel attention‐augmented bilinear network, a novel deep learning framework for AD diagnosis. Based on a 3D convolutional neural network, the framework directly learns both global and local features from sMRI scans without any prior knowledge. The framework is lightweight and suitable for end‐to‐end training. We evaluate the framework on two public datasets (ADNI‐1 and ADNI‐2) containing 1,340 subjects. On both the AD classification and mild cognitive impairment conversion prediction tasks, our framework achieves competitive results. Furthermore, we generate heat maps that highlight discriminative areas for visual interpretation. Experiments demonstrate the effectiveness of the proposed framework when medical priors are unavailable or the computing resources are limited. The proposed framework is general for 3D medical image analysis with both efficiency and interpretability.
A lightweight and effective neural network for early Alzheimer's disease (AD) diagnosis. An integrated framework of feature extraction, classification, and localization. Use an asymmetrically parallel structure to extract better representations. Process whole‐brain structural MRI scans without the need of any prior knowledge.
Spaghetti meat (SM), woody breast (WB), and white striping (WS) are myopathies affecting breast muscle of broiler chickens, and are characterized by a loss of myofibers and an increase in fibrous ...tissue. The conditions develop in intensive broiler chicken production systems, and cause poor meat process-ability and negative customer perception leading to monetary losses. The objectives of the present study were to describe the physical and histological characteristics of breast myopathies from commercial broiler chicken flocks in Ontario, Canada, and to assess the associations between the severity of myopathies with the physical and histological characteristics of the affected breast muscle fillets. Chicken breast fillets (n = 179) were collected over 3 visits from a processing plant and scored macroscopically to assess the severity of myopathies, following an established scoring scheme. For each fillet, the surface area, length, width, thickness, weight, and hardness (compression force) were measured. A subset of 60 fillets was evaluated microscopically. Multinomial logistic regression models were built to evaluate associations between physical parameters and macroscopic scores. The odds of SM co-occurring with severe WB (SM1WB2) were significantly associated with increased fillet thickness (OR = 1.59, 95% CI 1.31–1.94) and weight (OR = 1.06, 95% CI 1.03–1.09). Histologically, myopathies had overlapping lesions consisting of polyphasic myodegeneration, perivascular inflammatory cuffing and accumulation of fibrous tissue and fat. The pairwise correlation between macroscopic and microscopic scores was moderate (rho 0.45, P < 0.001). This is the first study to characterize breast myopathies in Canadian broiler flocks. Results show that the morphologic and microscopic changes of fillets from this cohort are similar to data from other countries, and provide database to benchmark these parameters in future studies. Our standardized categorization can be applied to broiler breast fillets in other regions of the world.
Intertidal groundwater and seawater were sampled to analyze the distribution characteristics, the contamination status and the submarine groundwater discharge (SGD)-associated fluxes of heavy metals ...Cu, Pb, Zn, Cd, Cr, and Hg as well as the metalloid As at four typical intertidal wetlands (including a sandy beach, a mud flat, a tidal marsh and an estuarine intertidal zone) of Jiaozhou Bay, China. Results show that the surface water near the Dagu River estuary suffers from a severe Cu pollution. The groundwater in the sandy beach and mud flat has stronger enrichment abilities of heavy metals than those at the other two sites. The contents of Pb and Zn in groundwater are mainly controlled by the sulfate reduction. At the mud flat, human activities may cause potential Pb contamination to groundwater. The heavy metal effluxes in the sandy beach are the largest of all the four wetlands.
•Heavy metals in four typical intertidal wetlands of Jiaozhou Bay were investigated.•Surface water near the Dagu River estuary suffers from a severe Cu pollution.•Groundwater at the sandy beach and mud flat has strong metal-enrichment abilities.•SGD-associated metal fluxes at the sandy beach are the largest in the four wetlands.
Coastal dynamic forces, such as waves and tides, altogether drive groundwater circulation and salt transport in the intertidal zone. However, few numerical studies have ever considered the combined ...effects of waves and tides. In this study, the fluctuations of wave height are integrated with tidal level together using an iterative least squares fitting method, in which the wave height can be acquired from measured sea level in the surf zone, and this fitted wave height is further verified by wind speed. Groundwater flow and salt transport were then simulated using the MARUN code to evaluate the impacts after considering wave effect. The simulated equivalent freshwater head and salinity of the model with wave effect presented less difference with the measured data compared with the simulated results of the model without wave effect. After incorporating the wave effect in a model, submarine groundwater discharge (SGD) was increased, among which recirculated SGD grew more rapidly than the fresh, leading to the proportion of the fresh SGD accounting for a small proportion (1.1%). The water influx and efflux rates increased greatly especially during the period of high wave height. Most of the influx occurred in the intertidal zone, while a considerable amount of efflux occurred in the subtidal zone. The iterative algorithm to separate the wave height from the mixed field data can be employed to identify and quantify the respective effects of tides and the combined effects of waves and tides on the density‐dependent beach groundwater flow.
Key Points
A new iterative algorithm was proposed to quantify tidal level and wave height using hourly measured surf zone sea level data
Beach groundwater flow and salinity distribution were numerically reproduced using wave height and setup as boundary conditions
Waves may significantly change groundwater flow and salinity distribution in a beach aquifer
Spaghetti meat (SM), woody breast (WB), and white striping (WS) are myopathies that affect the pectoral muscle of fast-growing broiler chickens. The prevalence and possible risk factors of these ...myopathies have been reported in other countries, but not yet in Canada. Thus, the objective of this study was to assess the prevalence and risk factors associated with these myopathies in a representative population of Canadian broilers. From May 2019 to March 2020, 250 random breast fillets from each of 37 flocks (total, 9,250) were obtained from two processing plants and assessed for the presence and severity of myopathies. Demographic data (e.g., sex and average live weight), environmental conditions during the grow-out period (e.g., temperature), and husbandry parameters (e.g., vaccination) were collected for each flock. Associations between these factors and the myopathies were tested using logistic regression analyses. The prevalence of SM, severe WB, and mild or moderate WS was 36.3% (95% CI: 35.3-37.3), 11.8% (95% CI: 11.2-12.5), and 96.0% (95% CI: 95.6-96.4), respectively. Most (85.1%) of the fillets showed multiple myopathies. Regression analyses showed that the odds of SM increased with live weight (OR = 1.30, 95% CI 1.01-1.69) and higher environmental temperature during the grow-out period (OR = 1.75, 95% CI 1.31-2.34). The odds of WB increased with live weight (OR = 1.23, 95% CI 1.03-1.47) and when flocks were not vaccinated against coccidia (OR = 1.86, 95% CI 1.51-2.29). This study documents for the first time a high prevalence of myopathies in Ontario broilers, and suggests that these lesions may have a significant economic impact on the Canadian poultry industry. Our results indicate that environmental conditions and husbandry are associated with the development of breast myopathies, in agreement with the current literature. Future studies are needed to determine how risk factors can promote the occurrence of these conditions, in order to implement possible mitigating strategies.