Converter Capacitor Temperature Estimation Based on Continued Training LSTM under Variable Load Conditions.

Capacitors are crucial components in power electronic converters, responsible for harmonic elimination, energy buffering, and voltage stabilization. However, they are also the most susceptible to damage due to their operational environment. Accurate temperature estimation of capacitors is essential for monitoring their condition and ensuring the reliability of the converter system. This paper presents a novel method for estimating the core temperature of capacitors using a long short-term memory (LSTM) algorithm. The approach incorporates a continued training mechanism to adapt to variable load conditions in converters. Experimental results demonstrate the proposed method's high accuracy and robustness, making it suitable for real-time capacitor temperature monitoring in practical applications.

Plasma Deoxycholic Acid Levels are Associated with Hemodynamic and Clinical Outcomes in Acute Pulmonary Embolism Patients.

This study aimed to evaluate the correlation of plasma deoxycholic acid (DCA) levels with clinical and hemodynamic parameters in acute pulmonary embolism (APE) patients. Total 149 APE adult patients were prospectively recruited. Plasma DCA levels were measured using rapid resolution liquid chromatography-quadrupole time-of-flight mass spectrometry. Baseline clinical and hemodynamic parameters were evaluated according to plasma DCA levels. The plasma DCA levels were significantly lower in APE patients than in those without APE (P < 0.001). APE patients with adverse events had lower plasma DCA levels (P < 0.001). Low DCA group patients presented more adverse cardiac function, higher NT-proBNP levels (P = 0.010), and higher WHO functional class levels (P = 0.023). Low DCA group also presented with an adverse hemodynamic status, with higher pulmonary vascular resistance levels (P = 0.027) and lower cardiac index levels (P = 0.024). Both cardiac function and hemodynamic parameters correlated well with plasma DCA levels. Kaplan-Meier survival analysis demonstrated that APE patients with lower plasma DCA levels had a significantly higher event rate (P = 0.009). In the univariate and multivariate Cox regression analyses, the plasma DCA level was an independent predictor of clinical worsening events after adjusting for age, sex, WHO functional class, NT-proBNP level, pulmonary vascular resistance, and cardiac index (HR 0.370, 95% CI 0.161, 0.852; P = 0.019). Low plasma DCA levels predicted adverse cardiac function and hemodynamic collapse. A low DCA level was correlated with a higher clinical worsening event rate and could be an independent predictor of clinical outcomes in multivariate analysis.

Dietary Bacillus velezensis KNF-209 supplementation improves growth performance, enhances immunity, and promotes gut health in broilers.

This study aimed to investigate the effects of dietary Bacillus velezensis KNF-209 (BV-KNF-209) on the growth performance, immunity, and gut health of broilers. A total of 540 one-day-old male Cobb-500 broilers were randomly divided into 5 groups of 6 replicates with 18 broilers per replicate. Dietary treatments were corn-soybean meal basal diets supplemented with 0, 50, 100, 200, and 400 mg/kg BV-KNF-209 (CON, BV 50, BV 100, BV 200, and BV 400 groups, respectively) for 42 d. Compared with the CON group, the average daily gains (ADG) at 0 to 42 d in the BV 100 and BV 200 groups were significantly increased (P < 0.01), and the feed-to-gain (F:G) ratios were significantly decreased at 0 to 21 d (P < 0.01) and 0 to 42 d (P < 0.05). The BV 200 and BV 400 groups had higher serum immunoglobulin M (IgM) levels at d 21 and 42 (P < 0.05). The serum levels of tumor necrosis factor-α (TNF-α), interleukin-1ß (IL-1ß), and interleukin-6 (IL-6) were significantly decreased in the BV 50, BV 100, and BV 200 groups at d 21 (P < 0.05), and serum IL-1ß and IL-6 levels were also reduced in the BV 100 and BV 200 groups at d 42 (P < 0.05). Meanwhile, increased interleukin-10 (IL-10) levels in the jejunal and ileal mucosa at d 42 were observed in the BV 100, BV 200, and BV 400 groups (P < 0.05), while the IL-1ß and IL-6 levels (P < 0.01) were decreased. The BV 200 and BV 400 groups showed significantly higher activities of lipase and trypsin (P < 0.05) in jejunal digesta as well as higher activities of amylase and trypsin (P < 0.01) in ileal digesta at d 42. The cecal acetic acid and propionic acid levels in the BV groups and lactic acid levels in the BV 50, BV 100, and BV 200 groups (P < 0.05) were significantly higher compared to those in the CON group. Overall, dietary BV-KNF-209 supplementation significantly improved broiler growth performance, an effect that may have been achieved by heightening immunity, increasing digestive enzyme activity, and raising intestinal short-chain fatty acids and lactic acid levels.

Biocontrol agents modulate phyllosphere microbiota interactions against pathogen Pseudomonas syringae.

The pathogen Pseudomonas syringae, responsible for a variety of diseases, poses a considerable threat to global crop yields. Emerging biocontrol strategies employ antagonistic microorganisms, utilizing phyllosphere microecology and systemic resistance to combat this disease. However, the interactions between phyllosphere microbial dynamics and the activation of the plant defense system remain poorly understood. Here we show significant alterations in phyllosphere microbiota structure and plant gene expression following the application of biocontrol agents. We reveal enhanced collaboration and integration of Sphingomonas and Methylobacterium within the microbial co-occurrence network. Notably, Sphingomonas inhibits P. syringae by disrupting pathogen chemotaxis and virulence. Additionally, both Sphingomonas and Methylobacterium activate plant defenses by upregulating pathogenesis-related gene expression through abscisic acid, ethylene, jasmonate acid, and salicylic acid signaling pathways. Our results highlighted that biocontrol agents promote plant health, from reconstructing beneficial microbial consortia to enhancing plant immunity. The findings enrich our comprehension of the synergistic interplays between phyllosphere microbiota and plant immunity, offering potential enhancements in biocontrol efficacy for crop protection.

Insights from the neural guidance factor Netrin-1 into neurodegeneration and other diseases.

Netrin-1 was initially discovered as a neuronal growth cue for axonal guidance, and its functions have later been identified in inflammation, tumorigenesis, neurodegeneration, and other disorders. We have recently found its alterations in the brains with Alzheimer's disease, which might provide important clues to the mechanisms of some unique pathologies. To provide better understanding of this promising molecule, we here summarize research progresses in genetics, pathology, biochemistry, cell biology and other studies of Netrin-1 about its mechanistic roles and biomarker potentials with an emphasis on clinical neurodegenerative disorders in order to expand understanding of this promising molecular player in human diseases.

Predicting Antibiotic Tolerance in hvKP and cKP Respiratory Infections Through Biofilm Formation Analysis and Its Resistance Implications.

Introduction: Respiratory infections are a major global health concern, with Klebsiella pneumoniae standing out due to its evolving antibiotic resistance. This study compares the resistance profiles of hypervirulent Klebsiella pneumoniae (hvKP) and classical Klebsiella pneumoniae (cKP), aiming to shed light on their clinical implications. Methods: We analyzed 86 cases, comprising 42 hvKP and 44 cKP strains, using comprehensive antimicrobial susceptibility testing and clinical data evaluation to assess antibiotic tolerance and resistance mechanisms. Results: Our findings reveal distinct resistance patterns between hvKP and cKP, highlighting the role of chromosomal mutations and plasmid-mediated gene transfer in conferring antibiotic resistance. Notably, hvKP strains exhibited unique resistance trends, including the production of extended-spectrum ß-lactamases (ESBLs) and carbapenemases, differing from those of cKP. Discussion: This research underscores the importance of continuous surveillance and the development of targeted therapies against antibiotic-resistant Klebsiella pneumoniae. It emphasizes the critical need for judicious antibiotic use and novel therapeutic approaches to combat respiratory infections caused by these increasingly resistant pathogens.

Application of pectin hydrolyzing bacteria in tobacco to improve flue-cured tobacco quality.

To study the relationship between the diversity of the surface microbial community and tobacco flavor, and to improve tobacco quality using microorganisms. The microbial community composition and diversity of 14 samples of flue-cured tobacco from tobacco-producing areas in Yunnan with varying grades were analyzed by high-throughput sequencing. PICRUSt was used for predicting microbial functions. A strain of Bacillus amyloliquefaciens W6-2 with the ability to degrade pectin was screened from the surface of flued-cured tobacco leaves from Yunnan reroasted tobacco leave. The enzyme preparation was prepared through fermentation and then applied for treating flue-cured tobacco. The improvement effect was evaluated by measuring the content of macromolecule and the changes in volatile components, combined with sensory evaluations. The bacterial communities on the surface of flue-cured tobacco exhibited functional diversity, consisting primarily of Variovorax, Pseudomonas, Sphingomonas, Burkholderia, and Bacillus. These bacterial strains played a role in the aging process of flue-cured tobacco leaves by participating in amino acid metabolism and carbohydrate metabolism. These metabolic activity converted complex macromolecules into smaller molecular compounds, ultimately influence the smoking quality and burning characteristics of flue-cured tobacco. The pectinase preparation produced through fermentation using W6-2 has been found to enhance the aroma and sweetness of flue-cured tobacco, leading to improved aroma, reduced impurities, and enhanced smoothness. Additionally, the levels of pectin, cellulose, and hemicellulose decreased, while the levels of water-soluble sugar and reducing sugar increased, and the contents of esters, ketones, and aldehydes increased, and the contents of benzoic acid decreased. The study revealed the correlation between surface microorganisms and volatile components of Yunnan tobacco leaves, and the enzyme produced by the pectin-degrading bacteria W6-2 effectively improved the quality of flue-cured tobacco.

What role do biocontrol agents with Mg2+ play in the fate of antibiotic resistome and pathogenic bacteria in the phyllosphere?

The contamination of the plant phyllosphere with antibiotics and antibiotic resistance genes (ARGs), caused by application of antibiotics, is a significant environmental issue in agricultural management. Alternatively, biocontrol agents are environmentally friendly and have attracted a lot of interest. However, the influence of biocontrol agents on the phyllosphere resistome remains unknown. In this study, we applied biocontrol agents to control the wildfire disease in the Solanaceae crops and investigated their effects on the resistome and the pathogen in the phyllosphere by using metagenomics. A total of 250 ARGs were detected from 15 samples, which showed a variation in distribution across treatments of biocontrol agents (BA), BA with Mg2+ (T1), BA with Mn2+ (T2), and kasugamycin (T3) and nontreated (CK). The results showed that the abundance of ARGs under the treatment of BA-Mg2+ was lower than that in the CK group. The abundance of cphA3 (carbapenem resistance), PME-1 (carbapenem resistance), tcr3 (tetracycline antibiotic resistance), and AAC (3)-VIIIa (aminoglycoside antibiotic resistance) in BA-Mg2+ was significantly higher than that in BA-Mn2+ (P < 0.05). The abundance of cphA3, PME_1, and tcr3 was significantly negatively related to the abundance of the phyllosphere pathogen Pseudomonas syringae (P < 0.05). We also found that the upstream and downstream regions of cphA3 were relatively conserved, in which rpl, rpm, and rps gene families were identified in most sequences (92%). The Ka/Ks of cphA3 was 0 in all observed sequences, indicating that under the action of purifying selection, nonsynonymous substitutions are often gradually eliminated in the population. Overall, this study clarifies the effect of biocontrol agents with Mg2+ on the distribution of the phyllosphere resistome and provides evolutionary insights into the biocontrol process. IMPORTANCE: Our study applied metagenomics analysis to examine the impact of biocontrol agents (BAs) on the phyllosphere resistome and the pathogen. Irregular use of antibiotics has led to the escalating dissemination of antibiotic resistance genes (ARGs) in the environment. The majority of BA research has focused on the effect of monospecies on the plant disease control process, the role of the compound BA with nutrition elements in the phyllosphere disease, and the resistome is still unknown. We believe BAs are eco-friendly alternatives for antibiotics to combat the transfer of ARGs. Our results revealed that BA-Mg2+ had a lower relative abundance of ARGs compared to the CK group, and the phyllosphere pathogen Pseudomonas syringae was negatively related to three specific ARGs, cphA3, PME-1, and tcr3. These three genes also present different Ka/Ks. We believe that the identification of the distribution and evolution modes of ARGs further elucidates the ecological role and facilitates the development of BAs, which will attract general interest in this field.

Metabolomics-based study of chemical compositions in cellulase additives derived from a tobacco-origin Bacillus subtilis and their impact on tobacco sensory attributes.

To enhance the quality of tobacco leaves and optimize the smoking experience, diverse strains of functional bacteria and their associated metabolites have been used in tobacco aging. Exogenous cellulase additives are frequently employed to facilitate the degradation of cellulose and other macromolecular matrices and enhance the quality of the tobacco product. However, little is known about how microbial metabolites present in exogenous enzyme additives affect tobacco quality. In this study, crude cellulase solutions, produced by a tobacco-originating bacterium Bacillus subtilis FX-1 were employed on flue-cured tobacco. The incorporation of cellulase solutions resulted in the reduction of cellulose crystallinity in tobacco and the enhancement of the overall sensory quality of tobacco. Notably, tobacco treated with cellulase obtained from laboratory flask fermentation demonstrated superior scent and flavor attributes in comparison to tobacco treated with enzymes derived from industrial bioreactor fermentation. The targeted and untargeted metabolomic analysis revealed the presence of diverse flavor-related precursors and components in the cellulase additives, encompassing sugars, alcohols, amino acids, organic acids, and others. The majority of these metabolites exhibited significantly higher levels in the flask group compared to the bioreactor group, probably contributing to a pronounced enhancement in the sensory quality of tobacco. Our findings suggest that the utilization of metabolic products derived from B. subtilis FX-1 as additives in flue-cured tobacco holds promise as a viable approach for enhancing sensory attributes, establishing a solid theoretical foundation for the potential development of innovative tobacco aging additives.

Complete genome sequence analysis of Bacillus velezensis A5, a promising biocontrol agent from the Pacific Ocean.

Tobacco bacterial wilt (TBW) caused by Ralstonia solanacearum is a serious soil-borne disease, which seriously damages the growth of tobacco crops. Bacillus velezensis A5 was isolated from 3000 m deep-sea sediments of the Pacific Ocean, and was found to be antagonistic to TBW. Here, we report the complete genome sequence of strain A5, which has a 4,000,699-bp single circular chromosome with 3827 genes and a G + C content of 46.44%, 87 tRNAs, and 27 rRNAs. A total of 12 gene clusters were identified in the genome of strain A5, which were responsible for the biosynthesis of antibacterial compounds, including surfactin, bacillaene, fengycin, difficidin, bacillibactin, and bacilysin. Additionally, strain A5 was found to contain a series of genes related to the biosynthesis of carbohydrate-active enzymes and secreted proteins. Our results indicate that strain A5 can be considered a promising biocontrol agent against TBW in agricultural fields.

FedEYE: A scalable and flexible end-to-end federated learning platform for ophthalmology.

Data-driven machine learning, as a promising approach, possesses the capability to build high-quality, exact, and robust models from ophthalmic medical data. Ophthalmic medical data, however, presently exist across disparate data silos with privacy limitations, making centralized training challenging. While ophthalmologists may not specialize in machine learning and artificial intelligence (AI), considerable impediments arise in the associated realm of research. To address these issues, we design and develop FedEYE, a scalable and flexible end-to-end ophthalmic federated learning platform. During FedEYE design, we adhere to four fundamental design principles, ensuring that ophthalmologists can effortlessly create independent and federated AI research tasks. Benefiting from the design principles and architecture of FedEYE, it encloses numerous key features, including rich and customizable capabilities, separation of concerns, scalability, and flexible deployment. We also validated the applicability of FedEYE by employing several prevalent neural networks on ophthalmic disease image classification tasks.

Intraoperative blood pressure and cardiac complications after aneurysmal subarachnoid hemorrhage: a retrospective cohort study.

BACKGROUND: Previous studies report that intraoperative hypotension worsens outcomes after aneurysmal subarachnoid hemorrhage (aSAH). However, the hypotensive harm threshold for major adverse cardiovascular events (MACE) remains unclear. METHODS: The authors included aSAH patients who had general anesthesia for aneurysmal clipping/coiling. MACE were defined by a composite of acute myocardial injury, acute myocardial infarction, and other cardiovascular complications identified by electrocardiogram and echocardiography. The authors initially used logistic regression and change-point analysis based on the second derivative to identify mean arterial pressure (MAP) of 75 mmHg as the harm threshold. Thereafter, our major exposure was MAP below 75 mmHg characterized by area, duration, and time-weighted average. The area below 75 mmHg represents the severity and duration of exposure and was defined as the sum of all areas below a specified threshold using the trapezoid rule. Time-weighted average MAP was derived by dividing area below the threshold by the duration of anesthesia. All analyses were adjusted for baseline risk factors including age greater than 70 years, female sex, severity of intracerebral hemorrhage, history of cardiovascular disease, and preoperative elevated myocardial enzymes. RESULTS: Among 1029 patients enrolled, 254 (25%) developed postoperative MACE. Patients who experienced MACE were slightly older (59±11 vs. 54±11 years), were slightly more often women (69 vs. 58%), and had a higher prevalence of cardiovascular history (65 vs. 47%). Adjusted cardiovascular risk increased nearly linearly over the entire range of observed MAP. However, there was a slight inflexion at MAP of 75 mmHg. MACE was significantly associated with area [adjusted odds ratios (aOR) 1.004 per 10 mmHg.min, 95% CI: 1.001-1.007, P =0.002), duration (aOR 1.031 per 10 min, 95% CI: 1.009-1.054, P =0.006), and time-weighted average (aOR 3.516 per 10 mmHg, 95% CI: 1.818-6.801, P <0.001) of MAP less than 75 mmHg. CONCLUSIONS: Lower blood pressures were associated with cardiovascular complications over the entire observed range, but worsened when MAP was less than 75 mmHg. Pending trial data to establish causality, it may be prudent to keep MAP above 75 mmHg in patients having surgical aSAH repairs to reduce the risk of MACE.

CYP8B1 downregulation mediates the metabolic effects of vertical sleeve gastrectomy in mice.

BACKGROUND AND AIMS: Although the benefits of vertical sleeve gastrectomy (VSG) surgery are well known, the molecular mechanisms by which VSG alleviates obesity and its complications remain unclear. We aim to determine the role of CYP8B1 (cytochrome P450, family 8, subfamily B, polypeptide 1) in mediating the metabolic benefits of VSG. APPROACH AND RESULTS: We found that expression of CYP8B1, a key enzyme in controlling the 12α-hydroxylated (12α-OH) bile acid (BA) to non-12α-OH BA ratio, was strongly downregulated after VSG. Using genetic mouse models of CYP8B1 overexpression, knockdown, and knockout, we demonstrated that overexpression of CYP8B1 dampened the metabolic improvements associated with VSG. In contrast, short hairpin RNA-mediated CYP8B1 knockdown improved metabolism similar to those observed after VSG. Cyp8b1 deficiency diminished the metabolic effects of VSG. Further, VSG-induced alterations to the 12α-OH/non-12α-OH BA ratio in the BA pool depended on CYP8B1 expression level. Consequently, intestinal lipid absorption was restricted, and the gut microbiota (GM) profile was altered. Fecal microbiota transplantation from wild type-VSG mice (vs. fecal microbiota transplantation from wild-type-sham mice) improved metabolism in recipient mice, while there were no differences between mice that received fecal microbiota transplantation from knockout-sham and knockout-VSG mice. CONCLUSIONS: CYP8B1 is a critical downstream target of VSG. Modulation of BA composition and gut microbiota profile by targeting CYP8B1 may provide novel insight into the development of therapies that noninvasively mimic bariatric surgery to treat obesity and its complications.

Complete genome sequence of Bacillus cereus Z4, a biocontrol agent against tobacco black shank, isolated from the Western Pacific Ocean.

Bacillus species have been considered as promising biological control agents due to their excellent antimicrobial ability. Bacillus cereus strain Z4 was isolated from 2000 m deep sea sediments of the Western Pacific Ocean, which possesses significant antifungal activity against Phytophthora nicotianae, the pathogenic fungus of tobacco black shank disease. To reveal the underlying antifungal genetic mechanisms, here, we report the complete genomic sequence of the strain Z4. The genome has one circular chromosome of 5,664,309 bp with a G + C content of 35.31%, 109 tRNAs, and 43 rRNAs. Genomic analysis identified 10 gene clusters related to the biosynthesis of biocontrol active compounds, including bacillibactin, petrobactin, fengycin, and molybdenum cofactor. Meanwhile, 6 gene clusters were responsible for the biosynthesis of metabolites with unknown functions. Strain Z4 also contains a large number of genes encoding carbohydrate-active enzymes and secreted proteins, respectively. The whole genomic analysis of Bacillus cereus Z4 may provide a valuable reference for elucidating its biocontrol mechanism against tobacco black shank.

Enhancing plant growth in biofertilizer-amended soil through nitrogen-transforming microbial communities.

Biofertilizers have immense potential for enhancing agricultural productivity. However, there is still a need for clarification regarding the specific mechanisms through which these biofertilizers improve soil properties and stimulate plant growth. In this research, a bacterial agent was utilized to enhance plant growth and investigate the microbial modulation mechanism of soil nutrient turnover using metagenomic technology. The results demonstrated a significant increase in soil fast-acting nitrogen (by 46.7%) and fast-acting phosphorus (by 88.6%) upon application of the bacterial agent. This finding suggests that stimulated soil microbes contribute to enhanced nutrient transformation, ultimately leading to improved plant growth. Furthermore, the application of the bacterial agent had a notable impact on the accumulation of key genes involved in nitrogen cycling. Notably, it enhanced nitrification genes (amo, hao, and nar), while denitrification genes (nir and nor) showed a slight decrease. This indicates that ammonium oxidation may be the primary pathway for increasing fast-acting nitrogen in soils. Additionally, the bacterial agent influenced the composition and functional structure of the soil microbial community. Moreover, the metagenome-assembled genomes (MAGs) obtained from the soil microbial communities exhibited complementary metabolic processes, suggesting mutual nutrient exchange. These MAGs contained widely distributed and highly abundant genes encoding plant growth promotion (PGP) traits. These findings emphasize how soil microbial communities can enhance vegetation growth by increasing nutrient availability and regulating plant hormone production. This effect can be further enhanced by introducing inoculated microbial agents. In conclusion, this study provides novel insights into the mechanisms underlying the beneficial effects of biofertilizers on soil properties and plant growth. The significant increase in nutrient availability, modulation of key genes involved in nitrogen cycling, and the presence of MAGs encoding PGP traits highlight the potential of biofertilizers to improve agricultural practices. These findings have important implications for enhancing agricultural sustainability and productivity, with positive societal and environmental impacts.

Combination effects of baicalin with linezolid against Staphylococcus aureus biofilm-related infections: in vivo animal model.

Staphylococcus aureus is a gram-positive bacterium that can produce biofilm, and biofilm-associated infections are difficult to control. Biofilm prevents antibiotics from penetrating and killing the bacteria. Combined use of antimicrobials is a common strategy to treat S. aureus biofilm-related infections. In this in vivo study, the clinically isolated strain of S. aureus 17546 (t037) was selected to establish a biofilm-associated infection rat model, and baicalin and linezolid were used to treat the infection. CFU counting was used to determine the bacteria within the biofilm, the biofilm structure was viewed using scanning electron microscopy (SEM), histopathology was performed, and inflammatory factors were analyzed by ELISA. Baicalin was efficient in destroying the biofilm and exerted a synergistic bactericidal effect when combined with linezolid. Based on these findings, baicalin combined with linezolid may be efficacious in controlling S. aureus biofilm-related infections.

The Influence of Special Environments on SiC MOSFETs.

In this work, the influences of special environments (hydrogen gas and high temperature, high humidity environments) on the performance of three types of SiC MOSFETs are investigated. The results reveal several noteworthy observations. Firstly, after 500 h in a hydrogen gas environment, all the SiC MOSFETs exhibited a negative drift in threshold voltage, accompanied by an increase in maximum transconductance and drain current (@ VGS/VDS = 13 V/3 V). This phenomenon can be attributed to that the hydrogen atoms can increase the positive fixed charges in the oxide and increase the electron mobility in the channel. In addition, high temperature did not intensify the impact of hydrogen on the devices and electron mobility. Instead, prolonged exposure to high temperatures may induce stress on the SiO2/SiC interface, leading to a decrease in electron mobility, subsequently reducing the transconductance and drain current (@ VGS/VDS = 13 V/3 V). The high temperature, high humidity environment can cause a certain negative drift in the devices' threshold voltage. With the increasing duration of the experiment, the maximum transconductance and drain current (@ VGS/VDS = 18V (20 V)/3 V) gradually decreased. This may be because the presence of moisture can lead to corrosion of the devices' metal contacts and interconnects, which can increase the devices' resistance and lead to a decrease in the devices' maximum transconductance and drain current.

Two-step hierarchical neural network for classification of dry age-related macular degeneration using optical coherence tomography images.

Purpose: The aim of this study is to apply deep learning techniques for the development and validation of a system that categorizes various phases of dry age-related macular degeneration (AMD), including nascent geographic atrophy (nGA), through the analysis of optical coherence tomography (OCT) images. Methods: A total of 3,401 OCT macular images obtained from 338 patients admitted to Shenyang Aier Eye Hospital in 2019-2021 were collected for the development of the classification model. We adopted a convolutional neural network (CNN) model and introduced hierarchical structure along with image enhancement techniques to train a two-step CNN model to detect and classify normal and three phases of dry AMD: atrophy-associated drusen regression, nGA, and geographic atrophy (GA). Five-fold cross-validation was used to evaluate the performance of the multi-label classification model. Results: Experimental results obtained from five-fold cross-validation with different dry AMD classification models show that the proposed two-step hierarchical model with image enhancement achieves the best classification performance, with a f1-score of 91.32% and a kappa coefficients of 96.09% compared to the state-of-the-art models. The results obtained from the ablation study demonstrate that the proposed method not only improves accuracy across all categories in comparison to a traditional flat CNN model, but also substantially enhances the classification performance of nGA, with an improvement from 66.79 to 81.65%. Conclusion: This study introduces a novel two-step hierarchical deep learning approach in categorizing dry AMD progression phases, and demonstrates its efficacy. The high classification performance suggests its potential for guiding individualized treatment plans for patients with macular degeneration.

Federated Data Quality Assessment Approach: Robust Learning With Mixed Label Noise.

Federated learning (FL) has been an effective way to train a machine learning model distributedly, holding local data without exchanging them. However, due to the inaccessibility of local data, FL with label noise would be more challenging. Most existing methods assume only open-set or closed-set noise and correspondingly propose filtering or correction solutions, ignoring that label noise can be mixed in real-world scenarios. In this article, we propose a novel FL method to discriminate the type of noise and make the FL mixed noise-robust, named FedMIN. FedMIN employs a composite framework that captures local-global differences in multiparticipant distributions to model generalized noise patterns. By determining adaptive thresholds for identifying mixed label noise in each client and assigning appropriate weights during model aggregation, FedMIN enhances the performance of the global model. Furthermore, FedMIN incorporates a loss alignment mechanism using local and global Gaussian mixture models (GMMs) to mitigate the risk of revealing samplewise loss. Extensive experiments are conducted on several public datasets, which include the simulated FL testbeds, i.e., CIFAR-10, CIFAR-100, and SVHN, and the real-world ones, i.e., Camelyon17 and multiorgan nuclei challenge (MoNuSAC). Compared to FL benchmarks, FedMIN improves model accuracy by up to 9.9% due to its superior noise estimation capabilities.

MetaFed: Federated Learning Among Federations With Cyclic Knowledge Distillation for Personalized Healthcare.

Federated learning (FL) has attracted increasing attention to building models without accessing raw user data, especially in healthcare. In real applications, different federations can seldom work together due to possible reasons such as data heterogeneity and distrust/inexistence of the central server. In this article, we propose a novel framework called MetaFed to facilitate trustworthy FL between different federations. obtains a personalized model for each federation without a central server via the proposed cyclic knowledge distillation. Specifically, treats each federation as a meta distribution and aggregates knowledge of each federation in a cyclic manner. The training is split into two parts: common knowledge accumulation and personalization. Comprehensive experiments on seven benchmarks demonstrate that without a server achieves better accuracy compared with state-of-the-art methods e.g., 10 %+ accuracy improvement compared with the baseline for physical activity monitoring dataset (PAMAP2) with fewer communication costs. More importantly, shows remarkable performance in real-healthcare-related applications.