Normal Regions Research Articles

Deep clustering hierarchical latent representation for anomaly-based cyber-attack detection

In the field of anomaly detection, well-known techniques and state-of-the-art models often face challenges when interpreting the latent space, which hinders their behavioral classification accuracy. Firstly, the sub-optimal distribution of data points within the latent space makes normal behavioral regions verbose and indistinguishable from abnormal regions. Secondly, within the latent space, it can be difficult to identify meaningful, separable, and indicative features. Finally, the processing time at the inference stage is still relatively slow.This paper aims to improve the accuracy of network anomaly detection mechanisms by proposing two novel deep hierarchical representation learning models: Deep Nested Clustering Auto-Encoder (DNCAE) and Deep Clustering Hierarchical Auto-Encoder (DCHAE). Both models adopt a nested branch structure, utilizing dual deep auto-encoders to establish hierarchical latent spaces; in each, clustering algorithms are used to spatially optimize and refine the data points. This approach results in improved separation between normal and abnormal data points, and easier identification of notable and/or indicative features.To ascertain the effectiveness of the approach and the quality of resulting features, both models were used in conjunction with ten different one-class anomaly detectors. Each of these ten anomaly detectors was evaluated on popular network intrusion datasets, notably: NSL-KDD, UNSW-NB15, CIC-IDS-2017, CSE-CIC-IDS-2018, and CTU13. Experimental results have confirmed that both of the proposed models produced higher levels of accuracy than existing baselines and current state-of-the-art models. Additionally, the processing time at the inference stage shows a significant reduction.

Physics-guided self-supervised learning for retrospective T1 and T2 mapping from conventional weighted brain MRI: Technical developments and initial validation in glioblastoma.

To develop a self-supervised learning method to retrospectively estimate T1 and T2 values from clinical weighted MRI. A self-supervised learning approach was constructed to estimate T1, T2, and proton density maps from conventional T1- and T2-weighted images. MR physics models were employed to regenerate the weighted images from the network outputs, and the network was optimized based on loss calculated between the synthesized and input weighted images, alongside additional constraints based on prior information. The method was evaluated on healthy volunteer data, with conventional mapping as references. The reproducibility was examined on two 3.0T scanners. Performance in tumor characterization was inspected by applying the method to a public glioblastoma dataset. For T1 and T2 estimation from three weighted images (T1 MPRAGE, T1 gradient echo sequences, and T2 turbo spin echo), the deep learning method achieved global voxel-wise error ≤9% in brain parenchyma and regional error ≤12.2% in six types of brain tissues. The regional measurements obtained from two scanners showed mean differences ≤2.4% and correlation coefficients >0.98, demonstrating excellent reproducibility. In the 50 glioblastoma patients, the retrospective quantification results were in line with literature reports from prospective methods, and the T2 values were found to be higher in tumor regions, with sensitivity of 0.90 and specificity of 0.92 in a voxel-wise classification task between normal and abnormal regions. The self-supervised learning method is promising for retrospective T1 and T2 quantification from clinical MR images, with the potential to improve the availability of quantitative MRI and facilitate brain tumor characterization.

Numerical investigation on the deformation of railway embankment under normal faulting

Active faults in the earthquake region are consistently regarded as a potential geological hazard to the construction and operation of railway engineering. However, the effects of normal faulting on railway embankments have not been investigated thoroughly. For bridging this knowledge gap, three-dimensional finite element analysis considering the influence of faulting offset, the soil layer’s thickness, the fault dip angle and the embankment cross-fault angle are conducted to clarify the normal faulting effects on the railway embankment. Emphasis is given to the stress and strain characteristic in the fault rupture outcropping regions on the embankment, the deformation of the embankment centerline for design purposes, and the determination of the affected zones for railway embankment preservation. The analysis shows that the normal fault rupture outcropping regions on railway embankment are tensile yield in most cases. The existence of the soil layer and its thickening would widen the affected zones and the regions where the fault ruptures outcrops. The fault dip angle and the cross-fault angle of the embankment have a complex effect on the behaviors of the crossing embankment. The depth of the subsidence zone of the embankment would increase with the decrease of the fault dip angle and the large fault dip angle would change the primary fault rupture to be a compressive one directly above the fault line. If the embankment crosses the fault line obliquely, the curvature radius of the centerline would hardly meet the design code.

Evaluation of the mechanism of action of paracetamol, drotaverine, and peppermint oil and their effects in combination with hyoscine butylbromide on colonic motility: human ex-vivo study.

Drotaverine, paracetamol, and peppermint oil are often prescribed for the treatment of gastrointestinal spasm and pain. This study aimed to evaluate the effect of these drugs alone and combined with the well-known antispasmodic hyoscine butylbromide on the human colon. Colon samples were obtained from macroscopically normal regions of 68 patients undergoing surgery and studied in muscle bath. Drotaverine, paracetamol, and peppermint oil were tested alone and in combination with hyoscine butylbromide on (1) spontaneous contractility induced by isometric stretch (in the presence of 1µM tetrodotoxin) and (2) contractility induced by 10-5M carbachol and after (3) electrical field stimulation-induced selective stimulation of excitatory (in the presence of 1mMNω-nitro-L-arginine and 10µM MRS2179) and (4) inhibitory (under non-adrenergic, non-cholinergic conditions) pathways. (5) Drotaverine alone was also tested on cAMP-dependent pathway activated by forskolin. Compared with the vehicle, drotaverine and paracetamol (10-9-10-5M) did not modify spontaneous contractions, carbachol-induced contractions, and responses attributed to selective activation of excitatory pathways. The addition of hyoscine butylbromide (10-7-10-5M), concentration-dependently reduced myogenic contractions and carbachol- and electrical field stimulation-induced contractile responses. The association of paracetamol (10-4M) and hyoscine butylbromide (10-7-10-5M) was not different from hyoscine butylbromide alone (10-7-10-5M). At higher concentrations (10-3M-3*10-3M), paracetamol decreased myogenic and carbachol-induced contractions. The adenylate cyclase activator, forskolin, concentration-dependently reduced contractility, leading to smooth muscle relaxation. The effect of forskolin 10-7M was concentration-dependently enhanced by drotaverine (10-6M-10-5M). Peppermint oil reduced myogenic activity and carbachol- and electrical field stimulation-induced contractions. The association of hyoscine butylbromide and peppermint oil was synergistic since the interaction index measured with the isobologram was lower than 1. No effect was seen on the neural-mediated inhibitory responses with any of the drugs studied although peppermint oil reduced the subsequent off-contraction. Drotaverine and hyoscine butylbromide have a complementary effect on human colon motility as one stimulates the cAMP inhibitory pathway and the other inhibits the excitatory pathway. Peppermint oil is synergic with hyoscine butylbromide suggesting that a combination therapy may be more effective in treating patients. In contrast, at therapeutic concentrations, paracetamol does not modify colonic contractility, suggesting that the association of paracetamol and hyoscine butylbromide has independent analgesic and antispasmodic properties.

Size-Dependent Glioblastoma Targeting by Polymeric Nanoruler with Prolonged Blood Circulation.

Currently, there is no effective treatment for glioblastoma multiforme (GBM), the most frequent and malignant type of brain tumor. The blood-brain (tumor) barrier (BB(T)B), which is composed of tightly connected endothelial cells and pericytes (with partial vasculature collapse), hampers nanomedicine accumulation in tumor tissues. We aimed to explore the effect of nanomedicine size on passive targeting of GBM. A series of size-tunable poly(ethylene glycol) (PEG)-grafted copolymers (gPEGs) were constructed with hydrodynamic diameters of 8-30 nm. Biodistribution studies using orthotopic brain tumor-bearing mice revealed that gPEG brain tumor accumulation was maximized at 10 nm with ∼14 dose %/g of tumor, which was 19 times higher than that in the normal brain region and 4.2 times higher than that of 30-nm gPEG. Notably, 10-nm gPEG exhibited substantially higher brain tumor accumulation than 11-nm linear PEG owing to the prolonged blood circulation property of gPEGs, which is derived from a densely PEG-packed structure. 10 nm gPEG exhibited deeper penetration into the brain tumor tissue than the larger gPEGs did (>10 nm). This study demonstrates, for the first time, the great potential of a nanomedicine downsizing strategy for passive GBM targeting.

Anomalous to normal dispersion nonlinear optical dephasing switch in electromagnetically induced transparency using a Kerr effect

The atomic decoherence effect (DE) on a Kerr nonlinear (KNL) electromagnetically induced transparency (EIT)is studied in a Δ system. The DE between the ground state hyperfine levels is caused by the dephasing rate γ d which dramatically modifies the medium response. It controls the normal dispersive region which shows steep positive slopes for linear response at the line center while the nonlinear response experiences steep negative slopes for low γ d . The microwave field strength and γ d modify the nonlinear response from the anomalous dispersion to normal dispersion. The calculations show that room-temperature atoms are used to quantify the quantum interference (QI) on linear and nonlinear absorption with γ d . The EIT spectrum explores the understanding of the subluminal and superluminal wave propagation of probe signal and this study opens a new pathway for the understanding of the QI devices and their nonlinearities based on EIT.

CHARGE CARRIERS STATES IN A MODEL OF CuO SUPERCONDUCTIVE CERAMICS

<p>The translational symmetry of the distribution of atoms (ions) of the charge carriers (electrons or holes)<br />system is broken by sputtering (doping) due to the existence of two boundary surfaces. This is a model<br />of high-temperature superconductors in which the observed symmetry breaking orthogonal to the CuO<br />plane is treated as a perturbation. Single-particle fermion wave functions and possible charge carrier<br />energies were determined. The competing existence of superconducting and normal regions in such a<br />sample is shown in agreement with experimental data. The conditions for the formation of<br />superconducting states and the limits of the current density values in the planes parallel to the<br />boundary surfaces (in the CuO planes) were obtained and discussed.</p>

SHRINKAGE-BASED WEAKLY-SUPERVISED FEATURE LEARNING TO ENHANCE IoT ANOMALY DETECTION

IoT anomaly detection faces challenges due to the rarity of IoT anomalies and the limited availability of labels. Recent weakly-supervised approaches, like Feature Encoding with AutoEncoder and Weakly-supervised Anomaly Detection (FeaWAD) and an improvement on FeaWAD (iFWAD), address this scarcity by constructing detectors from a combination of unlabeled data and a small labeled anomalous set. While effective, these methods lack constraints during the feature learning stage to delineate normal regions from anomalies. Notably, the Shrink AutoEncoder promotes clustering of normal data around the origin while preserving space for anomalies. Drawing inspiration from the Shrink AutoEncoder, the study aims to introduce Shrink iFWAD (called sFWAD), embedding a shrink regularizer into iFWAD. This term compels the feature encoder of sFWAD to learn penalizing normal data that is close to zero, while simultaneously pushing IoT anomalies further away from zero. This process facilitates the anomalous score generator of sFWAD in efficiently identifying IoT anomalies. The proposed method is evaluated against state-of-the-art weakly-supervised techniques and other common anomaly detection methods using the N-BaIoT dataset. Experimental results indicate that sFWAD often surpasses recent weakly-supervised methods as well as the common techniques in IoT anomaly detection performance. For identifying unknown/new IoT anomalies, Missed Detection Rate from sFWAD (0.008) is much lower than those from iFWAD (0.026) and RoSAS (0.015).

Photonic crystal fiber temperature sensor based on self-phase modulation and solitons

In this paper, we investigated the temperature sensing properties of self-phase modulation (SPM) combined with solitons in photonic crystal fibers by experimental verification. Pumped in the normal dispersion region close to the zero-dispersion point, SPM allows the resulting spectrum to extend into the normal dispersion region, generating solitons. By detecting the wavelength shift of the soliton at 900 nm, 2.366 W, the maximum sensitivity is 0.98 nm/°C. To the best of our knowledge, this is the first study of temperature sensing using SPM in combination with solitons, which broadens the boundaries of nonlinear-based sensors and holds considerable promise for high-performance temperature detection in a variety of demanding scenarios, such as railway safety and national security.

Humanin activates integrin αV–TGFβ axis and leads to glioblastoma progression

The role of mitochondria peptides in the spreading of glioblastoma remains poorly understood. In this study, we investigated the mechanism underlying intracranial glioblastoma progression. Our findings demonstrate that the mitochondria-derived peptide, humanin, plays a significant role in enhancing glioblastoma progression through the intratumoral activation of the integrin alpha V (ITGAV)–TGF beta (TGFβ) signaling axis. In glioblastoma tissues, humanin showed a significant upregulation in the tumor area compared to the corresponding normal region. Utilizing multiple in vitro pharmacological and genetic approaches, we observed that humanin activates the ITGAV pathway, leading to cellular attachment and filopodia formation. This process aids the subsequent migration and invasion of attached glioblastoma cells through intracellular TGFβR signaling activation. In addition, our in vivo orthotopic glioblastoma model provides further support for the pro-tumoral function of humanin. We observed a correlation between poor survival and aggressive invasiveness in the humanin-treated group, with noticeable tumor protrusions and induced angiogenesis compared to the control. Intriguingly, the in vivo effect of humanin on glioblastoma was significantly reduced by the treatment of TGFBR1 inhibitor. To strengthen these findings, public database analysis revealed a significant association between genes in the ITGAV–TGFβR axis and poor prognosis in glioblastoma patients. These results collectively highlight humanin as a pro-tumoral factor, making it a promising biological target for treating glioblastoma.

Improving radiology report generation with multi-grained abnormality prediction

Conventional data-driven approaches for radiology report generation face a problem that the descriptions of the normal regions in the real data are much more than the that for abnormal ones, which potentially leads to the bias of losing focus on abnormalities. Previous work showed promising results owing to the fact that most of the content in reports are described within normal range, which, although fluent, has the limitation of tending to favor the evaluation metrics rather than produce useful hints for human judgment and model learning. To this end, we propose to explicitly predict abnormalities for radiology report generation, following a multi-task learning scheme to drive the model paying more attention on the abnormal regions with multi-grained information, including abnormalities in different granularities, so as to tackle the aforementioned limitation for report generation. In doing so, we propose a disease detector (DD) to identify coarse-grained abnormality, and a medical concept detector (MCD) to associate the predicted disease with the predefined fine-grained pathological concepts. To integrate the information from the proposed abnormality prediction, we design a dual-stream adaptive decoder that takes such information into account with a gate unit controls the integration at each generation step. Extensive experiment results on two widely used benchmark datasets indicate that our method achieves 29.8% and 34.2% performance improvement over the baseline on the natural language generation (NLG) metrics and clinical efficacy (CE) metrics respectively, demonstrate the superiority of the proposed approach.

Spatial proteomic landscape of primary and relapsed hepatocellular carcinoma reveals immune escape characteristics in early relapse.

Surgical resection serves as the principal curative strategy for hepatocellular carcinoma (HCC), yet the incidence of postoperative recurrence remains alarmingly high. However, the spatial molecular structural alterations contributing to postoperative recurrence in HCC are still poorly understood. We employed imaging mass cytometry to profile the in-situ expression of 33 proteins within 358,729 single cells of 92 clinically annotated surgical specimens from 46 patients who were treated with surgical resections for primary and relapsed tumors. We revealed the recurrence progression of HCC was governed by the dynamic spatial distribution and functional interplay of diverse cell types across adjacent normal, tumor margin, and intratumor regions. Our exhaustive analyses revealed an aggressive, immunosuppression-related spatial ecosystem in relapsed HCC. Additionally, we illustrated the prominent implications of the TME of tumor margins in association with relapse HCC. Moreover, we identified a novel subpopulation of dendritic cells (PDL1+CD103+ DCs) enriched in the peritumoral area that correlated with early postoperative recurrence, which was further validated in an external cohort. Through the analysis of scRNA-seq data, we found the interaction of PDL1+CD103+ DCs with regulatory T cells and exhausted T cells enhanced immunosuppression and immune escape via multiple ligand-receptor pathways. We comprehensively depicted the spatial landscape of single-cell dynamics and multicellular architecture within primary and relapsed HCC. Our findings highlight spatial organization is a prominent determinant of HCC recurrence and provide a valuable insight into the immune evasion mechanisms driving recurrence.

Label-Free Fiber-Optic Raman Spectroscopy for Intravascular Coronary Atherosclerosis and Plaque Detection.

The rupture of atherosclerotic plaques remains one of the leading causes of morbidity and mortality worldwide. The plaques have certain pathological characteristics including a fibrous cap, inflammation, and extensive lipid deposition in a lipid core. Various invasive and noninvasive imaging techniques can interrogate structural aspects of atheroma; however, the composition of the lipid core in coronary atherosclerosis and plaques cannot be accurately detected. Fiber-optic Raman spectroscopy has the capability of in vivo rapid and accurate biomarker detection as an emerging omics technology. Previous studies demonstrated that an intravascular Raman spectroscopic technique may assess and manage the therapeutic and medication strategies intraoperatively. The Raman spectral information identified plaque depositions consisting of lipids, triglycerides, and cholesterol esters as the major components by comparing normal region and early plaque formation region with histology. By focusing on the composition of plaques, we could identify the subgroups of plaques accurately and rapidly by Raman spectroscopy. Collectively, this fiber-optic Raman spectroscopy opens up new opportunities for coronary atherosclerosis and plaque detection, which would assist optimal surgical strategy and instant postoperative decision-making. In this paper, we will review the advancement of label-free fiber-optic Raman probe spectroscopy and its applications of coronary atherosclerosis and atherosclerotic plaque detection.

Multi-pulse femtosecond laser ablation of SiC/SiC composites: Morphological evolution, oxidation mechanisms and component characterization

SiC fiber reinforced SiC matrix (SiosC/SiC) composites processed by femtosecond laser play a pivotal role in thermal protection systems. This paper reveals the ablation characteristics of multi-pulse femtosecond laser ablation of SiC/SiC composites, including morphological evolution, oxidation mechanisms and component characterization by heat accumulation simulation and advanced experimental analysis. The results show that the ablation depth induced by the heat accumulation effect is an exponential function of laser parameters. Morphological analysis and component characterization indicate that two types of ablation regions are formed, normal ablation region is an oxidized zone induced by passive oxidation mechanism, and central ablation region is a non-oxidized zone induced by active oxidation mechanism. The increase of single pulse energy and decrease of scanning speed enhance the driving force of chemical reactions, thus dramatically affecting the gas hole and graphite properties in the normal ablation region. Furthermore, the propagation path of thermal stress-induced cracks is determined which emerges from the center ablation region, extending through gas holes to the normal ablation region.

Ultrasound Biomarkers: Contrast-enhanced Ultrasound and Nakagami Imaging to Differentiate Benign and Malignant Choroidal Tumor

Purpose We hypothesized that contrast-enhanced ultrasound (CEUS) using a microbubble technique to quantify microvascular changes and Nakagami imaging for tissue characterization would provide a new approach for diagnosing and differentiating benign and malignant choroidal lesions. Methods Five patients with choroidal melanoma (CM) and five patients with choroidal hemangioma (CH) were selected. Definity®, which contains perflutren microbubbles, was administered as a slow IV bolus (1 ml). CEUS was performed for 1 min postinjection of the contrast agent with ultrasound radiofrequency data acquired from 10 s to 60 s. The contrast value was calculated for the whole tumor region. A gradient magnitude method was used for each postcontrast frames with 1-second interval, and the time-averaged value in pixel intensity gradient of postinjection frames was estimated and reported. Based on the Nakagami statistical distribution model, two Nakagami parameters, m and Ω, where m (shape parameter), representing tissue heterogeneity, and Ω (scale parameter), representing the average energy of backscattered signals, were studied. Results CEUS analysis showed that the time-averaged estimated contrast was significantly higher (p = 0.008) for CH compared to CM. Furthermore, the time-averaged contrast within the normal choroidal region was significantly higher than the choroidal tumor region for both CH and CM (p = 0.001 for CH cases and p < 0.0001 for CM cases). Nakagami analysis showed that the m estimates were significantly higher (p = 0.032) for CH (m = 0.61) than for CM (m = 0.28), indicating that CH is a more heterogeneous tumor than CM. The Ω estimates were significantly higher (p = 0.0019) for CH (Ω = 0.15) compared to CM (Ω = 0.03). These results may be due to the more vascular structures in CH compared to CM. Conclusions Quantitative intensity-based perfusion analysis using CEUS and backscattering tissue analysis using Nakagami imaging can provide valuable insights to differentiate benign and malignant choroidal lesions.

Diffusion Analysis of Intracranial Epidermoid, Head and Neck Epidermal Inclusion Cyst, and Temporal Bone Cholesteatoma.

Intracranial epidermoid tumors, temporal bone cholesteatomas, and head and neck epidermoid inclusion cysts are typically slow-growing, benign conditions arising from ectodermal tissue. They exhibit increased signal on DWI. While much of the imaging literature describes these lesions as showing diffusion restriction, we aimed to investigate these qualitative signal intensities and interpretations of restricted diffusion with respect to normal brain structures. This study aimed to quantitatively evaluate the ADC values and histogram features of these lesions. This retrospective study included children with histologically confirmed diagnoses of intracranial epidermoid tumors, temporal bone cholesteatomas, or head and neck epidermoid inclusion cysts. Lesions were segmented, and voxelwise calculation of ADC values was performed along with histogram analysis. ADC calculations were validated with a second analysis software to ensure accuracy. Normal brain ROIs-including the cerebellum, white matter, and thalamus-served as normal comparators. Correlational analysis and Bland-Altman plots assessed agreement among software tools for ADC calculations. Differences in the distribution of values between the lesions and normal brain tissues were assessed using the Wilcoxon rank sum and Kruskal-Wallis tests. Forty-eight pathology-proved cases were included in this study. Among them, 13 (27.1%) patients had intracranial epidermoid tumors, 14 (29.2%) had head and neck epidermoid inclusion cysts, and 21 (43.7%) had temporal bone cholesteatomas. The mean age was 8.67 (SD, 5.30) years, and 27 (52.9%) were female. The intraclass correlation for absolute agreement for lesional ADC between the 2 software tools was 0.997 (95% CI, 0.995-0.998). The intracranial epidermoid tumor, head and neck epidermoid inclusion cyst, and temporal bone cholesteatoma median ADC values were not significantly different (973.7 versus 875.7 versus 933.2 ×10-6 mm2/s, P = .265). However, the ADCs of the 3 types of lesions were higher than those of 3 normal brain tissue types (933 versus 766, × 10-6 mm2/s, P < .0001). The ADC values of intracranial epidermoid tumors, temporal bone cholesteatomas, and head and neck epidermoid inclusion cysts are higher than those of normal brain regions. It is not accurate to simply classify these lesions as exhibiting restricted diffusion or reduced diffusivity without considering the tissue used for comparison. The observed hyperintensity on DWI compared with the brain is likely attributable to a relatively higher contribution of the T2 shinethrough effect.

Anomalous supercurrent and diode effect in locally perturbed topological Josephson junctions

The simultaneous breaking of time-reversal and inversion symmetry can lead to peculiar effects in Josephson junctions, such as the anomalous Josephson effect or supercurrent rectification, which is a dissipationless analog of the diode effect. Due to their impact in new quantum technologies, it is important to find robust platforms and external means to manipulate the above-mentioned effects in a controlled way. Here, we theoretically consider a Josephson junction based on a quantum spin Hall system as the normal channel, subjected to a magnetic field in the direction defined by spin-momentum locking, and in the presence of a local tip in close proximity to one of the metallic edges in the normal region. We consider different local perturbations, model normal and magnetic tips, and study how they affect the Josephson response of the device. In particular, we argue that magnetic tips are a useful tool that allows for tunability of both ϕ0 response and supercurrent rectification.

Towards interactive AI-authoring with prototypical few-shot classifiers in histopathology

A vast multitude of tasks in histopathology could potentially benefit from the support of artificial intelligence (AI). Many examples have been shown in the literature and first commercial products with FDA or CE-IVDR clearance are available. However, two key challenges remain: (1) a scarcity of thoroughly annotated images, respectively the laboriousness of this task, and (2) the creation of robust models that can cope with the data heterogeneity in the field (domain generalization). In this work, we investigate how the combination of prototypical few-shot classification models and data augmentation can address both of these challenges. Based on annotated data sets that include multiple centers, multiple scanners, and two tumor entities, we examine the robustness and the adaptability of few-shot classifiers in multiple scenarios. We demonstrate that data from one scanner and one site are sufficient to train robust few-shot classification models by applying domain-specific data augmentation. The models achieved classification performance of around 90% on a multiscanner and multicenter database, which is on par with the accuracy achieved on the primary single-center single-scanner data. Various convolutional neural network (CNN) architectures can be used for feature extraction in the few-shot model. A comparison of nine state-of-the-art architectures yielded that EfficientNet B0 provides the best trade-off between accuracy and inference time. The classification of prototypical few-shot models directly relies on class prototypes derived from example images of each class. Therefore, we investigated the influence of prototypes originating from images from different scanners and evaluated their performance also on the multiscanner database. Again, our few-shot model showed a stable performance with an average absolute deviation in accuracy compared to the primary prototypes of 1.8% points. Finally, we examined the adaptability to a new tumor entity: classification of tissue sections containing urothelial carcinoma into normal, tumor, and necrotic regions. Only three annotations per subclass (e.g., muscle and adipose tissue are subclasses of normal tissue) were provided to adapt the few-shot model, which obtained an overall accuracy of 93.6%. These results demonstrate that prototypical few-shot classification is an ideal technology for realizing an interactive AI authoring system as it only requires few annotations and can be adapted to new tasks without involving retraining of the underlying feature extraction CNN, which would in turn require a selection of hyper-parameters based on data science expert knowledge. Similarly, it can be regarded as a guided annotation system. To this end, we realized a workflow and user interface that targets non-technical users.

Current spectral norm and phase variation based fault region identification for active distribution network

The paper presents a fault region identification method for the active distribution network (ADN) with limited PMU. First, PMU configuration and region division strategies are proposed based on the network topology. Next, the difference in three-phase current variations between the normal and fault regions of the ADN is analyzed. A multi-dimensional state monitoring matrix is built using the measured current data. The spectral norm ratio coefficient is constructed based on the 2-norm to lower the complexity of the multi-dimensional state monitoring matrix and quantify the difference in state changes before and after the fault occurs in each region. Then, the fault region is identified by combining each region’s spectral norm ratio coefficient and the change of the current phase. Finally, an IEEE 33-node simulation model is created to simulate faults under different working conditions. According to the simulation results, the suggested approach is less impacted by fault type, neutral point grounding mode, and transition resistance. Furthermore, even if the communication does not match the rigorous synchronization requirements, the proposed method can still complete the fault identification of the distribution network correctly and has high robustness.

Evaluation of deep learning-based reconstruction late gadolinium enhancement images for identifying patients with clinically unrecognized myocardial infarction

BackgroundThe presence of infarction in patients with unrecognized myocardial infarction (UMI) is a critical feature in predicting adverse cardiac events. This study aimed to compare the detection rate of UMI using conventional and deep learning reconstruction (DLR)-based late gadolinium enhancement (LGEO and LGEDL, respectively) and evaluate optimal quantification parameters to enhance diagnosis and management of suspected patients with UMI.MethodsThis prospective study included 98 patients (68 men; mean age: 55.8 ± 8.1 years) with suspected UMI treated at our hospital from April 2022 to August 2023. LGEO and LGEDL images were obtained using conventional and commercially available inline DLR algorithms. The myocardial signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and percentage of enhanced area (Parea) employing the signal threshold versus reference mean (STRM) approach, which correlates the signal intensity (SI) within areas of interest with the average SI of normal regions, were analyzed. Analysis was performed using the standard deviation (SD) threshold approach (2SD–5SD) and full width at half maximum (FWHM) method. The diagnostic efficacies based on LGEDL and LGEO images were calculated.ResultsThe SNRDL and CNRDL were two times better than the SNRO and CNRO, respectively (P < 0.05). Parea−DL was elevated compared to Parea−O using the threshold methods (P < 0.05); however, no intergroup difference was found based on the FWHM method (P > 0.05). The Parea−DL and Parea−O also differed except between the 2SD and 3SD and the 4SD/5SD and FWHM methods (P < 0.05). The receiver operating characteristic curve analysis revealed that each SD method exhibited good diagnostic efficacy for detecting UMI, with the Parea−DL having the best diagnostic efficacy based on the 5SD method (P < 0.05). Overall, the LGEDL images had better image quality. Strong diagnostic efficacy for UMI identification was achieved when the STRM was ≥ 4SD and ≥ 3SD for the LGEDL and LGEO, respectively.ConclusionsSTRM selection for LGEDL magnetic resonance images helps improve clinical decision-making in patients with UMI. This study underscored the importance of STRM selection for analyzing LGEDL images to enhance diagnostic accuracy and clinical decision-making for patients with UMI, further providing better cardiovascular care.