Deep generative learning cannot only be used for generating new data with statistical characteristics derived from input data but also for anomaly detection, by separating nominal and anomalous instances based on their reconstruction quality. In this paper, we explore the performance of three unsupervised deep generative models-variational autoencoders (VAEs) with Gaussian, Bernoulli, and Boltzmann priors-in detecting anomalies in multivariate time series of commercial-flight operations. We created two VAE models with discrete latent variables (DVAEs), one with a factorized Bernoulli prior and one with a restricted Boltzmann machine (RBM) with novel positive-phase architecture as prior, because of the demand for discrete-variable models in machine-learning applications and because the integration of quantum devices based on two-level quantum systems requires such models. To the best of our knowledge, our work is the first that applies DVAE models to anomaly-detection tasks in the aerospace field. The DVAE with RBM prior, using a relatively simple-and classically or quantum-mechanically enhanceable-sampling technique for the evolution of the RBM's negative phase, performed better in detecting anomalies than the Bernoulli DVAE and on par with the Gaussian model, which has a continuous latent space. The transfer of a model to an unseen dataset with the same anomaly but without re-tuning of hyperparameters or re-training noticeably impaired anomaly-detection performance, but performance could be improved by post-training on the new dataset. The RBM model was robust to change of anomaly type and phase of flight during which the anomaly occurred. Our studies demonstrate the competitiveness of a discrete deep generative model with its Gaussian counterpart on anomaly-detection problems. Moreover, the DVAE model with RBM prior can be easily integrated with quantum sampling by outsourcing its generative process to measurements of quantum states obtained from a quantum annealer or gate-model device.

Anomaly detection in commercial aviation is an extremely challenging yet crucial task. Accurately detecting operationally significant anomalies can save civilian lives and/or result in significant savings in maintenance cost. The current practice uses manually tuned rule-based mechanisms to flag exceedances from predefined safety boundaries. However, this system cannot identify unknown risks and emerging vulnerabilities. Recently, innovative approaches based on machine learning have been used to automate anomaly detection. However, there are limits to their applicability in the field of aviation due to several challenges: 1) Properly reviewed data are scarce in aviation and, as a result, supervised learning cannot reach optimal performance. 2) Operationally significant anomalies do not coincide with statistically significant ones and, as a result, unsupervised learning fails to provide reliable and robust performance. In this paper, we propose a semi-supervised active learning framework for anomaly detection (SALAD), which detects operationally significant anomalies in flight operational quality assurance data. The developed framework works with vast amounts of unlabeled data as well as a small quantity of labeled data reviewed by subject matter experts to reliably identify safety anomalies in flight operations. Moreover, the model’s active learning strategy allows it to detect unknown anomalies that might emerge in the system. We validate the performance of the SALAD with a real-world case study of anomaly detection during the approach to landing of commercial aircraft. We specifically show that the proposed framework reaches reliable performance when only 1% of the data is labeled and can identify unknown anomalies effectively.

View Video Presentation: https://doi.org/10.2514/6.2022-2542.vid Anomaly detection in commercial aviation is an extremely challenging yet crucial task. Accurately detecting operationally significant anomalies in operational data can save civilian lives and/or result in significant savings in aircraft/maintenance cost. The current practice uses manually tuned rule-based mechanisms to flag exceedances from pre-defined safety boundaries. However, this system cannot identify unknown risks and emerging vulnerabilities. Recently, innovative approaches based on data science and machine learning have been utilized to automate anomaly detection. However, there are limits to their applicability in the field of aviation due to several fundamental challenges: (1) Properly reviewed data is scarce in aviation and, as a result, supervised learning approaches cannot reach optimal performance. (2) Operationally significant anomalies do not coincide with statistically significant ones and, as a result, unsupervised learning approaches fail to provide reliable and robust performance. In this paper, we propose SALAD, a Semi-supervised Active Learning framework for Anomaly Detection, which detects operationally significant anomalies in flight operational quality assurance data. The developed framework works with vast amounts of unlabeled data as well as a small quantity of labeled data reviewed by subject matter experts (SMEs) to reliably identify safety anomalies in flight operations. Moreover, the model's active learning strategy allows it to detect unknown risks and anomalies that might emerge in the system earlier with the help of SME reviews. We validate performance of the SALAD framework and its components with a real-world case study of multi-class anomaly detection during the approach to landing of commercial aircraft. We specifically show that the proposed framework reaches reliable anomaly-detection accuracy when only one percent of the data is reviewed and labeled and can identify unknown anomalies effectively.

The identification of precursors to safety incidents in aviation data is a crucial task, yet extremely challenging. The main approach in practice leverages domain expertise to define expected tolerances in system behavior and flags exceedances from such safety margins. However, this approach is incapable of identifying unknown risks and vulnerabilities. Various machine-learning approaches have been investigated and deployed to identify anomalies, with the great challenge of procuring enough labeled data to achieve reliable and accurate performance. This paper presents an explainable deep semi-supervised model for anomaly detection in aviation, building upon recent advancements described in the machine-learning literature. The proposed model combines feature engineering and classification in feature space, while leveraging all available data (labeled and unlabeled). Our approach is validated with case studies of anomaly detection during the takeoff and landing phases of commercial aircraft. Our model outperforms the state-of-the-art supervised anomaly-detection model, reaching significantly higher accuracy and fewer false alarms, even if only small proportion of data in the training set is labeled.

A significant target for radiation-induced effects is the microvascular system, which is critical to healthy tissue function and its pathology is linked to disrupted endothelial barrier function. Low-linear energy transfer (LET) ionizing radiation is a source of noncancer pathologies in humans and little is known about the early events that could initiate subsequent diseases. However, it is well known that gamma radiation causes a very early disruption of the endothelial barrier at doses below those required for cytotoxic effects. After irradiation of human umbilical vein endothelial cells (HUVECs) to doses as low as 2 Gy, transendothelial electrical resistance (TEER) is transiently reduced at 3 h, and the platelet-derived endothothelial cell adhesion molecule (PECAM-1 or CD31) is uncoupled from the cells along with the release of endothelial microparticles (EMPs). In this study, we measured TEER reduction as an indicator of barrier function loss, and specifically examined the shedding of EMPs from human endothelial barrier models after a variety of low-LET irradiations, including photons and charged particles. Our findings showed two TEER responses, dependent on radiation type and environmental conditions. The first response was diminishing oscillations of TEER, which occurred during the first 10 h postirradiation. This response occurred after a 5 Gy proton or helium-ion (1 GeV/n) dose in addition to a 5 Gy gamma or X radiation dose. This occurred only in the presence of multiple growth factors and did not show a dose response, nor was it associated with EMP release. The second response was a single acute drop in TEER at 3 h after photon irradiation. Dose response was observed and was associated with the shedding of EMPs in 2D barrier cultures and in 3D vessel models. In this case, helium-ion and proton irradiations did not induce a drop in TEER or shedding of EMPs. The photon radiation effects was observed both in serum-free media and in the presence of multiple growth factors, indicating that it occurs under a range of environmental conditions. These results show an acute response of the human endothelial barrier that is relevant to photon irradiation. Significantly, it involves the release of EMPs, which have recently attracted attention due to their emerging clinical importance.

A limiting factor in the treatment of cancer with radiotherapy is the damage to surrounding normal tissue, particularly the vasculature. Vessel pathologies are a major feature of the side effects of radiotherapy and little is known about early events that could initiate subsequent diseases. We tested the hypothesis that gamma radiation has early damaging effects on the human endothelial barrier. Two models were used; Human Brain Microcapillary Endothelial Cells (HBMEC), and Human Umbilical Vein Endothelial Cells (HUVEC). Endpoints included Trans-Endothelial Electrical Resistance (TEER), barrier permeability to 10 kDa and 70 kDa tracer molecules, and the localization of F-actin, and junction proteins and the Platelet Endothelial Cell Adhesion Molecule (PECAM-1). Radiation induced a rapid and transient decrease in TEER at 3 h, with effects also seen at the radiotherapy doses. This dip in resistance correlated to the transient loss of PECAM-1 in discrete areas where cells often detached from the monolayer leaving gaps. Redistribution of PECAM-1 was also seen in 3-D human tissue models. By 6 h, the remaining cells had migrated to reseal the barrier, coincident with TEER returning to control levels. Resealed monolayers contained fewer cells per unit area and their barrier function was weakened as evidenced by an increased permeability over 24 h. This is the first demonstration of a transient and rapid effect of gamma radiation on human endothelial barriers that involves cell detachment and the loss of PECAM-1. Considering the association of cell adhesion molecules with vasculopathies, such an effect has the potential to be clinically relevant to the longer-term effects of radiotherapy.

PURPOSE:Previously, we showed that microRNA (miRNA) signatures derived from the peripheral blood of mice are highly specific for both radiation energy (γ-rays or high linear energy transfer [LET] (56)Fe ions) and radiation dose. Here, we investigate to what extent miRNA expression signatures derived from mouse blood can be used as biomarkers for exposure to 600 MeV proton radiation.MATERIALS AND METHODS:We exposed mice to 600 MeV protons, using doses of 0.5 or 1.0 Gy, isolated total RNA at 6 h or 24 h after irradiation, and used quantitative real-time polymerase chain reaction (PCR) to determine the changes in miRNA expression.RESULTS:A total of 26 miRNA were differentially expressed after proton irradiation, in either one (77%) or multiple conditions (23%). Statistical classifiers based on proton, γ, and (56)Fe-ion miRNA expression signatures predicted radiation type and proton dose with accuracies of 81% and 88%, respectively. Importantly, gene ontology analysis for proton-irradiated cells shows that genes targeted by radiation-induced miRNA are involved in biological processes and molecular functions similar to those controlled by miRNA in γ ray- and (56)Fe-irradiated cells.CONCLUSIONS:Mouse blood miRNA signatures induced by proton, γ, or (56)Fe irradiation are radiation type- and dose-specific. These findings underline the complexity of the miRNA-mediated radiation response.

Purpose: Biomarkers of ionising radiation exposure are useful in a variety of scenarios, such as medical diagnostic imaging, occupational exposures, and spaceflight. This study investigates to what extent microRNA (miRNA) expression signatures in mouse peripheral blood can be used as biomarkers for exposures to radiation with low and high linear energy transfers. Materials and methods: Mice were irradiated with doses of 0.5, 1.5, or 5.0 Gy γ-rays (dose rate of 0.0136 Gy/s) or with doses of 0.1 or 0.5 Gy 56Fe ions (dose rate of 0.00208 Gy/s). Total RNA was isolated from whole blood at 6 h or 24 h after irradiation. Three animals per irradiation condition were used. Differentially expressed miRNA were determined by means of quantitative real-time polymerase chain reaction. Results: miRNA expression signatures were radiation type-specific and dose- and time-dependent. The differentially expressed miRNA were expressed in either one condition (71%) or multiple conditions (29%). Classifiers based on the differentially expressed miRNA predicted radiation type or dose with accuracies between 75% and 100%. Gene-ontology analyses show that miRNA induced by irradiation are involved in the control of several biological processes, such as mRNA transcription regulation, nucleic-acid metabolism, and development. Conclusion: miRNA signatures induced by ionising radiation in mouse blood are radiation type- and radiation dose-specific. These findings underline the complexity of the radiation response and the importance of miRNA in it.

PURPOSE:MicroRNAs (miRNAs), a class of noncoding small RNAs that regulate gene expression, are involved in numerous physiologic processes in normal and malignant cells. Our in vivo study measured miRNA and gene expression changes in human blood cells in response to ionizing radiation, to develop miRNA signatures that can be used as biomarkers for radiation exposure. METHODS AND MATERIALS:Blood from 8 radiotherapy patients in complete remission 1 or 2 was collected immediately before and 4 hours after total body irradiation with 1.25 Gy x-rays. Both miRNA and gene expression changes were measured by means of quantitative polymerase chain reaction and microarray hybridization, respectively. Hierarchic clustering, multidimensional scaling, class prediction, and gene ontology analysis were performed to investigate the potential of miRNAs to serve as radiation biomarkers and to elucidate their likely physiologic roles in the radiation response. RESULTS:The expression levels of 45 miRNAs were statistically significantly upregulated 4 hours after irradiation with 1.25 Gy x-rays, 27 of them in every patient. Nonirradiated and irradiated samples form separate clusters in hierarchic clustering and multidimensional scaling. Out of 223 differentially expressed genes, 37 were both downregulated and predicted targets of the upregulated miRNAs. Paired and unpaired miRNA-based classifiers that we developed can predict the class membership of a sample with unknown irradiation status, with accuracies of 100% when all 45 upregulated miRNAs are included. Both miRNA control of and gene involvement in biologic processes such as hemopoiesis and the immune response are increased after irradiation, whereas metabolic processes are underrepresented among all differentially expressed genes and the genes controlled by miRNAs. CONCLUSIONS:Exposure to ionizing radiation leads to the upregulation of the expression of a considerable proportion of the human miRNAome of peripheral blood cells. These miRNA expression signatures can be used as biomarkers of radiation exposure.

In many vertebrate species, the superior olive in the auditory brainstem plays an essential role in sound source localization. Little is known, however, about the structural and functional changes in this nucleus during development when alterations in head size and shape as well as in inner ear projections are expected to affect the perception of binaural cues. Using stereological techniques, we investigated the changes in several cellular and spatial features of the bullfrog superior olive across metamorphosis, the time period during which the animal transforms from a totally aquatic larva to a semiterrestrial adult. The total number of cells shows a strongly linear increase from hatchling through late larval stages. The number of neurons decreases during metamorphic climax stages, and recovers to pre-metamorphic climax levels in the early post-metamorphic froglet stage. The number of glial cells increases during the early larval period, and remains relatively stable, with no systematic variation, from late larval to froglet stages. The volume of the superior olive increases rapidly in early larval stages, followed by a much-attenuated rate of growth between late larval and froglet stages. These morphological changes may provide a substrate for the functional restructuring of the bullfrog superior olive, shortly before the switch from aquatic to mostly atmospheric hearing.