Defence Research and Development Canada

Emotional and workload-related states unfold dynamically during immersive virtual reality (VR) experiences, yet reliable physiological modeling in such environments remains challenging. We investigated whether multi-channel facial electromyography (fEMG), combined with spatio-temporal deep learning, can (i) accurately classify calibrated facial expressions across participants and (ii) transfer to spontaneous, task-elicited behavior in immersive VR. Twelve adults completed a calibration phase involving four intentional expressions (smile, frown, raised eyebrow, neutral), followed by VR scenes designed to elicit emotional, cognitive, physical, and dual task demands. After participant-level physiological normalization, a single shared Convolutional Neural Network-Temporal Convolutional Network (CNN-TCN) model was trained and evaluated using leave-one-participant-out (LOPO) validation. The model achieved strong cross-participant performance (Macro-F1 = 0.88 ± 0.13; ROC-AUC = 0.95 ± 0.06). When applied to unlabeled spontaneous VR task-elicited fEMG recordings, the trained model generated continuous expression classes. Derived static and temporal expression features showed scene-dependent modulation and False Discovery Rate (FDR)-surviving associations, primarily with perceived physical demand (NASA-TLX). The observed muscle activation patterns were physiologically plausible and aligned with Facial Action Coding System (FACS)-based interpretations of underlying muscle activity. These findings demonstrate that end-to-end spatio-temporal modeling of raw fEMG enables facial expression sensing in immersive VR using a single shared model following physiological normalization. The proposed framework bridges calibrated expression learning and spontaneous task-elicited behavior, supporting privacy-preserving, continuous and physiologically grounded monitoring in human-centered VR applications.

We propose a simple and cost-effective distributed force sensing technique based on intermodal coupling in polarization-maintaining few-mode fibers. A broadband linearly polarized optical source is injected into a single-mode-few-mode-single-mode fiber structure, and the resulting interferometric signal is analyzed using an optical spectrum analyzer. Localized external perturbations are detected via the induced coupling to higher-order modes, with spatial information extracted through Fourier-domain analysis of the spectrum, in a manner similar to optical frequency domain reflectometry. Experimental results demonstrate a spatial resolution of 8 cm of a maximum of 38 m sensing length, directional sensitivity, and the capability to resolve multiple force points along meter-scale fiber lengths. Transverse forces ranging from 0.65 N to 3.8 N have been tested, yielding a measurement precision of approximately 0.03 N. This technique offers a compact, temperature-independent method for real-time monitoring, providing an alternative to conventional distributed sensing methods.

BackgroundMilitary aircrew can be exposed to extreme occupational and physiological stressors, such as extreme G‐forces, decompression stress, and intermittent hypoxia and/or hyperoxia. These stressors may contribute to cognitive dysfunction and neurobiological damage. Previous studies on U.S. Air Force U‐2 pilots found a higher burden of white matter hyperintensities (WMH) on MRI, commonly associated with cerebrovascular disease, were linked to lower cognitive test performance in otherwise healthy, highly functioning individuals. Prompted by findings in U‐2 pilots, we investigated the prevalence and neurocognitive correlates of WMH in Royal Canadian Air Force (RCAF) fighter pilots and aircrew.MethodAll study participants were Canadian Armed Forces personnel, including 42 high performance aircraft pilots and 6 high altitude, non‐pilot aircrew. Prior to testing, detailed medical, environmental and occupational exposure histories were acquired. The testing included brain MRI (3T) and neurocognitive assessments. A standardized and validated volumetric image segmentation algorithm was used to acquire WMH and enlarged perivascular spaces (PVS). The Mann‐Whitney U test was used to compare RCAF WMH volumes with a convenience sample of 12 age‐matched normal controls from a previous scan‐rescan study. Regression models were used to examine associations between WMH and PVS with neurocognitive test performance in the RCAF study participants.ResultThe RCAF group had a higher mean volume of WMH compared to the normal control group (RCAF: 497.62mm3 vs. NC: 158.58mm3, p = 0.001). High performance flight hours were significantly correlated with PVS volumes (r = 0.352, p = 0.022). Regression models revealed WMH were associated with Shipley‐2 vocabulary scores (crystallized intelligence) standardized for age group (β=‐0.307, p = 0.012), short‐term visual memory (Delayed Matching‐to‐sample, Average Trial Accuracy, β=‐.406, p = 0.001), and working memory updating (D‐Prime for 1 back trials: hit rate – false alarm rate, β=‐.320, p = 0.008).ConclusionCanadian military aircrew in the RCAF have a higher burden of WMH compared to normal controls. The findings suggest that increases in WMH volumes, potentially due to extreme occupational and physiological stressors from long exposure to high performance flights, are associated with subtle neurocognitive impairment. Future analyses will examine the potential modulating effects related to changes in blood‐based neurological injury biomarkers.

Chronic alcohol exposure may trigger astrogliosis-a process involving hypertrophy and upregulation of astrocyte-specific markers following neuronal stress or injury-as suggested by trends in preclinical studies. However, in vivo evidence of astrogliosis in alcohol use disorder (AUD) is lacking. Here, we investigated the status of the astrocyte marker MAO-B -an enzyme predominantly expressed in astrocytes and upregulated during astrogliosis - using [11C]SL25.1188 positron emission tomography imaging in healthy controls (n = 28) and people with AUD after 3.5 ± 3.7 (n = 24) and 24 ± 7 days (in a subset: n = 8) of abstinence. Clinical symptoms of AUD were assessed, alongside peripheral markers of astrocyte activation and neuronal injury: glial fibrillary acidic protein (GFAP) and neurofilament light chain (NF-L), respectively. While mean [11C]SL25.1188 binding did not differ significantly between people with and without AUD at either abstinence time-point, binding was notably more variable in AUD and inversely correlated with AUD severity, withdrawal and anxiety (p < 0.05). Plasma GFAP and NF-L were elevated in people with AUD. Daily cigarette use was associated with lower [11C]SL25.1188 binding in people with AUD (-40%) and control participants (-33%). These findings reflect variability in MAO-B binding in AUD and support a potential link between lower MAO-B and greater clinical severity. The observed association between cigarette use and lower MAO-B binding replicates prior reports and extends this observation to AUD. The relationship between higher MAO-B binding and lower AUD severity may reflect compensatory reactive astrogliosis. Understanding whether MAO-B status reflects a beneficial or detrimental astrocytic response in AUD may be important for glial-targeted treatments.

<title>Abstract</title> We present BiRLNN, a bidirectional molecular design framework that combines recurrent neural networks with reinforcement learning to optimize drug-like properties of generated compounds. Unlike traditional unidirectional models, our approach generates molecular sequences synchronously in both forward and backward directions, enabling more balanced exploration of chemical space while maintaining constraint requirements during molecular design. We employ Self-Referencing Embedded Strings representations to ensure 100% syntactic validity of generated molecules. To guide generation towards desirable pharmacological targets, we implement a multi-objective reward function based on quantitative estimate of drug-likeness and synthetic accessibility, and apply policy gradient-based reinforcement learning for fine-tuning. Simulation results show that our model covers a much larger part of the chemical space compared to unidirectional ones, allowing it to explore regions that contain molecules unreachable by the latter. Moreover, the reinforcement learning process successfully steers the generation process toward desirable compound classes with improved reward metrics. Our results demonstrate that BiRLNN offers a robust and flexible strategy for navigating chemical space in multi-objective drug design tasks.

Abstract As cyber threats become increasingly automated and sophisticated, novel solutions must be introduced to improve defense of enterprise networks. Deep reinforcement learning (DRL) has demonstrated potential in mitigating these advanced threats. Single DRL agents have proven utility toward execution of autonomous cyber defense. Despite the success of employing single DRL agents, this approach presents significant limitations, especially regarding scalability within large enterprise networks. An attractive alternative to the single‐agent approach is the use of multi‐agent reinforcement learning (MARL). However, developing MARL agents is costly with few options for examining MARL cyber defense techniques against adversarial agents. This paper presents a MARL network security environment, the fourth iteration of the cyber autonomy gym for experimentation (CAGE) challenges. This challenge was specifically designed to test the efficacy of MARL algorithms in an enterprise network. Our work aims to evaluate the potential of MARL as a robust and scalable solution for autonomous network defense.

A molecularly imprintedpolymer (MIP)-based electrochemical sensorfor the rapid detection of fentanyl is reported. The sensor was preparedby electrochemically grafting polydopamine on a carbon nanofiber–Ptnanoparticle composite-modified screen-printed electrode. Dopaminewas identified as a suitable functional monomer via in-silico modelingand was electropolymerized via cyclic voltammetry in the presenceof fentanyl to form the MIP sensor. The properties and morphologyof the sensing material were characterized with spectroscopy, microscopy,and electrochemical techniques. Factors influencing the sensor performancewere studied and optimized. Under optimized conditions, the MIP sensorresponse followed the Langmuir–Freundlich binding isothermwith a dissociation constant (kd) of 16.13μM and a limit of detection of 0.094 μM fentanyl. Thesensor displayed good run-to-run repeatability and batch-to-batchperformance reproducibility with relative standard deviations of 6.7%(n = 5) and 9.1% (n = 3), respectively.Three sensors, prepared and tested in parallel, showed excellent storagestability in a fridge under a humidified environment for 4 weeks withrelative standard deviations of ≤10%. The developed MIP sensorpresented suitable selectivity when interrogated with solutions composedof equimolar concentrations of fentanyl and glucose, acetaminophen,theophylline, morphine, naloxone, codeine, or norfentanyl. The sensorwas also successfully tested in artificial urine samples, indicatingthat it is a promising candidate as a rapid testing method in fentanylinvestigation.

Interactive information visualisation (IViz) and visual analytics (VA) that adopt a complex system perspective would increase the understanding of the maritime traffic network (MTN) and improve maritime situational awareness (MSA).In this work, we concentrate on the relationships between the MTN components at different scales of the system, illustrating examples of IViz/VA tools that integrate visual elements to help focus the end user’s attention on these complex system features. The approach facilitates the discovery of "unknown unknowns" and the detection of emerging behaviours, facilitating the understanding of the elaborated relationships among system elements which often produce outcomes difficult to explain if individual components are examined in isolation. IViz/VA tools in support of use cases where the user is required to assess a maritime security or safety situation are illustrated.

BackgroundBlood transfusion is increasingly utilised to manage haemorrhagic shock in prehospital environments. This practice is particularly relevant in settings where delays to definitive treatment are common due to extended evacuation timelines, limited resupply, and challenging environmental conditions. Safe and effective transfusion in these contexts depends on competent, well-prepared providers. Non-physician personnel may be required to perform transfusions independently in high-stakes situations without direct physician supervision. This scoping review synthesizes current literature on blood transfusion training for prehospital providers, with a focus on instructional design, simulation modalities, knowledge retention, and outcome evaluation.MethodsWe conducted a scoping review following Joanna Briggs Institute methodology and reported in accordance with the PRISMA-ScR framework. Seven databases were systematically searched through 20 February 2025. Eligible studies described transfusion training for non-physician healthcare providers in prehospital or austere environments. Data were extracted on instructional strategies, simulation modalities, and outcome measures. Outcomes were categorised using the Kirkpatrick Model, and instructional design features mapped to simulation-based education frameworks.ResultsSix studies involving 475 participants were included. Participants included combat medics, paramedics, registered nurses, physician assistants, and medical students. Training was delivered across environments including simulation centres, field-based exercises, and in-theatre deployments. All studies featured face-to-face instruction and hands-on skills training. Simulation modalities included part-task trainers in four studies, high-fidelity mannequins in two, live human models in two, and real-world transfusions in one. Instructional design features such as team-based learning, repeated practice, and structured feedback were reported in most studies. Outcomes were reported across all four Kirkpatrick levels. Four studies assessed learner satisfaction and confidence (Level 1), five evaluated knowledge and procedural skill acquisition (Level 2), three assessed behavioural change in practice (Level 3), and one reported patient-level outcomes during operational missions (Level 4). None assessed long-term retention. Variability in instructional methods and limited evaluation at higher outcome levels constrained generalizability.ConclusionsBlood transfusion training for prehospital providers appears feasible and associated with short-term improvements in knowledge, skills, and confidence. However, inconsistent instructional design and limited evaluation of long-term or clinical outcomes indicate important gaps. Structured, simulation-informed programs aligned with operational needs may improve training consistency and effectiveness.Supplementary InformationThe online version contains supplementary material available at 10.1186/s13049-025-01440-0.

This article proposes a novel design for a tunable coaxial dual-band filter capable of simultaneously changing the center frequency of both bands. The resonator consists of a metallic rod with two fins of different dimensions inserted inside a cavity housing. Each fin represents a <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$(\lambda /4)$ </tex-math></inline-formula> resonator that is responsible for covering one band. The housing is shaped to change the capacitive loading on the two <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$(\lambda /4)$ </tex-math></inline-formula> resonators and change their resonance frequencies as the rod rotates inside the housing. Specific iris shapes are designed to provide proper coupling between adjacent dual-band resonator and realize the tunable dual-band filter. A feature of the proposed design is that different shapes of housings can result in different directions of movement of the passbands within each range. Additionally, this design utilizes a single tuning element in the middle of the housing to rotate the resonators, independently of the filter’s order. To verify this concept, two reconfigurable dual-band six-pole filters were designed and tested. The only difference between these two designs is the shape of their housings, which demonstrates different directions of movement of the two passbands. The proposed concept demonstrates the feasibility of realizing tunable dual-band filters using a single tuning element for the first time. This article also shows how the concept can extended to realize a tunable diplexer with a single tuning element.