Abstract: The growing importance of integrating renewable energy sources (RES) into mainline railway traction networks stems from the sector's substantial electricity demands, traditionally met by carbon-intensive thermal generation. This paper addresses the potential of wind and solar power to enhance energy efficiency and reduce emissions in rail transport. It details the devel-opment of digital models for simulating DC traction power systems (TPS) coupled with RES, specifically wind turbines. Given the complexity of TPS, effective integration requires digital modeling that accounts for their unique properties. The proposed methodology, based on phase coordinates algorithms, offers a universal and comprehensive framework. It enables the identi-fication of various operational modes (normal, emergency, special) for diverse network com-ponents, including traction networks, transmission lines, and transformers. These models were used to simulate real-world train operations, generating data on electrical parameter dynamics and transformer thermal conditions. The results confirm that wind integration can improve energy efficiency, validating the methodology's practical applicability for RES projects in DC traction networks, including advanced high-voltage systems.

Abstract: Background/Objectives: Artificial intelligence (AI) has transformed healthcare delivery by revolutionizing the offering opportunities in prognosis, diagnosis, personalize treatment, improving patient outcomes. However, little is known about the nurses’ perceptions and attitudes toward the integration of AI-driven conversational technology, AI chatbots into clinical practice. The aim of our study was to investigate nurses’ perceptions regarding the use of AI chatbots as a tool for mental health support. Additionally, the study aimed to evaluate their levels of acceptance and fear toward AI, while examining the influence of demographic variables on these attitudes. Methods: A cross-sectional study was conducted. Attitudes toward the use of AI-powered chatbots for mental health support were measured using the Artificial Intelligence in Mental Health Scale (AIMHS). Additionally, the Attitudes Towards Artificial Intelligence Scale (ATAI) was employed to assess nurses’ levels of acceptance and fear regarding artificial intelligence. Results: AIMHS score reflected moderately positive attitudes toward AI chatbots for mental health support, while ATAI scores indicated a moderate level of acceptance and fear toward AI. Multivariable analysis showed that increased age and increased daily engagement with social media and websites were significantly associated with more favorable technical perceptions of AI-based mental health chatbots. Also, male nurses exhibited significantly more favorable attitudes toward AI-based mental health chatbots in terms of perceived personal benefits. Higher levels of digital technology competence were significantly associated with greater acceptance of artificial intelligence. Additionally, male nurses reported significantly higher acceptance of AI compared to their female counterparts. We found that lower financial status was significantly associated with heightened fear of AI. Conclusions: Nurses generally held moderately positive attitudes toward both AI-based mental health chatbots and AI more broadly. Several demographic factors were found to significantly influence these attitudes.

Abstract: Buildings located in highly urbanized areas have not been considered for photovoltaic (PV) deployment on building walls due to limitation of ground and rooftops space. As the need for increasing energy demand due to population growth in cities, and the advancements in the efficiency of semi-transparent (ST-PV) solar cell technology, the integration of ST-PV modules into building windows, become feasible. The present article proposes a novel methodology for calculating the incident solar energy on PV vertical modules deployed on building walls and windows facing the southern direction and obscured by a nearby building in front. The present work analyses analytically, for the first time, the incident energy and its distribution on PV vertical modules along a wall height. Monthly and annually direct beam, diffuse and global energies are calculated for different wall height, building separation and orientation. The results shows, for example, that both the front and rear building walls receive the same amount of annual direct beam energy 913 kWh/m² for a distance 25 m between the buildings. Decreasing the distance from 25 m to 10 m, decreases the annual incident global energy on the rear-building wall by 15 %.

Abstract: In order to respond to the national " double carbon " strategic goal, promote the green and low-carbon transformation of the building materials industry, and develop low-carbon and environmentally friendly grouting materials, an AACGMs was prepared in this study. The effects of CG content, alkali activator modulus and alkali activator content on material fluidity, setting time, compressive strength and impermeability were systematically studied by orthogonal test. The optimal mix ratio was determined, and its internal mechanism was revealed by microscopic analysis. The results show that the comprehensive performance is the best when the content of CG is 50%, the modulus of alkali activator is 1.6 and the content of alkali activator is 14%. The primary and secondary order of the influence of various factors on the performance is : CG content > alkali activator content > alkali activator modulus. Microscopic analysis reveals that the hydrolysis polymerization products of the material are mainly C-S-H, C- (N) -A-S-H gel and zeolite-like phase, forming a dense three-dimensional network structure, which is the internal mechanism of its good mechanical and impermeability properties. This study provides a new idea for the utilization of CG, and the prepared materials are of great significance in the field of grouting reinforcement in underground engineering.

Abstract: Background: High-dose chemotherapy (HDCT) followed by autologous stem cell transplantation (ASCT) is an established salvage therapy for relapsed or refractory (R/R) germ cell tumors (GCTs). Methods: We conducted a retrospective analysis of 9 patients (23-41 years of age) with relapsed/refractory (R/R) GCTs treated according to the Swedish-Norwegian Testicular Cancer Group Clinical Protocol (SWENOTECA), at the Specialized Hospital for Active Treatment of Hematological Diseases (SHATHD) in Sofia, Bulgaria. The study evaluates the efficacy and safety of HDCT followed by ASCT in R/R GCTs eligible for second-line consolidation. Results: The median follow-up was 39.3 months. Following ASCT, the overall response rate (ORR) was 33.3%, consisting entirely of complete responses (CR). The 1-year OS and PFS were both 44.4%. Notably, Kaplan-Meier curves for OS and PFS reached a plateau after 24 months (33.3%), showing that long-term durable remission in these patients is achievable. Although grade 3–4 toxicities were observed, there was no treatment-related mortality. Conclusion: Tandem HDCT followed by ASCT is a safe and effective salvage strategy for R/R GCTs, offering predictable toxicity and curative potential, which is proven by the observed survival plateau. While tandem cycles are the current standard, the upcoming TIGER trial results will determine the future necessity of further intensification.

Abstract: This review examines implementation dimensions of integrated lemon biorefinery systems, including cascade valorisation design, circular-economy integration, life-cycle assessment, techno-economic feasibility, and regulatory frameworks. Bibliometric analysis of Web of Science data (2015–2025) reveals exponential growth in citrus-biorefinery research, with lemon representing a burgeoning subset. Techno-economic assessments indicate that cascade biorefineries recovering essential oils, pectin, polyphenols, nanocellulose, and bioenergy can achieve cumulative revenues of USD 400–650 per tonne of dry peel. Whilst small-scale units (<500 tonnes/year) struggle to achieve viability, industrial simulations demonstrate Internal Rates of Return exceeding 18% at processing scales above 100,000 tonnes annually (2025 basis). Life-cycle assessments confirm environmental benefits, with greenhouse gas reductions of 60–85% relative to conventional disposal. Critical success factors include adopting green extraction technologies to preserve bioactive integrity and mitigating D-limonene inhibition in downstream anaerobic digestion. These findings establish lemon biorefineries as technically mature, economically viable pathways for circular bioeconomy transitions, provided regulatory hurdles—Novel Foods authorisation (EU) and GRAS determination (US)—are effectively navigated.

Abstract: Background: Dying in the preferred place is considered an indicator of quality of end-of-life care. Advance care planning and home palliative care may increase the likelihood of dying at home. This study aimed to assess whether recording patients’ preferred place of care or death is associated with the actual place of death among patients followed by home palliative care teams. Methods: We conducted a retrospective observational study including adult patients who died in 2022 and were followed by a home palliative care team in Madrid, Spain. Sociodemographic and clinical variables, recorded preferred place of care or death, and actual place of death were extracted from electronic health records. Associations were analysed using bivariate tests and multivariable logistic regression. Results: A total of 464 patients were included (53% women; mean age 80.8 years). Overall, 82.5% died at home. A preferred place of care or death was recorded for 64% of patients; among them, 97.6% expressed a preference for home, and 89% of these patients died at home. In the multivariable analysis, older age and female sex were independently associated with death at home. Recording a preferred place of care or death was not independently associated with the place of death. Conclusions: Most patients followed by home palliative care teams died at home. Older age and female sex were associated with a higher probability of home death. Although most patients with a recorded preference for home died at home, recording preferences alone was not independently associated with the place of death. Systematic documentation of preferences may support advance care planning and patient-centred decision-making.

Abstract: Background/Objectives: Research examining the emotional and psychological challenges experienced by couples undergoing in vitro fertilisation (IVF) remains limited. Existing evidence suggests that women undergoing IVF often report elevated levels of depression, anxiety, and emotional distress, while men may experience feelings of anger, inadequacy, and self-doubt, especially following unsuccessful treatment cycles. Successful IVF outcomes are commonly associated with intense joy, relief, and fulfilment, as couples realise their aspiration to become parents. Given the limited qualitative research in Greece, the present study aimed to explore the lived experiences of couples undergoing IVF treatment, with particular attention to emotional, relational, and systemic dimensions. Methods: A qualitative research design was employed. Semi-structured, in-depth interviews were conducted with six heterosexual couples (aged 18–49 years) residing in Athens and Karditsa, Greece, all of whom had undergone IVF treatment. Interviews were audio-recorded, transcribed verbatim, and analyzed using Interpretative Phenomenological Analysis. Results: Analysis revealed five interrelated superordinate themes with associated subordinate themes: (1) attitudes towards infertility and IVF, (2) the impact of IVF on couple relationships, (3) the IVF experience, (4) challenges related to healthcare and the insurance system, and (5) expectations for the future. Lived experiences of infertile couples undergoing IVF treatment, highlighted a range of emotions, social pressure, and attitudes towards IVF and related policies. Conclusions: In Greece, where declining birth rates are increasingly prominent, IVF has gained societal and policy attention as a potential solution. Although IVF constitutes a demanding psychological journey, it represents a hopeful pathway for couples striving to achieve parenthood.

Abstract: This paper examines the Three-Dimensional Discrete Cosine Transform (3D-DCT), an extension of the widely used 1D- and 2D-DCT families that underpin modern audio, image, and video compression standards. Although extensively studied in theory, the 3D-DCT remains far less explored in practical coding systems. In this work, we develop a simple yet complete 3D-DCT encoder and investigate its performance in several application domains. By stacking video frames into 3D blocks and applying cubic or non-cubic 3D-DCT transforms, we construct a video coder that is significantly simpler than MPEG-x/H.26x methods while achieving comparable compression efficiency. The proposed approach is also well-suited for video editing scenarios, where small GOP sizes and the absence of motion dependencies provide improved frame-level accessibility. In addition, we evaluate the use of 3D-DCT for compressing volumetric medical images (CT studies), demonstrating strong quality and compression performance. Since 3D-DCT naturally applies to any 3D dataset, the method is likewise applicable to MRI or LIDAR volumetric data. Overall, the results show that 3D-DCT offers a versatile and computationally simple alternative for both video compression and 3D image coding.

Abstract: A digital pre-distortion (DPD) scheme for concurrent multi-beam transmission in fully digital multiple-input, multiple-output (MIMO) systems, using fast Fourier transform (FFT) beamforming and so-called virtual array processing, is proposed. In a MIMO array with nonlinear power amplifiers (PAs), transmitting multiple beams concurrently yields intermodulation products that end up in both user and non-user directions. In the setting with few users in a large array, the array dimension will typically be much larger than the number of generated intermodulation products. At the same time, linearization per-PA is excessively costly for large arrays. This work shows that is instead possible to linearize the system by producing predistorted user beams, and non-user intermodulation products, through DPD processing in a virtual array, of a much smaller dimension than the physical array. Theoretical derivations and simulation examples show how this approach can lead to manyfold reductions in DPD complexity.

Abstract: Accurate determination of heavy metals in marine hydrate-associated muds is crucial for tracing methane seepage, yet it faces challenges from complex matrices. This study developed a matrix-matched microwave digestion ICP-OES method. By comparing XRF spectral profiles and statistical tests, the feasibility of using soil certified reference materials to simulate marine mud matrices was demonstrated, thereby optimizing digestion parameters (power/temperature) and ICP-OES spectral line selection. Method validation revealed detection limits of 0.0004 – 0.0105 mg/L for Cu, Zn, Pb, and Cr, with spike recoveries ranging from 95.5% – 103.7%. Accuracy was further verified using soil reference materials (GSS-4/5) and comparative t-tests with ICP-MS data. This efficient and reliable method provides a practical analytical tool for geochemical exploration of marine gas hydrates.

Abstract: Electricity service navigation in Bangladesh is hindered by opaque billing, complex complaint resolution, and limited access to information, often requiring inefficient manual processes. To address these challenges, this paper introduces TaritBandhu, a hybrid AI and database-driven service system designed to streamline customer support for Bangla-speaking users. The system features a three-tier architecture comprising an Interface Layer for multimodal interaction (text and voice), a Logic Layer that orchestrates AI-driven query resolution and deterministic complaint matching, and a Data Layer that grounds responses in user-specific billing records and historical complaint logs. TaritBandhu employs a Bangla-first, voice-integrated conversational model, leveraging the Bangladesh Government's Speech APIs for inclusive access. It utilizes a Large Language Model (GPT-4.1) for generating contextual responses to general queries and a TF-IDF-based semantic matching algorithm to map user complaints to pre-existing solutions. A key innovation is its hybrid automation model, which escalates unresolved or complex issues to human agents via a token-based queuing system, managed through an admin panel for dynamic content control. While initial implementation demonstrates the system's viability, limitations concerning large-scale data handling, conversation context length, and pending real-world deployment are acknowledged. TaritBandhu presents a scalable, locally adapted framework that balances AI automation with human oversight, aiming to enhance transparency, accessibility, and efficiency in utility customer service.

Abstract: Modern science explains how structures evolve once space, time, and dynamical laws are assumed. We ask a prior question: which geometric forms are admissible as the first coherent differentiations of a maximally symmetric (isotropic), undifferentiated state? The analysis is deliberately pre-dynamical and pre-physical: no temporal evolution, field equations, energetics, or mechanism are assumed.“Zero” is interpreted operationally as nondifferentiation (maximal isotropy), not emptiness and not a physically extant point. The sphere appears only as a symmetry object encoding “all directions are equivalent” once differentiation is contemplated. Coherence is a closure-compatibility constraint (standing-wave–like only in the sense of global consistency under closure).Under isotropy-preserving closure and minimality, continuous differentiation is disfavored and a finite set of extrema is forced. The smallest non-degenerate configuration requires four extrema; imposing single-scale maximal symmetry uniquely selects the regular tetrahedron (up to rotation). Minimal conjugate completion yields the star tetrahedron, and the cube/octahedron arise as induced envelopes. We record admissible extension pathways toward packing and Voronoi structure (preview only), deferring substrate selection to later work. The results are admissibility claims conditional on explicit postulates, not assertions of physical necessity.

Abstract: Myeloid sarcoma (MS) is a malignant extramedullary tumor that occurs in patients with acute myeloid leukemia (AML), myelodysplastic syndrome (MDS), or chronic myeloid leukemia (CML). The standard first-line treatment for MS is intensive chemotherapy according the AML protocol, regardless of bone marrow involvement. The role of allogeneic hematopoietic stem cell transplantation (alloHSCT) in the treatment of pediatric patients with MS requires further investigation. The aim of the study was to evaluate treatment outcomes for MS in pediatric patients with a focus on assessing the impact of allogeneic hematopoietic stem cell transplantation (alloHSCT) on treatment efficacy. The study included 64 patients aged 0 to 19 years from 15 pediatric oncology centers in Poland, who were diagnosed with MS between 1998 and 2024. The probability of 5-year overall survival (pOS) for the entire cohort was 0.63 ± 0.07, while the 5-year event-free survival (pEFS) and 5-year relapse-free survival (pRFS) were 0.62 ± 0.07 and 0.72 ± 0.07, respectively. Treatment outcomes were compared between patients who underwent allogeneic hematopoietic stem cell transplantation (alloHSCT) in first complete remission (ICR) (n1 = 17/64; 27%) and those who did not receive alloHSCT (n2 = 47/64; 73%). In the alloHSCT group (n1), the estimated survival probabilities were pOS = 0.49 ± 0.13, pEFS = 0.44 ± 0.14, and pRFS = 0.40 ± 0.14. In the non-alloHSCT group (n2), these values were pOS = 0.68 ± 0.08, pEFS = 0.68 ± 0.08, and pRFS = 0.84 ± 0.06. The difference in pRFS between groups n1 and n2 was statistically significant (p = 0.0049). Extramedullary relapse was more frequently observed in patients who had undergone allogeneic hematopoietic stem cell transplantation (alloHSCT) (p = 0.0001). Further research is needed to identify effective strategies for sustaining remission in patients with MS after alloHSCT.

Abstract: The emergence and persistence of life pose a profound paradox. Statistical estimates of abiogenesis under standard prebiotic models yield extremely low probabilities (10⁻⁷⁸–10⁻¹⁰⁰), although such values are strongly model‑dependent and do not constitute evidence against naturalistic origins. Rather, they highlight a gap between current physical chemistry and the observed robustness of biological organization. Here we propose that both phenomena can be explained by the action of a hitherto unobserved informational reservoir that subtly “leaks” into biological systems, biasing microstate probabilities in real time. While quantum coherence and nonlocality currently represent the most plausible physical substrates, the hypothesis deliberately remains agnostic about the ultimate origin of this reservoir. Crucially, the transfer need not be intentional; it may constitute an unintended “crosstalk” across an ontological boundary—analogous to sound leaking through a wall between apartments. This framework offers a strictly naturalistic alternative to intelligent design theories while generating falsifiable predictions distinguishable from both pure chance and directed panspermia.

Abstract: Brain size correlates weakly with intelligence within species yet strongly across species, and several taxa—from corvids to honeybees—exhibit cognitive abilities that appear disproportionate to their brain mass. The Strong Electromagnetic Field Hypothesis (SEFH) proposes that consciousness and higher cognition emerge from hierarchically nested electromagnetic (EM) field dynamics in neural tissue, with neural firing serving primarily as an energy source for these fields rather than as the primary computational medium. This framework generates specific, quantitative predictions based on two variables: (i) wattage density—the EM field production intensity per unit volume of integrative tissue, driven by neuron density—and (ii) harmonic capacity—the number of distinct geometric eigenmodes (resonant standing-wave patterns) that the field-permeable tissue can sustain, determined by the geometry and volume of the ephaptically coupled neural medium. We systematically test these predictions by mining existing comparative neuroscience datasets, including isotropic fractionator studies of cortical/pallial neuron counts and densities across primates, corvids, parrots, cetaceans, elephants, carnivores, rodents, and invertebrates (honeybees). After excluding cerebellar neurons (which serve motor control, not integrative cognition), we calculate estimated EM field production density (watts per cubic centimetre) for associative tissue across taxa. We find that SEFH predictions are strongly confirmed in several key comparisons: corvids and parrots achieve primate-rival cognition with 3–5× higher pallial wattage density than human cortex; honeybees achieve remarkable cognitive feats with the highest neural density measured in any animal (~960,000 neurons/mg); and elephants dramatically underperform their total neuron count when cerebellar (motor control) neurons are excluded. Drawing on recent work showing that brain geometry—rather than connectome topology—fundamentally constrains neural dynamics (Pang et al., 2023) and that harmonic brain modes govern spatiotemporal dynamics of cognition and consciousness (Atasoy et al., 2016, 2018), we propose a two-variable predictive model: cognitive capacity ∝ wattage density × log(harmonic capacity). A honeybee’s mushroom body is an exquisitely tuned tiny drum—remarkable domain-specific performance from a handful of harmonic modes—while the human cortex is a cathedral, sustaining thousands of resonant modes across its vast field-permeable geometry. This framework accounts for cross-species cognitive patterns better than any single neural measure and, unlike models built on the McCulloch–Pitts neuron-as-logic-gate framework, is fully native to field-based physics. A preliminary cross-species regression using these two variables explains R² = 91.8% of cognitive variance across ten focal taxa (Spearman ρ = 0.976, p < 0.00001), compared with 39.2% for brain mass and 64.8% for neuron count alone.

Abstract: This study presents a comprehensive framework combining finite element analysis, machine learning, and generative AI for aluminum cold spray deposition analysis. Abaqus explicit dynamic simulations modeled high-velocity particle impact at 700 m/s, capturing stress tensor components and von Mises equivalent stress distributions. The maximum von Mises stress of 537.73 MPa exceeded aluminum yield strength by 3.6 times, confirming successful deposition through severe plastic deformation. Three machine learning algorithms were trained on stress tensor components (S11, S22, S33, S12, S13, S23) to predict von Mises stress. Random Forest, Gradient Boosting, and Neural Network models achieved exceptional accuracy with R² values of 0.9975, 0.9955, and 0.9922 respectively. Hyperparameter optimization further improved performance to R² = 0.9977, 0.9887, and 0.9985. Feature importance analysis identified S22 transverse stress as the dominant predictor with 80% importance. Google Gemini generative AI provided engineering insights confirming bonding mechanisms through adiabatic shear instability and oxide disruption. Process optimization recommendations addressed velocity control, particle distribution, and substrate preparation. This integrated approach enables rapid stress prediction and intelligent process optimization for industrial cold spray applications.

Abstract: The proliferation of marine data is an opportunity for ocean governance and contributes to fragmentation in the disciplines, institutions, and sectors. Marine Spatial Data Infrastructure (MSDI) stands out as the major framework in integrating marine information; however, its intellectual framework and development are not well synthesised. The research applies the two-step systematic review and bibliometric analysis of Scopus-indexed literature (2000-2024) to trace the trends in publications, collaboration patterns, thematic cohesion, and time-related changes in MSDI research. Results suggest that the MSDI scholarship is growing faster,, with most of the products being European-made,, with policy frameworks like INSPIRE and the Marine Strategy Framework Directive leading the pack. It is divided into four pillars of themes, namely technical implementation, governance and policy, data management, and stakeholder applications. This development of MSDI can be characterised by five consecutive stages: fundamental technical standardisation, the implementation of the model of governance, semantic interoperability improvement, the integration of the policy, and the sophisticated application of the principles of FAIR/CARE and AI. The paper concludes that MSDI is moving to a more socio-technical approach that requires consideration of a technical-focused tool in the present-day ocean governance. In the future, combining semantic AI, decentralised architectures, polycentric governance models, and impact assessment frameworks to align the MSDI development with the objectives of equity, inclusion, and sustainability should be considered.

Abstract: Vitamin D is traditionally recognized for its role in calcium homeostasis and skeletal health; however, growing molecular and clinical evidence indicates that vitamin D signaling is a central regulator of biological barrier integrity across multiple organ systems. Epithelial and endothelial barriers—including the intestinal mucosa, vascular endothelium, blood–brain barrier, pulmonary epithelium, renal filtration barrier, and cutaneous barrier—depend on intact tight junctions, adherens junctions, and immune–redox homeostasis to maintain systemic health.We propose that vitamin D deficiency may contribute to a unifying pathophysiological state characterized by multi-barrier dysfunction, recently conceptualized as Systemic Leaky Barrier Syndrome (SLBS). Through regulation of junctional proteins (e.g., claudins, occludin, ZO-1), modulation of innate and adaptive immunity, suppression of chronic inflammation, and maintenance of redox balance, vitamin D plays a pivotal role in preserving barrier resilience. Failure of these protective mechanisms promotes translocation of microbial products, inflammatory mediators, and metabolic toxins, driving chronic diseases including cardiovascular disease, neurodegeneration, autoimmune disorders, cancer progression, and metabolic dysfunction.This review synthesizes mechanistic, translational, and clinical evidence supporting vitamin D as a barrier-protective hormone and positions SLBS as a systems-level framework for understanding the broad disease consequences of vitamin D insufficiency.

Abstract: Technical indicators derived from historical price data have long been central to quantitative trading strategies, yet their actual contribution to modern deep learning forecasting models remains an open empirical question. This study presents a large-scale ablation analysis examining whether technical indicators improve next-day price prediction when used as inputs to recurrent neural networks. We conduct 500 controlled experiments across 10 assets spanning five asset classes—commodities (Crude Oil, Gold), cryptocurrencies (Bitcoin, Ethereum), equities (Apple, Microsoft), foreign exchange (EUR/USD, USD/JPY), and market indices (NASDAQ, S&P 500)—using daily OHLCV data from 2010 to 2025. Five feature configurations are evaluated: a raw OHLCV baseline and four indicator-augmented variants incorporating momentum (RSI, Stochastic Oscillator), trend (SMA, EMA, ADX, MACD), volatility (ATR, Bollinger Bands), and a combined all-indicator set. Each configuration is tested with both LSTM and GRU architectures across five random seeds to ensure statistical robustness. Our results show that technical indicators do not improve—and frequently degrade— forecasting performance relative to raw price data. The baseline OHLCV configuration achieves the lowest mean RMSE (0.166 ± 0.148) and highest mean directional accuracy (55.7% ± 5.5%). Every indicator-augmented configuration produces higher prediction error, with the comprehensive all-indicators variant exhibiting statistically significant degradation (34.6% RMSE increase, p < 0.001, Cohen’s d = −0.29). All four indicator categories show significant performance reduction at α = 0.05. GRU models achieve marginally higher directional accuracy than LSTM (55.3% vs. 51.0%), although RMSE differences between the two architectures are not statistically significant (p = 0.846). Foreign exchange stands out as the only asset class where volatility indicators improve performance (4.2% RMSE reduction), while high-volatility assets (cryptocurrencies, commodities) exhibit 83% higher mean prediction error than their low-volatility counterparts. These findings suggest that deep recurrent architectures implicitly learn the patterns captured by conventional technical indicators, making explicit indicator features redun- dant or even harmful. The results carry practical implications for feature engineering in neural network-based trading systems and highlight the importance of rigorous baseline comparisons in applied financial machine learning.