Introduction. Knowledge of orbital anthropometry is essential for reconstructive surgery, ophthalmology and forensic identification, yet normative multislice computed tomography (MSCT) data for this region are scarce. The aim was to determine the principal orbital parameters in adults without orbital pathology and to examine sex, age and side differences. Method. In this retrospective study, 154 orbits in 77 adults (51 men, 26 women; aged 24-86 years) were measured on three-dimensional reconstructions from a 64-slice MSCT scanner. Orbital height, width, volume, orbital index and the interorbital and biorbital distances were recorded and compared by sex, age group and side. Results. Men had significantly greater orbital width, volume and biorbital and interorbital distances than women bilaterally (p ≤ 0.009), whereas orbital height did not differ by sex. The right orbital index was higher in women (90.02 ± 6.45 versus 86.36 ± 7.26; p = 0.034). In the older group the right orbit showed greater height, volume and orbital index (p ≤ 0.031), with no significant age effect on the left. The right orbit was significantly wider and had a higher orbital index than the left (p ≤ 0.011), while orbital volume was symmetrical (p = 0.099). Conclusion. Orbital morphometry is sex-, age- and side-dependent, and the MSCT-derived values reported here provide population-specific reference data for clinical and forensic use.

Radiological imaging modalities, particularly computed tomography (CT) and magnetic resonance imaging (MRI), play a pivotal role in the diagnosis of orbital diseases. CT remains the gold standard for the evaluation of orbital trauma, including fractures, calcifications, and metallic foreign bodies, whereas MRI is superior for assessing soft tissues and the optic nerve. Characteristic radiological signs include the “I’M SLOW” pattern in Graves’ ophthalmopathy, the “tram-track” sign in optic nerve sheath meningioma, the “teardrop” sign in blowout fractures, and the “empty delta” sign in cavernous sinus thrombosis. An apparent diffusion coefficient (ADC) threshold value of 0.84 × 10⁻³ mm²/s has been shown to differentiate malignant from benign orbital lesions. The presence of a metallic intraocular foreign body represents an absolute contraindication to MRI examination. Artificial intelligence is increasingly contributing to orbital imaging through automated muscle segmentation, fracture detection, and prediction of tumor histology. Appropriate selection of imaging modality and recognition of these characteristic findings significantly improve diagnostic accuracy.

This study investigated the distribution of 40K, 137Cs and 90Sr during the processing of cow milk into cottage cheese from central Bosnia and Herzegovina and Livno cheese from the southwestern production area, with particular emphasis on physicochemical composition and regional differences. Raw milk (n = 15 per group), corresponding cheeses (n = 15 per group), and composite whey samples (one per cheese type) were analyzed for protein, fat, dry matter, water content, 40K, and 137Cs, as well as titratable acidity in milk samples. In a random second-stage subset, because of the cost and complexity of the radiochemical method, 90Sr was determined. Mean activity concentrations in milk used for cottage and Livno cheese production were 42.31 and 40.77 Bq/L for 40K, 0.030 and 0.057 Bq/L for 137Cs, and 0.015 and 0.030 Bq/L for 90Sr, respectively. Corresponding cheese means were 26.45 and 28.23 Bq/kg for 40K, 0.038 and 0.087 Bq/kg for 137Cs, and 0.073 and 0.369 Bq/kg for 90Sr. Processing retention factors showed limited transfer of 40K and 137Cs into cheese, whereas 90Sr was retained mostly in the curd. Significant regional differences were observed for 137Cs in both milk and cheese, whereas 40K did not differ significantly between regions. Differences between regions for 90Sr were also indicated, but these results should be interpreted cautiously because it was analyzed only in a limited second-stage subset. These findings show that radionuclide behavior in artisanal dairy systems is jointly shaped by radionuclide chemistry, cheese technology, and local environment, highlighting the need for site-specific food-chain assessment.

This paper examines the origins and current frontiers of behavioural macroeconomics, using expectation formation as a conceptual framework linking early psychological insights to modern formal models. The review traces developments from animal spirits, bounded rationality and adaptive expectations to heterogeneous heuristics, sticky information, rational inattention and behavioural New Keynesian models. Its main contribution is a comparative assessment of these approaches in terms of their underlying mechanisms, empirical predictions, identification challenges and implications for monetary and fiscal policy. Particular attention is given to heterogeneity stemming from different economic circumstances and to behavioural departures from rational expectations. The analysis shows that competing approaches can explain similar macroeconomic patterns yet imply different causal mechanisms and policy responses. Future progress depends on direct model comparisons, causal identification, out-of-sample validation and the integration of survey, experimental, administrative and high-frequency data into theoretically disciplined models.

Speculative decoding losslessly accelerates large language model inference by having a lightweight draft model predict future tokens for verification by the target model. Recent block diffusion drafters further reduce drafting latency by predicting multiple tokens in parallel. However, existing block drafters project target hidden states at every input position into a separate drafter-side KV cache, incurring per-request memory and KV-write overhead that grow with concurrency; directly reusing target KVs in place removes this cache but fails to sustain draft quality throughout the block. We propose a hybrid target-context injection method that complements direct target KV reuse with target hidden states only at the last input position, requiring no separate drafter-side KV cache. Building on this design, we propose H-Spec, a hybrid Mamba-attention parallel drafter that consumes the two target-context sources through complementary modules. Mamba modules are initialized with projected last-token target hidden states, while attention modules reuse target KVs in place. Despite its recurrent formulation, Mamba's parallel scan allows H-Spec to preserve block-parallel drafting. Across three target models and diverse tasks, H-Spec improves over the best baseline by 5.0--13.3% in mean accepted length and 5.3--12.6% in batch-size-1 inter-token latency speedup. Under concurrent serving, H-Spec consistently achieves higher throughput while maintaining lower KV cache utilization than baselines across evaluated concurrency levels.

Real-time monitoring of safety-critical interior states remains an open problem across energy, environmental, and industrial systems where direct instrumentation is infeasible. Existing approaches based on governing equations, discrete state vectors, or fixed sensor locations fail to provide mesh-independent, field-level reconstructions at arbitrary interior coordinates under real-time constraints. Here we introduce neural operator-based virtual sensing as a general framework for recovering inaccessible interior fields from sparse, heterogeneous measurements and operating conditions, instantiated with MIMONet, a multi-input, multi-output neural operator that fuses heterogeneous inputs and decodes coupled physical fields through a shared latent representation. Across three engineering-grade evaluations of escalating complexity, from confined recirculating flows to pressurized water reactor subchannels and compact power-system heat exchangers, MIMONet achieves ≤5% relative reconstruction error, while generating full-field predictions with sub-millisecond inference on an NVIDIA H200, at least five orders of magnitude faster than traditional high-fidelity CFD simulations, with calibrated uncertainty and noise resilience demonstrated on the heat-exchanger case. We validate the framework on three independent real-world datasets spanning electrochemical energy, atmospheric science, and physical oceanography: current-density mapping in a hydrogen fuel cell, wind-speed profiling on instrumented meteorological towers, and North Atlantic ocean-state from a global ocean reanalysis. MIMONet improves over classical virtual-sensing baselines by 72–84%: 84% on fuel-cell internal-field recovery from 10% sensor coverage and 83% on hub-height wind-speed recovery from lower-level tower sensors. Most notably, it recovers sea-surface height, a field with zero direct sensors, from its learned coupling to observed temperature and salinity at 4.1 × lower error than the best classical predictor, a cross-modal field recovery beyond the reach of classical interpolation. These results establish operator-based virtual sensing as a practical route to real-time field observability in systems where the states that matter most cannot be directly measured. The authors introduce an operator-learning framework for real-time virtual sensing that reconstructs inaccessible interior physical fields from sparse, heterogeneous measurements and operating conditions.

Exercise is increasingly positioned as an adjunct to oncology care, yet its clinical benefits and proposed antitumour mechanisms are often discussed in parallel rather than as a unified biological model. This focused mini review integrates 20 unique full-text sources spanning meta-analyses, systematic and narrative reviews, consensus guidance, and translational studies. The evidence supports a multiscale framework in which muscle contraction generates acute neuroendocrine and myokine signals, improves tumour perfusion and oxygenation, alters metabolite competition within the tumour microenvironment, mobilises cytotoxic immune cells, and preserves host muscle and functional reserve. Clinical evidence is strongest for improvements in cardiorespiratory fitness, strength, fatigue, physical function, and quality of life, whereas direct tumour control and enhancement of immunotherapy remain supported mainly by preclinical and early human mechanistic data. Important tensions include the context-dependent biology of interleukin-6, the distinction between vascular normalisation and tumour-promoting angiogenesis, the uncertain optimal exercise dose, and the risk that excessive energy expenditure may be inappropriate in severe cachexia. We propose a precision exercise oncology model that matches modality and dose to tumour phenotype, treatment phase, symptoms, physiological reserve, and dynamic monitoring. Exercise should be treated as evidence-based supportive care with plausible disease-modifying potential, not as a substitute for anticancer therapy.

Background: Cancer remains one of the leading global public health challenges, with a growing number of patients living during and after oncological treatment. Physical activity has increasingly been recognized as an important non-pharmacological strategy in cancer prevention, treatment support, rehabilitation, and survivorship care. Objective: The aim of this review is to examine the impact of physical activity on cancer prevention, treatment-related outcomes, and quality of life among individuals diagnosed with cancer. Methods: A structured literature search was conducted in Web of Science, PubMed, and Scopus databases. The search included studies published in English and focused on physical activity, exercise interventions, cancer prevention, cancer treatment, cancer-related fatigue, survivorship, and quality of life. Systematic reviews, meta-analyses, randomized controlled trials, cohort studies, and clinical guidelines were considered. Results: Previous evidence suggests that regular physical activity is associated with a reduced risk of several cancer types, especially breast, colorectal, and endometrial cancers. During and after treatment, exercise interventions may improve physical functioning, reduce cancer-related fatigue, support mental health, and enhance health-related quality of life. Conclusion: Physical activity represents a safe and effective supportive strategy across the cancer continuum. Its integration into oncology care may contribute to prevention, better treatment tolerance, rehabilitation, and improved quality of life among cancer patients and survivors.

While conversational AI agents (CAs) have the potential to support drivers through hands-free and eyes-free interaction, their role in long-haul trucking remains unexplored. This paper investigates how long-haul truck drivers perceive CAs and identifies design considerations for their successful deployment. We conducted ride-along observations and semi-structured interviews with five long-haul truck drivers during regular driving operations. The results show that drivers today regulate verbal communication according to driving demands and expect communication partners to adapt to changes in workload and attention. Participants were most receptive to CAs supporting driving-related tasks, including traffic information, rerouting and route planning. Trust, perceived complexity, and discomfort with conversational interaction emerged as key barriers to adoption. We conclude that CAs may be best suited to supporting strategic and selected tactical tasks, while being less appropriate for operational tasks. Effective CA design must adapt to context and workload, support interruptions, and provide transparent and trustworthy support.

Current AI-based conversational agents (CAs) largely lack situational awareness, limiting their ability to understand driving context, driver state, and immediate driver needs. Consequently, the use of CA may increase drivers’ cognitive load and impair safety and user experience. This study investigates the role of situationally aware CA in long-haul trucks and how such CA should be designed to support drivers in an efficient way. The insights were gathered through a Wizard-of-Oz driving simulator study exploring two CA concepts. The results show that CA are well-suited for strategic tasks, could be useful for some tactical tasks depending on the cognitive load, and are generally inappropriate at the operational level, where they interfere with core driving actions. Effective CA design must prioritize driving as the primary task, adapt interaction timing to drivers’ cognitive load, and ensure transparency and predictability to build trust.

Annotated medical imaging data remain scarce, and labels are often weak and noisy: in mammography, an examination comprises several high-resolution views, yet the diagnostic outcome is recorded only at the breast level. Such problems are naturally cast as Multiple Instance Learning (MIL), where the model must infer instance-level structure from bag-level labels alone. Although contemporary foundation encoders supply strong general-purpose embeddings, augmenting MIL data in this representation space remains an open problem as established techniques act on one instance at a time and ignore the statistical dependencies binding a bag together. We address this with SetFlow, a generative model that learns the distribution of complete MIL bags directly in a frozen encoder’s embedding space. SetFlow couples flow-matching training with a Set Transformer-inspired backbone, making it invariant to instance ordering while modeling intra-bag relationships. Generation is conditioned jointly on class label and per-instance scale, yielding coherent, semantically faithful bags rather than isolated vectors. Evaluating on two large public mammography datasets and two encoders, we assess distributional fidelity, nearest-neighbor behavior, and downstream augmentation utility. We show that generated bags reproduce real-data statistics and improve classification in certain configuration, with performance gains varying on the amount of synthetic data. An architecture ablation confirms each design choice contributes to performance.

Coal remains an important component of the electricity system of Bosnia and Herzegovina, while the country is simultaneously undergoing a gradual energy transition toward lower-carbon technologies. This study evaluates the quality of coal from major mining areas in Tuzla Canton using laboratory results for composite samples from Kreka, Đurđevik and Banovići. The principal indicators were total moisture, ash, combustible matter and lower heating value (LHV). Eight mining areas were compared using five analytical samples per area, giving 40 observations for the primary statistical analysis. An additional 15 observations from five Banovići separation products were evaluated separately. Marked differences were identified among mining areas. Mean LHV ranged from 7,836 kJ/kg at PK Dubrave to 19,794 kJ/kg at Jama Omazići. Lignite from the Kreka area had substantially higher mean total moisture than the analysed brown coals, while several brown-coal areas were constrained by high ash content. Across the 40 primary observations, LHV showed a moderate negative correlation with total moisture (r = -0.43) and with ash (r = -0.37). A multiple linear model using moisture and ash explained approximately 98.4% of the observed LHV variation, although the small number of samples and the compositional nature of the variables require cautious interpretation. The findings demonstrate that coal reserves alone are insufficient for evaluating the future role of coal: fuel quality, preparation efficiency, plant efficiency, emissions intensity, economic performance and regulatory conditions must be considered together. The results support a transition strategy that combines improved coal characterisation and combustion efficiency in the medium term with accelerated deployment of renewable generation, storage, grid modernisation and just-transition measures.

Using on-shell methods, we derive an all-order non-renormalization theorem for general four-dimensional effective field theories, relying on Poincar\'e invariance and unitarity. Each operator $\mathcal O$ is assigned the weights $(w,\bar w)=(\ell-h,\ell+h)$, where $\ell$ and $h$ are the number of particles and total helicity of its minimal configuration. At $L$ loops, the mixing $\mathcal O_j\to\mathcal O_i$ vanishes when $\ell_j-\ell_i=L-1$ and either $w_i<w_j-2(L-1)$ or $\bar w_i<\bar w_j-2(L-1)$, in the absence of non-holomorphic Yukawa couplings. These zeros follow entirely from tree-level data and are scheme independent under finite renormalizations that preserve the operator-length selection rule. For operators of dimensions five through eight, we identify broad classes of previously unknown zeros at higher loop orders, with particular emphasis on two loops, where the theorem provides direct checks on current calculations, as in the Standard Model Effective Field Theory.

This paper evaluates the energy performance and lifecycle environmental impact of three retrofit pathways—conventional insulation, an extensive green roof, and rooftop photovoltaics—for two lamella-type residential buildings (three-story, five-story; over 40 years old) in Blocks 45 and 70, Novi Beograd, Serbia. The study builds on an architectural concept treating the flat roof as an active environmental surface. Energy demand was quantified via calibrated dynamic simulation for a baseline and three scenarios—insulation (S1), insulation with an extensive green roof (S2), and insulation with rooftop photovoltaics (S3)—with environmental performance assessed using cradle-to-use LCA (1 m2 heated floor area, 50-year period) covering global warming potential, cumulative energy demand, and Environmental Footprint 3.1 indicators. Insulation and the green roof achieve nearly identical reductions in heating demand (27–31%) and lifecycle GWP/CED (10–13%), the green roof performing marginally better while also offering evapotranspirative cooling and urban co-benefits. Photovoltaics deliver the largest lifecycle GWP reduction (26% three-story, 15% five-story) and cut acidification by up to 36%, but increase mineral resource depletion, water and land use. Benefits of both roof-level measures scale with roof-to-floor-area ratio, favoring lower-rise buildings, supporting prioritization of green roof and photovoltaic retrofit across the district lamella stock.