Abstract: This study looks into the predictive performance of linear econometric and deep learning methodologies for the South African unemployment rate quarterly data. In this paper, the Autoregressive Integrated Moving Average with exogenous variables (ARIMAX) model was compared to the Long Short-Term Memory (LSTM) network using unemployment rate quarterly data. Exploratory Data Analysis (EDA) suggested that the unemployment rate series is non-stationary, with structural breaks around 2020 and time-varying volatility. Stationarity tests established the need for differencing, whereas diagnostic tests revealed the presence of autocorrelation and ARCH effects in the raw data. The ARIMAX model added labour market covariates, and the differenced Not Economically Active (NEA) variable was statistically significant, whereas Discouraged workers were not. Although the ARIMAX model provided a good in-sample fit, residual diagnostics showed deviations from normality. Out-of-sample forecast study revealed moderate predictive accuracy, with relatively substantial forecast errors and increasing prediction intervals over time. In contrast, the LSTM model showed significant learning capacity, with early convergence and well-behaved residuals that meet both independence and homoskedasticity criteria. The model achieved significantly lower forecast errors, with RMSE, MAE, and MAPE values much lower than those of the ARIMAX model. Comparative forecast analysis using Diebold-Mariano (DM) test and model confidence Set (MCS) method and bootstrap confidence intervals consistently demonstrated the statistical superiority of the LSTM model. The findings give strong evidence that the LSTM model outperformed the ARIMAX model for projecting South African unemployment rate. The findings emphasise the importance of nonlinear modelling approaches in capturing complex labour market dynamics while also demonstrating the limitations of classic linear models. These findings also emphasise the importance of using nonlinear machine learning algorithms in macroeconomic forecasting.

Abstract: Introduction: In hemodialysis patients, Body Mass Index is insufficient in assessing their nutritional status due to the ‘obesity paradox’ and the impact of body composition on inflammation. The aim of the study was to assess the relationship between body composition, traditional inflammatory markers, and the new NETosis indicators (neutrophil extracellular traps), as well as to determine their impact on 12-month mortality. Methods: The study included 99 patients with end-stage renal disease. Their body composition was assessed using bioelectrical impedance analysis (Seca mBCA 525). Blood serum was tested for inflammatory markers (hs-CRP, IL-6, TNF-α, IL-1B), NETosis markers (citrullinated histone H3, MPO, elastase), and nutritional status parameters (albumin, transferrin). Results: No correlation between BMI and inflammation was demonstrated. Higher contents of the adipose tissue, particularly visceral, were significantly associated with increased levels of IL-6 and hs-CRP, while muscle mass was negatively correlated with inflammation. The use of dialysis catheters stimulated NETosis (higher CH3 levels), which had a negative effect on albumin concentrations. Low albumin levels and high TNF-α levels were independent predictors of death. Conclusions: It is body composition, and not BMI, that determines the severity of inflammation. Visceral obesity promotes inflammation, while muscle mass has a protective effect. Dialysis catheters, by stimulating NETosis, contribute to a decrease in albumin levels and a poorer prognosis.

Abstract: Genomic prediction is now routine in crop improvement, but its main bottleneck has shifted from marker density to environmental complexity. Breeders rarely need predictions for one fixed environment; they need to rank genotypes across target populations of environments that differ in weather, soils, management, and stress timing. This makes genotype-by-environment interaction a primary breeding problem rather than a secondary statistical nuisance. This review examines how genomic, environmental, and phenomic information can be integrated to improve multi-environment prediction in crop breeding pipelines. The review is narrative rather than PRISMA-style, but the literature search and selection logic were structured and explicitly defined. Peer-reviewed English-language studies were identified through structured searches of Web of Science Core Collection and Scopus, supplemented by backward citation screening, with emphasis on literature published from January 2023 to March 2026. Four conclusions emerge. First, environmental information is most useful when it is developmentally aligned, biologically interpretable, and matched to the target population of environments. Second, strong structured statistical baselines remain highly competitive, especially in moderate-sized or highly unbalanced datasets, whereas gains from more flexible machine-learning and deep-learning approaches are most evident in large, sparse, heterogeneous, and multimodal settings. Third, phenomic markers often improve prediction for complex traits, especially yield, because they capture realized crop responses not fully represented by markers alone. Fourth, practical value depends less on isolated gains in predictive accuracy than on evaluation under realistic deployment scenarios, including untested genotype and untested environment settings. Progress therefore requires transparent reporting, benchmark design, stage-aware envirotyping, multimodal integration, uncertainty reporting, and cost-aware deployment.

Abstract: Aerobic composting is an important pathway for the resource utilization of agricultural waste. However, nitrogen loss during composting not only reduces the nutrient value of the final product but also causes environmental burdens, particularly through ammonia (NH₃) volatilization and nitrous oxide (N₂O) emissions. This review critically examines the sources, pathways, and mechanisms of nitrogen loss during aerobic composting of agricultural waste, with emphasis on nitrogen transformation and the major loss routes, including NH₃ volatilization, N₂O emissions, and nitrate leaching. From a multiscale perspective, the review synthesizes control strategies spanning feedstock pretreatment (e.g., C/N ratio optimization, adsorbent amendment, and microbial inoculation), in-process regulation (e.g., aeration, moisture, temperature, pH), and post-treatment approaches for nitrogen stabilization and resource recovery. The supporting roles of reactor innovation, intelligent process control, and policy and regulatory measures are also discussed. Finally, current bottlenecks and future research directions are summarized from environmental and economic perspectives, with particular emphasis on interdisciplinary integration and technological innovation to enhance nitrogen retention during composting.

Abstract: Refugees are exposed to cumulative pre-migration, migration, and post-migration stressors that increase vulnerability to depressive disorders and impaired quality of life. Aim of this study was to assess the prevalence and severity of depressive symptoms among adult refugees in Serbia and associations with sociodemographic characteristics, traumatic experiences, social support, and Health Related Quality of Life (HQoL). This study included 324 refugees in four reception centers in Serbia. Data were collected between November 2022 and April 2023 using self-report questionnaires. Depressive symptoms were assessed using the Patient Health Questionnaire-9 (PHQ-9), while HQoL was evaluated using the SF-36 Health Survey. Sociodemographic, migration-related, and psychosocial variables were collected through a structured questionnaire. The mean PHQ-9 score indicated mild to moderate depressive symptomatology. Significant depressive symptoms were present in 41.4% of participants, while more than 70% reported mild symptoms. Depressive symptom severity was negatively correlated with energy/fatigue, emotional well-being, social functioning, general health, and pain. Energy/Fatigue emerged as the strongest independent predictor of depressive symptom severity. Depressive symptoms are highly prevalent among refugees and are closely associated with impaired quality of life and psychosocial stressors. These findings highlight the need for systematic screening and psychosocial interventions targeting mental health issues in refugees.

Abstract: Drinking water quality is essential for public health and requires monitoring approaches able to capture both regulatory compliance and short-term variability. This study presents a high-frequency IoT-based comparative physicochemical assessment of two drinking-water sources in Bragança, NE Portugal: treated municipal water derived from surface water and groundwater abstracted from a decentralized supply system. A low-cost IoT monitoring system was used to measure pH, electrical conductivity, temperature, oxidation-reduction potential, and total dissolved solids. Monitoring campaigns were conducted between January and March 2026 at two treated-water points within the public supply system and three groundwater points, complemented by municipal records from 2023 to 2025. The treated municipal supply showed a more stable physicochemical profile and lower variability, whereas groundwater was associated with higher mineralization and stronger temporal fluctuations. Significant differences were found for electrical conductivity, total dissolved solids, oxidation-reduction potential, temperature, and pH. High-frequency monitoring enabled the identification of dynamic patterns and transient fluctuations that would be difficult to detect through discrete sampling alone.

Abstract: Let Γ be a commutative with identity multiplicative hyperring and S ⊆ Γ be a multiplicatively closed subset of Γ. In this study, we will discuss the definition and general properties of weakly Γ-prime hyperideals. Let Ω be a hyperideal of Γ disjoint from S. We say that Ω is a weakly S− prime hyperideal of Γ if there exists an s ∈ S such that for all ϱ,σ ∈ Γ , if {0} ̸ = ϱ⋄σ ⊆ Ω then s⋄ϱ ⊆ Ω or s◦y ⊆ Ω . We will also examine the relationship between weakly−S prime hyperideals, weakly prime, and S−prime hyperideals.

Abstract: Against the backdrop of rural revitalization, traditional villages in Guizhou's ethnic minority regions face the dual challenges of preservation and development. Existing research predominantly focuses on macro-scale morphological descriptions, lacking an operable spatial classification method that can directly guide planning and management. To address this gap, this paper takes Fengxi Village in Dejiang County as a case study, integrates Conzenian urban morphology theory with the concept of "management units", and proposes a spatial unit classification method for traditional villages based on the overlay analysis of "morphological regions + property parcels". First, the Conzenian plan analysis method is employed to systematically deconstruct Fengxi Village's land use, road system, plot combinations, and building types, thereby delineating its morphological regions. Subsequently, three evaluation factors—building value, quality, and appearance—are innovatively introduced. Through quantitative evaluation, all 702 buildings in the village are classified into five categories: preservation units, restoration and improvement units, comprehensive renovation units, demolition and renewal units, and new development units, with the quantities and proportions of each unit type statistically analyzed. Building on this foundation, differentiated control guidelines and development strategies are proposed for each unit category. The research indicates that this method achieves a transformation from "morphological description" to "implementable control", breaking down the vague goal of "holistic preservation" into concrete "unit-based guidance" actions, and provides a replicable technical pathway for the refined planning and management of traditional villages. The innovation of this paper lies in constructing a complete technical framework of "morphological analysis - factor evaluation - unit control", addressing the deficiency of existing research at the micro-operational level.

Abstract: Upper respiratory tract infections such as the common cold are highly prevalent and impose a substantial health and economic burden, with many individuals relying on over-the-counter (OTC) medications for symptomatic relief despite limited real‑world evidence on perceived effectiveness. This non‑interventional, retrospective, cross‑sectional survey evaluated the product attributes most valued when selecting cold and flu medications and assessed perceptions of the effectiveness, quality‑of‑life impact, and overall attitudes toward MED‑LEMON. A total of 249 adults completed a structured questionnaire covering symptom relief priorities, medication attributes, perceived effectiveness, quality of life outcomes, and post-intake attitudes. Relief of fever, sore throat, headache, and sinus congestion, along with fast action, long‑lasting relief, and ease of use, were identified as key drivers of OTC cold and flu medication choice. MED‑LEMON was widely perceived as effective, with over 90% of participants reporting overall symptom improvement and strong relief of pain, headache, and fever. Adherence to the recommended dosing regimen was associated with better symptom control and improved quality of life, including sleep, emotional well‑being, and daily functioning. Overall attitudes toward MED‑LEMON were highly positive. From a public health perspective, these findings highlight the role of effective OTC treatments in supporting health promotion and responsible self-care during cold/flu period.

Abstract: Dry reforming of methane (DRM) is an attractive route for H2 production and simultaneous CO₂ utilization, but its practical implementation is limited by catalyst deactivation. This study experimentally investigates the catalytic performance of Ni/Al₂O₃ and Gd-doped ceria–promoted Ni/GDC–Al₂O₃ catalysts for DRM in a fixed-bed quartz reactor over 400–800 °C at gas residence times of 0.1 s and 0.4 s. Increasing temperature and residence time enhanced CH₄ and CO₂ conversion as well as H₂ and CO yields for both catalysts. The GDC-promoted catalyst exhibited markedly improved activity, achieving conversions and product yields at 0.1 s comparable to those of Ni/Al₂O₃ at 0.4 s and reaching complete CH₄ conversion at about 650 °C, approximately 100 °C lower than the Ni/Al₂O₃. Long-term testing demonstrated high durability of Ni/GDC–Al₂O₃ at 650 °C with no detectable carbon deposition, consistent with thermodynamic equilibrium analysis.

Abstract: Background: China and India represent a large proportion of the Asian birth cohort and have produced extensive but heterogeneous evidence on neonatal hearing screening. This scoping review summarizes studies published between 2005 and 2025 on otoacoustic-emission-based neonatal hearing screening programs in these countries, with emphasis on program implementation, screening coverage, prevalence of congenital and bilateral hearing loss, follow-up, and intervention pathways. Methods: Searches were conducted in PubMed, Scopus, and Google Scholar using predefined keywords. Studies reporting screening protocols, coverage, prevalence, or follow-up outcomes were included. The standard English language filter was used. A total of 19 papers were considered for this review. Results: The data from the two assessed Asian states show two clearly different screening implementation profiles. In China, Universal hearing screening has evolved into a large-scale and increasingly standardized system, supported by technical specifications and regional or municipal databases; The reported screening coverage was 85.8% in early rural programs, 93.6% in Shanghai, and 97.9% in Liuzhou, while national institutional surveys indicate that UNHS has been substantially implemented in many regions. Reported Hearing Loss prevalence estimates generally ranged from 1.66 to 3.43 per 1,000 newborns, although follow-up and regional equity remain problematic, especially in rural settings. In India, the evidence is dominated by tertiary-hospital feasibility studies rather than a uniformly implemented national program. Reported Hearing loss prevalence estimates varied more widely, from 0.29 to 5.60 per 1,000 screened newborns, largely reflecting differences in study design, screening timing, referral completion, and population risk profile. Across both countries, OAE-based two-stage or sequential OAE+AABR protocols reduced referral rates and improved case identification, but loss to follow-up remained a recurrent limitation. Conclusions: China and India provide complementary models of neonatal hearing screening expansion: China demonstrates the effects of system-level scale-up, whereas India highlights the feasibility and constraints of hospital-based implementation in a highly diverse healthcare environment. Future priorities include stronger follow-up systems, harmonized reporting standards, and broader dissemination of outcome data through peer-reviewed publications.

Abstract: Background/Objectives: Children growing up in families with alcohol-related problems are considered a high-risk group for developmental, emotional, and cognitive difficulties, although this condition is not classified as a clinical diagnosis in DSM-5 or ICD-11. The aim of this study was to develop a neurofunctional profile of such children based on electroencephalographic (EEG) markers, in order to identify indicators of neurodevelopmental risk and explore their potential relevance for pedagogical and social interventions. Methods: The study employed resting-state EEG recordings in children aged 6–10 years from alcohol-affected families and a control group. Quantitative EEG (qEEG) indices were analyzed, including theta–beta ratio (TBR), frontal alpha asymmetry (FAA), temporal beta activity, and beta2 power in parietal regions. Standard preprocessing procedures were applied, and between-group comparisons were conducted using Welch’s t-tests with correction for multiple comparisons. Results: Children from alcohol-affected families exhibited significantly elevated TBR indices (global, frontal, prefrontal, and midline), increased temporal beta activity and SMR composite values, and higher beta2 power in parietal regions. Additionally, reduced alpha power in the prefrontal region (Fp1) was observed. These patterns are consistent with differences in attention, executive functioning, emotional regulation, and stress reactivity. No significant differences were found for frontal alpha asymmetry after correction. Conclusions: The findings indicate the presence of distinct group-level EEG patterns associated with children from alcohol-affected environments. These results may contribute to understanding developmental variability in high-risk populations; however, they should not be interpreted as indicators of individual impairment or causal mechanisms. The study highlights the potential, but still limited, applicability of EEG-based measures in informing educational and social support strategies and underscores the need for further research integrating neurophysiological and environmental perspectives.

Abstract: Soil liquefaction is a significant geotechnical hazard that can lead to severe structural damage during seismic events. Traditional liquefaction assessment methods, such as those based on the Standard Penetration Test (SPT) and Cone Penetration Test (CPT), rely on empirical correlations but often struggle to capture the complex, nonlinear interactions between soil properties and seismic parameters. Recent advancements in machine learning (ML) offer data-driven approaches that can improve liquefaction prediction accuracy. This study evaluates and compares the performance of Random Forest (RF) and Artificial Neural Networks (ANNs) for liquefaction potential prediction using a dataset containing 480 field observations derived from CPT-based studies. The dataset was preprocessed using min-max normalization, and models were trained and optimized through hyperparameter tuning. Model performance was assessed using accuracy, precision, recall, F-measure, Cohen’s kappa, and AUC-ROC analysis. The results show that RF achieved the highest accuracy (89%), outperforming both ANN (86%) and the traditional CPT-based liquefaction assessment method (87%). Additionally, ROC-AUC values of 0.932 for RF and 0.872 for ANN indicate the superior classification capability of machine learning models. Feature importance analysis in RF revealed that cone tip resistance (qc), cyclic stress ratio (CSR), and peak ground acceleration (amax) are the most influential factors in liquefaction prediction. These findings demonstrate that machine learning techniques, particularly RF, provide more reliable liquefaction predictions compared to conventional empirical methods. The study highlights the potential of ML models in improving seismic risk assessments and guiding engineering decision-making processes.

Abstract: During the exploration drilling process, maintaining a vertical well trajectory is a critical issue. In geological formations with complex conditions that are prone to well deviation, conventional drilling tool assemblies exhibit poor anti-deviation performance. To achieve anti-deviation and accelerate drilling in exploration wells, a pre-bent drilling tool assembly is proposed. In this study, a dynamic model of the pre-bent drilling tool assembly was established. The anti-deviation mechanism of the pre-bent drilling tool assembly was investigated. The deviation-reduction effects of the drilling tool assembly under different parameter conditions were analyzed. The results indicate that the deviation-reducing force initially increases and then decreases as the pre-bend angle of the anti-deviation drilling tool increases. When the bend angle is between 1° and 1.13°, a larger deviation-reducing force is generated at the drill bit. A shorter distance (L1) between the near-bit stabilizer and the drill bit, a smaller near-bit stabilizer diameter, and a larger upper stabilizer diameter result in a greater deviation-reducing force. The relationship between the deviation-reducing force and the distance between the two stabilizers (L2) is not explicitly linear, but a decreasing trend is observed after the distance exceeds 10 m. Compared with the conventional pendulum anti-deviation drilling tool assembly, the deviation-reducing force of the pre-bent drilling tool assembly has an advantage of more than two orders of magnitude. Based on the calculation results, the optimal design of the pre-bent drilling tool assembly was carried out. The bend angle was increased to 1.15°, the diameter of the near-bit stabilizer was reduced to 305 mm, L2 was reduced to 9–11 m, and L1 was reduced to 0.9 m. Field applications in 22 exploration wells show that the pre-bent drilling tool assembly provides excellent anti-deviation effects. It can fully release the weight on bit while ensuring a vertical trajectory, achieving a 14% increase in the drilling rate. This technology effectively replaces vertical steering tools. Tool costs are significantly saved, providing an effective method for anti-deviation in complex formations.

Abstract: Objectives: This study aimed to evaluate the forward translation of the maxillary base in adults undergoing 3D-guided midpalatal piezocorticotomy-assisted Miniscrew-Assisted Rapid Palatal Expansion (MARPE). Furthermore, the research investigated the contributing factors for forward maxillary movement and the subsequent immediate shift of the mandible. Methods: In this retrospective quasi-experimental study, cephalometric records of 80 adult patients (mean age 35.23 ± 8.76 years; 52 females, 28 males) were analyzed. Maxillary translation was assessed via SNA and A-Nperp(FH), while intermaxillary changes were measured using the ANB angle. Vertical and rotational changes were tracked through SN-MP, FH-MP, and various occlusal plane angles (OcP-FH, OcP-SN, OcP-GoMe). Facial height dimensions (TAFH, UAFH, LAFH, PFH) and dento-alveolar positions (U1-MP, U1LENGTH) were also recorded. Results: Following intervention, significant increases were observed in SNA (0.96°; 95% CI [0.48, 1.43]), ANB (1.42°; 95% CI [1.04, 1.80]), and A-Nperp(FH) (0.81 mm; 95% CI [0.24, 1.39]). The SN-GoMe angle increased by 0.98° and Posterior Facial Height (PFH) increased by 1.57 mm, while upper incisor length (U1LENGTH) significantly decreased by 0.71 mm. Conclusions: 3D-Guided midpalatal Piezocoroticotomy Assisted MARPE in adults is associated with the increase in SNA, ANB, SN-GoMe, Posterior Facial Height, A-Nperp(FH), and decrease in the maxillary incisor length. The amount of anterior midpalatal separation is not associated with the SNA increase while the latter is associated with the inclination of the maxillary plane (SN-MP).

Abstract: This work focuses on parametric optimisation and the prediction of performance for NiCr/WC-Co coatings prepared using high-velocity oxygen fuel (HVOF) spraying. An L18 orthogonal experimental design based on the Taguchi method and the response surface method (RSM) was adopted to examine how key process parameters affect the microstructure, phase composition and hardness of the coatings. A total of eight controllable factors were selected and the hardness, microstructure and phase characteristics of the coatings were evaluated using a Vickers hardness tester, scanning electron microscopy and X-ray diffraction. Analysis of variance (ANOVA) revealed that travel velocity, methane flow rate, powder feed rate and spraying distance were the dominant parameters affecting coating hardness, accounting for altogether 76.25% of the total variance.The model established in this study demonstrates remarkably high predictive accuracy, with a coefficient of determination (R²) of 0.985 and an average prediction error of just 1.16%. This model accurately reflects the nonlinear relationship between process parameters and coating hardness. Meantime, verification experiments were conducted under optimal conditions. The measured hardness was 1352.7 ± 75 HV, in close agreement with the predicted value of 1365 HV. This result has a relative error of 0.98%, which validates the reliability of the second-order model, and a dense layered structure, low porosity, and minimal decarburization of tungsten carbide are exhibited by the coating. Adding a NiCr intermediate layer improves interfacial bonding and reduces structural defects. It is demonstrated by the results that the Taguchi-RSM method is reliable for the optimization of HVOF spraying parameters and the prediction of coating hardness. Overall, this study provides technical support and industrial application for the preparation of high-performance NiCr/WC-Co ceramic-metal composite coatings.

Abstract: Background/Objectives: Artificial intelligence (AI) and its subfield deep learning (DL) have rapidly evolved into a central tool in modern medicine. The purpose of this work was to examine if DL neural networks can discriminate efficiently the microvessel network of post-COVID-19 patients from healthy individuals from. Methods: A non-contact, digital slit-lamp video capillaroscopy system was used to record high magnification images form the bulbar conjunctival microcirculation of 12 COVID-19 survivors (named “COVID-19 Group”) and 12 healthy volunteers (named “Control Group”). Four pretrained convolutional neural networks (CNNs) were fine-tuned by transfer learning and their performance was assessed by standard binary classification evaluation criteria. Results: A scene-centric CNN named GoogLeNet-Places365 excelled on all evaluation criteria with an average testing accuracy, sensitivity, specificity and AUC (area under the curve) of 92%, 92%, 91%, and 0.971, respectively. Conclusions: Post-COVID effects on the eye microcirculation can be detected by deep CNNs, and there is now evidence for the first time, that AI could provide a risk-free, painless, contactless, fast, and accurate detection method of viral effects that does not depend on the optical clarity of the eye.

Abstract: We study a class of wrapping operators acting on the space of formal languages over a fixed finite alphabet. The underlying space is equipped with a length-based ultrametric, in which two languages are close whenever they coincide on all sufficiently short words. We prove that every wrapping operator generated by a finite family of guards with positive total guard length is a contraction. As a consequence, Banach’s contraction principle yields existence and uniqueness of a fixed point for the corresponding recursive language equation, together with convergence of the Picard iteration from an arbitrary initial language. We also obtain an explicit quantitative estimate for the rate of convergence. This makes it possible to determine how many iterations are sufficient to recover the fixed point correctly on all words up to a prescribed length. Several examples illustrate the theory, including operators with different guard lengths and a case showing that convergence in the length-based ultrametric does not coincide with set-theoretic convergence. An application to recursive structures and document validation is also presented, including recursive data formats, abstract syntax trees, and a restricted fragment of JSON schemas. The results provide a formal foundation for validation together with explicit bounds for correctness on inputs of bounded length.

Abstract: The preparation of annotated datasets remains a critical bottleneck in the machine learning (ML) pipeline. Existing tools are fragmented across cloud-hosted services, self-hosted web applications, and lightweight desktop tools—none simultaneously ad-dressing diverse annotation modalities, offline-first operation, integrated training, and serverless collaboration. We present Annotix, an open-source, cross-platform desktop application built on a Rust backend (Tauri 2) and React 19 frontend, designed to unify the entire ML data preparation workflow within a single privacy-preserving environ-ment. To evaluate its practical utility, we conducted a controlled annotation efficiency study using 60 synthetic images (bounding box and mask tasks) annotated by three expert evaluators across Annotix, CVAT, and Label Studio, analyzed via Krus-kal-Wallis with Dunn–Bonferroni post-hoc tests, and a heuristic usability evaluation over standardized tasks on real medical images (retinographies and otoscopies). Re-sults demonstrate that Annotix achieves statistically significant annotation efficiency relative to established tools while offering substantially broader feature coverage, in-cluding 7 image annotation primitives, 19 ML training backends, ONNX-based infer-ence-assisted labeling, and serverless P2P collaboration. Annotix provides a complete, privacy-preserving ML data preparation workflow suited to regulated domains such as medical imaging and ecological monitoring and is freely available under the MIT license.

Abstract: Background:Advances in oncology have led to the development of novel targeted therapies with demonstrated efficacy in clinical trials; however, their real-world economic impact prior to and after market introduction remains insufficiently characterized [1,2]. Cancer-related healthcare costs vary significantly depending on disease stage, time since diagnosis, tumor type, and therapeutic approach[3–6], making inter-hospital comparisons challenging due to heterogeneity in patient populations and information systems [7]. Therefore, integrating cost analysis with clinically meaningful patient stratification is essential to improve resource allocation and outcome evaluation[8–12]. Methods: A multicentre working group comprising four tertiary hospitals in Madrid (Spain) was established to develop and validate a novel classification system for adult oncohematological patients. A standardized methodology was designed to stratify patients into homogeneous groups (PATONCO categories) based on tumor location, therapeutic objective, and clinically relevant biomarkers. A cost indicator was defined as the average cost per patient per month for each PATONCO category. Data were extracted from pharmacy dispensing systems and analyzed using descriptive and inferential statistics, including Kruskal–Wallis and post hoc Dunn tests. Results: A total of 3,659 patients were included (3,168 oncology; 491 hematology), distributed across 62 programmes (54 oncology; 8 hematology). The PATONCOS tool enabled the identification and validation of a cost indicator (average cost/patient/month per category), allowing inter-hospital comparison. Significant differences in costs were observed across most high-prevalence categories, reflecting variability in therapeutic strategies and adoption of innovative treatments. The model demonstrated its capacity to detect intra-group homogeneity and inter-group variability, improving the identification of high-cost patient subgroups and supporting benchmarking across centres. Conclusions: The PATONCOS tool provides a novel, clinically oriented stratification methodology that integrates pharmacotherapy, biomarkers, and disease stage with economic evaluation. This approach enables more accurate comparisons of oncology treatment costs between institutions and supports data-driven decision-making in resource allocation. Its implementation may contribute to more sustainable healthcare systems by aligning clinical practice with economic outcomes.