Search Results (3,068)

Background: This study aimed to compare the measurement agreement and intramethod reliability of four cephalometric analysis workflows: manual tracing, semi-automated digital analysis (Dolphin), fully automated AI-based analysis (WebCeph), and expert-adjusted AI analysis (WebCeph+). Methods: In this retrospective method-comparison study, 67 lateral cephalometric radiographs were initially included. After the exclusion of radiographs containing extreme values, 54 radiographs (35 females, 19 males; mean age: 15.0 ± 2.13 years) were analyzed. Twenty-one skeletal, dental, and soft-tissue parameters (13 angular, 8 linear) were evaluated across the four methods. Intramethod repeatability was assessed via the intraclass correlation coefficient (ICC). Intermethod comparisons were analyzed using ANOVA and post hoc pairwise tests. Pragmatic clinical relevance thresholds were predefined as ±2 degrees for angular measurements and ±2 mm for linear measurements. Results: All methods demonstrated high intramethod reliability, with ICC values exceeding 0.90 in 20 out of 21 parameters. Manual and Dolphin methods yielded statistically comparable results (p > 0.05). In contrast, WebCeph differed significantly from manual and/or Dolphin in seven parameters, including SNA, IMPA, Go-Gn length, Pog to N-perpendicular, Wits appraisal, nasolabial angle, and mentolabial angle (p < 0.05). Several discrepancies exceeded the predefined pragmatic thresholds (±2 degrees and ±2 mm), highlighting their potential clinical relevance. After expert adjustment (WebCeph+), statistically significant inter-workflow differences were no longer observed; however, residual individual-level variability remained for selected parameters. Conclusions: Fully automated WebCeph analysis showed limited agreement with manual and semi-automated methods for several clinically relevant measurements. Expert adjustment reduced systematic mean discrepancies and improved agreement with clinician-dependent workflows; however, residual individual-level variability remained for selected parameters. AI-driven cephalometric analysis should therefore be considered a supportive tool requiring specialist verification rather than an unsupervised replacement for conventional methods. Full article

CT perfusion (CTP) is widely used in acute ischemic stroke imaging, particularly for treatment selection beyond conventional time windows. However, automated perfusion maps are not direct measurements of irreversible tissue injury, but estimates shaped by deconvolution strategy, temporal correction, dispersion handling, and software-specific thresholds. This review provides a clinically oriented explanation of how CTP algorithms influence the estimation of ischemic core and hypoperfused tissue. Particular attention is given to singular value decomposition (SVD) methods, Bayesian approaches, and timing parameters, including time to maximum (Tmax), Delay, time to peak (TTP), and mean transit time (MTT). Differences in residue function estimation and threshold definition may generate variable outputs across software platforms, even from the same source dataset. Perfusion thresholds should therefore not be treated as universally interchangeable. CTP findings should be integrated with clinical status, non-contrast CT, CT angiography (CTA), collateral status, occlusion site, and imaging-to-treatment context, serving as decision-support tools rather than isolated measures of tissue viability. Full article

Background: Treatment monitoring in pulmonary tuberculosis increasingly requires assessment of residual inflammatory burden and structural lung damage beyond microbiologic response alone. High-resolution computed tomography (HRCT) can provide this information, but interpretation of serial examinations is time-consuming and partly subjective. This study did not aim to evaluate AI for the diagnosis of pulmonary tuberculosis. Instead, it explored whether artificial intelligence (AI)-assisted quantitative HRCT analysis could support longitudinal assessment of treatment-related imaging changes in patients with microbiologically confirmed pulmonary tuberculosis. Methods: We conducted a retrospective, single-center, exploratory longitudinal study of patients receiving treatment for pulmonary tuberculosis. HRCT examinations acquired at diagnosis and during follow-up were anonymized, reviewed by an expert thoracic radiologist, and processed using AVIEW Lung Texture (Coreline Soft v2.0). The software quantified total lung volume and six predefined parenchymal categories: normal lung, ground-glass opacity, consolidation, reticulation, honeycombing, and emphysema. Results: Ninety-six patients contributed 256 HRCT examinations. The most frequent software-detected abnormalities were ground-glass opacity, consolidation, and emphysema-labeled low-attenuation areas. Ground-glass opacity and consolidation showed the clearest decline across serial examinations, consistent with regression of active inflammatory disease during treatment. Reticulation showed a heterogeneous course, likely reflecting both inflammatory resolution and residual structural remodeling. Honeycombing was infrequent and quantitatively limited. Lung volume changed variably and did not consistently parallel visual improvement. A key methodological limitation was the absence of a dedicated cavity class. As a result, emphysema-labeled low-attenuation areas should not be interpreted as conventional emphysema alone, because tuberculous cavities and post-destructive abnormalities were frequently included in this category. Conclusions: AI-assisted HRCT quantification may support longitudinal assessment of pulmonary tuberculosis by providing structured and reproducible measures of interval change. However, tuberculosis-specific interpretation remains dependent on expert radiologic oversight, particularly in cavitary disease. Full article

Global particleboard production reached almost 123 million m³ in 2024. Such quantities require intensified search for alternative lignocellulosic materials in accordance with the principles of the circular economy. The research aim was to determine the total energy consumption (E_max) and specific cutting work (SCW) when drilling holes in boards made from 100% forest biomass, agricultural biomass, and wood-based post-production residues. The experiments were carried out on a CNC (computerize numerical control) center at a constant speed of 6000 rpm and variable feed rates of 0.2, 2, and 20 mm/min, using conventional particleboard as a reference. The results showed that the feed rate has a dominant influence on energy consumption. As the speed increased, the average SCW value decreased from 22.32 J/mm³ to 6.11 J/mm³. Conventional board required the highest energy input in all variants, and statistical analyses showed no significant differences (p > 0.05) between boards made from alternative raw materials and the reference. This proves that the variability of energy consumption depends mainly on the process parameters and not on the material itself. The research confirms the technological feasibility of using alternative biomass sources in the furniture industry without increasing energy consumption during processing, which promotes the implementation of sustainable development principles. Full article

Vegetation change in cryosphere-affected mountain basins reflects interacting climate and human pressures but their relative influence remains uncertain in the Upper Indus Basin. The novelty of this study is the integration of satellite vegetation, climate variables, human pressure indicators, residual attribution and diagnostic validation in a data-scarce high-mountain basin. We evaluated growing-season Normalized Difference Vegetation Index dynamics and associated drivers from 2001 to 2023 using trend analysis, correlation, Random Forest diagnostics, Sentinel 2 validation, and residual trend analysis. The results showed widespread greening across 96.59% of the basin, with stronger improvement in the lower and central areas. Significant greening covered 69.94% of the basin, while only 1.55% showed significant browning. Precipitation and temperature were predominantly positive drivers of vegetation change, whereas potential evapotranspiration and solar radiation were mostly negative. Soil moisture played a strong regulatory role along elevation gradients. Residual trend analysis provided approximate and method-dependent estimates of the possible anthropogenic influence on vegetation change at 73.09% and climatic drivers at 26.91% rather than direct causal decomposition. These values are approximate and method-dependent estimates, not direct causal decomposition. The findings highlight human-related greening in lower valleys and climate-controlled vegetation responses in high-mountain areas. Full article

The increasing adoption of Plant Factories with Artificial Lighting (PFAL) has intensified the reliance on Internet of Things (IoT) technologies for real-time monitoring and control of environmental and operational variables. While IoT-based architectures enable precise resource management and productivity optimization, PFAL systems remain highly vulnerable to component failures, sensor malfunctions, communication faults, and energy disruptions, which may compromise crop integrity and system reliability. These risks are particularly critical in low-cost and small-scale PFAL implementations, where maintenance capacity and redundancy are often limited. Existing IoT-based PFAL monitoring systems typically address either hardware or software redundancy in isolation and rarely incorporate a dedicated maintenance-oriented fault detection layer validated under realistic multi-failure scenarios. This study addresses these challenges by proposing a low-cost redundant system architecture for PFAL applications that simultaneously integrates (1) hardware redundancy through multi-sensor configurations; (2) analytical redundancy based on residual generation and threshold-based fault isolation; and (3) a maintenance-oriented fault detection layer capable of identifying abnormal internal device conditions. Experimental validation was conducted using four hardware configurations—Arduino Nano with Ethernet, ESP32, STM32 with Wi-Fi, and STM32 with Ethernet—evaluated across five fault scenarios: dust accumulation, water exposure, high temperature, fire detection, and physical impact. The STM32 with Ethernet configuration consistently achieved the fastest fault detection response times across all tested scenarios. Future work will focus on the integration of machine learning-based predictive maintenance algorithms, multi-node PFAL network deployments, and long-term field validation. Full article

The thermo-mechanical degradation of the base–subgrade interface in airport pavements was investigated using a three-dimensional sequentially coupled finite element framework in ABAQUS 2023, in which progressive interfacial debonding was described by a bilinear cohesive-zone model through the damage variable CSDMG. The results show that thermal loading markedly accelerates interface degradation when combined with moving wheel loads. Compared with the wheel-loading-only condition, thermo-mechanical coupling advances the first damage initiation from 0.04993 h to 0.00254 h and shortens the severe-degradation stage from 1.000 h to 0.00927 h. This acceleration is attributed to a thermal stress pre-weakening effect, whereby constrained thermal deformation partially consumes the available cohesive resistance and shifts the interface closer to the softening threshold before external loading is applied. A decomposition of the mixed-mode initiation criterion further indicates that the first damage event is governed by synergistic normal–shear interaction, with the normalized contribution ratio (t_n/t_n⁰)²:(t_s/t_s⁰)² = 0.38:0.62, showing that wheel-induced shear is the dominant trigger while tensile opening induced by thermal curling provides substantial preconditioning assistance. In addition, a representative normalized comparison between simulated average CSDMG and cumulative AE hit count demonstrates a consistent stage evolution from distributed deformation to accelerated localization and residual stabilization. These findings indicate that the base–subgrade interface should be treated as a temperature-sensitive weak layer in airport pavement assessment, particularly near joints and other discontinuity-controlled regions. Full article

Autoencoders (AEs) have been widely used for industrial process fault detection owing to their ability to learn nonlinear representations from normal operating data. However, conventional AE methods rely heavily on reconstruction errors and may miss weak faults due to overgeneralization. In addition, insufficient modeling of temporal evolution and operating condition variations may reduce their sensitivity to dynamic faults. To address these issues, this study proposes a memory-enhanced and prediction-assisted conditional variational autoencoder named MI-CVAE for unsupervised fault detection. In the proposed framework, statistical features extracted from sliding windows are used as condition information to describe variable operating states. A memory module stores representative normal prototypes to constrain reconstruction and reduce overgeneralization to faulty samples. Meanwhile, an Informer branch captures temporal dependencies and provides complementary prediction residuals. Reconstruction and prediction residuals are fused to construct squared prediction error and squared Mahalanobis distance statistics, with control limits determined by kernel density estimation. The proposed method is validated on the Benchmark Simulation Model No. 1 wastewater treatment benchmark and a real papermaking process dataset. The results show that MI-CVAE outperforms the evaluated comparison methods, particularly in detecting weak and dynamic faults, while maintaining a low false alarm rate. Full article

Qinghai Lake, the largest endorheic saline lake in China, has undergone a pronounced hydrological regime shift from a multi-decadal decline to a rapid post-2004 recovery, reflecting strong hydroclimatic non-stationarity in the northeastern Tibetan Plateau (TP). This paper supplements the current water level and lake area status of Qinghai Lake to provide basic background for future prediction. Reliable forecasting of such climate sensitive lake systems remains difficult because conventional statistical models often fail to capture non-linear fluctuations, whereas standalone deep learning models may overlook long-term deterministic evolution. To address this challenge, we developed a serial decomposition GeoAI framework that integrates autoregressive integrated moving average (ARIMA), one-dimensional convolutional neural networks (1D-CNNs), and long short-term memory (LSTM) networks for non-stationary water level forecasting. Using annual water level observations from 1960 to 2025, the ARIMA component was first used to extract the low-frequency deterministic trend, after which the CNN-LSTM module reconstructed the nonlinear residual variability. The model was trained on the 1960–2012 period and validated over 2013–2025, which represents the most dynamic expansion stage of Qinghai Lake. The hybrid framework outperformed the benchmark models, achieving a Root Mean Square Error (RMSE) of 0.2033 m, Mean Absolute Error (MAE) of 0.1727 m, and Mean Squared Error (MSE) of 0.0413 m² during validation. The decomposition strategy effectively reduced phase lag and amplitude attenuation, improving both predictive accuracy and process interpretability. Multi-step forecasting for 2026–2056 suggests that Qinghai Lake will continue to rise, reaching approximately 3204.08 m by 2056, although the growth rate is projected to slow as negative hydrological feedback strengthen. By explicitly separating deterministic climate scale signals from nonlinear short-term variability, the proposed framework provides a robust and transferable geoinformation based tool for forecasting water level dynamics and supporting adaptive management in climate sensitive, data scarce lake basins. Full article

Background: Diabetic peripheral neuropathy (DPN) is a progressive complication of type 2 diabetes mellitus (T2DM) for which no approved disease-modifying therapy exists. Palmitoylethanolamide/Baicalin (PEA/Bai; Neuridase^®) is a nutraceutical formulation with anti-neuroinflammatory and antioxidant properties; however, real-world evidence on its associations with objective neuropathy biomarkers remains limited; nutraceutical approaches to DPN remain exploratory and adjunctive in the absence of randomised controlled trial evidence of disease modification. Methods: We conducted a single-centre, retrospective, 1:1 matched-cohort study at an Internal Medicine outpatient clinic. Forty-eight T2DM patients with clinically diagnosed DPN who received PEA/Bai supplementation (Neuridase^® group) were matched to 48 untreated controls drawn from a large institutional database, using age, sex, BMI, and diabetes duration as matching variables. Acknowledged a priori limitations include baseline imbalance in neuropathy severity (VAS and ESC) and SGLT2 inhibitor use, reflecting real-world prescribing patterns (confounding by indication) and constituting potential sources of residual confounding that preclude causal inference. The primary outcome was change in VAS neuropathic pain score from baseline (T0) to 6-month follow-up (T6). Secondary outcomes were changes in electrochemical skin conductance (ESC, µS) in hands, feet, and four-limb sum measured by Sudoscan. Results: At baseline, the Neuridase^® group exhibited significantly greater neuropathic burden: higher VAS scores (median 5.5 [IQR 3.8–7.2] vs. 2.0 [0.0–5.0]; p < 0.001) and lower ESC in both hands (53.0 vs. 72.2 µS; p < 0.001) and feet (74.5 vs. 81.0 µS; p < 0.001), reflecting real-world prescribing patterns. Over 6 months, VAS decreased significantly in the Neuridase^® group (5.5→3.0; p < 0.0001; median Δ = −2.5 points, exceeding the clinically important difference), with no change in controls (2.0→2.0; p = 0.85). Differential Sudoscan trajectories were observed: the Neuridase^® group showed significant improvement in hand ESC (53.0→60.0 µS; p = 0.035) and preservation of foot ESC (p = 0.888), while controls exhibited significant deterioration across all three sudomotor indices (hand p = 0.038; foot p = 0.008; four-limb sum p = 0.004). In a complementary categorical pain trajectory analysis, VAS worsening occurred in 31.3% of controls compared with 0% of Neuridase^®-treated patients (p = 0.00022). Among patients with pathological hand ESC at baseline (<60 µS), 27.8% of Neuridase^® patients (n = 36) transitioned to non-pathological values at T6 versus 0% of controls (n = 32; p = 0.001). Conclusions: In a real-world matched cohort, PEA/Baicalin supplementation was associated with clinically meaningful pain reduction and with differential longitudinal sudomotor trajectories compared to matched untreated controls. These exploratory, hypothesis-generating findings from a retrospective non-randomised design are consistent with possible modulatory effects of PEA/Baicalin on objective sudomotor autonomic biomarkers in DPN. Confounding by indication, baseline severity imbalance, and residual confounders including SGLT2 inhibitor use preclude causal interpretation. These observations provide a rationale for adequately powered, prospective, randomised placebo-controlled trials with extended follow-up and structural neuropathy endpoints. Full article

This paper examines how international crude-oil price movements are transmitted to natural-rubber prices through the petrochemical–synthetic-rubber chain, with implications for Thailand as the world’s leading natural-rubber exporter and China as the dominant consumer. Using monthly data from April 2003 to March 2026 on the OPEC reference basket, butadiene, styrene–butadiene rubber (SBR), and the Shanghai natural-rubber benchmark, the analysis combines a nonlinear ARDL specification with a Pesaran–Shin–Smith bounds test, a long-run association decomposition into direct and synthetic-rubber-mediated components with bootstrap inference, and a threshold-NARDL extension that conditions the decomposition on the inventory state. Three findings stand out. First, the synthetic-rubber-mediated component accounts for approximately three-quarters of the estimated oil–natural rubber long-run association (73.5 percent, 95 percent bootstrap CI [60.6, 87.2]), with the residual direct component accounting for the remainder. Second, long-run pass-through is directionally consistent with concentration in the synthetic-rubber component, although Wald tests do not reject symmetry at conventional levels for either the synthetic-rubber component (Wald $p=0.135$) or the direct oil component ($p=0.166$). Third, the synthetic-rubber-mediated share is consistently larger in low-inventory regimes by 26 to 66 percentage points across three alternative regime variables, although the magnitude amplification of asymmetric pass-through itself is not robust. Asymmetric local projections and a Diebold–Yilmaz spillover analysis are reported as complementary horizon-indexed and network checks. The results imply that the synthetic–natural rubber spread, conditioned on the inventory state, may be more informative for natural-rubber price-risk monitoring than crude-oil prices alone. These findings have implications for commodity price-risk monitoring, export-income exposure, and stabilisation design in rubber-exporting economies. Because crude-oil shocks are not externally identified, all estimates are interpreted as decompositions of long-run association rather than causal mediation effects. Full article

In precision gear manufacturing, longitudinal crowning on tooth flanks is commonly produced by applying diagonal feed in worm-type generating processes using tools such as variable-tooth-thickness hobs and dressable grinding worms. However, precise twist control remains difficult because the geometric parameters of the generating tool are strongly coupled with the machine feed settings in the underlying generating kinematics. In addition, direct numerical optimization becomes unreliable near the standard tool state, where the sensitivity of the diagonal-feed coefficient degenerates and conventional linearized solvers may lose effectiveness. To address these issues, this study proposes a multi-variable optimization framework for twist-constrained worm-type gear generation. An iterative singular value decomposition (SVD) scheme is developed to construct and update the sensitivity matrix, while a warm-start continuation strategy is introduced to overcome the local singularity and improve numerical robustness. Two closed-form expressions for the diagonal-feed coefficient are also proposed as practically useful initial estimates, corresponding respectively to the minimum SVD topographic residual and the minimum tooth-flank twist. Numerical validation over a 60-case parameter sweep shows maximum relative errors below 1.6% within the tested range. The proposed framework coordinates the tool-geometry design and diagonal-feed selection to generate tooth flanks with prescribed crowning characteristics while satisfying a specified twist requirement and limiting the required diagonal shift. Numerical examples show that the iterative framework reduces the root-mean-square (RMS) topographic error from 1.14 μm to 0.027 μm relative to the analytical setting of Hsu and Fong. These results indicate that the proposed method provides a reliable computational basis for twist control and process-parameter design in advanced CNC gear generation. From a manufacturing standpoint, because the three design criteria are accessed by adjusting only the diagonal-feed ratio on the machine, a single generating-tool design can serve a range of crowning and twist requirements without retooling, reducing setup and tooling efforts in production. Full article

Concrete cracks are ubiquitous in practical engineering structures and continuously affect structural safety and durability. Crack images provide important visual evidence of damage evolution; however, crack images alone are insufficient to determine the residual load-bearing capacity of concrete members. Although the development of deep learning algorithms has significantly improved the automatic detection of concrete surface cracks, most existing methods remain limited to the extraction of crack geometric features and lack a direct connection with mechanical performance. To explore the relationship between image-based crack geometry and mechanical response, this study combines U-Net-based crack segmentation, OpenCV-based crack geometry extraction, and phase-field fracture simulation to establish a preliminary visual–mechanical framework for plain concrete beams. In this framework, surface crack images are first segmented using a U-Net model, and crack length, average width, and propagation path are extracted from the predicted binary masks. The extracted crack length is then used as the primary variable to match the observed crack state with the phase-field crack evolution sequence. Once the corresponding simulation stage is identified, the associated load level and residual load-bearing capacity can be obtained from the simulated load–crack mouth opening displacement (Load–CMOD) response. Through a mixed-mode I–II fracture test, the crack geometric features extracted by deep learning are compared with the phase-field simulation results. The results show that the error in crack length is within 2.5%. Meanwhile, the relative error between the simulated peak load and the experimental value was 1.57%, which preliminarily verified the correlation between image-based crack information and the load-bearing capacity of plain concrete beams. The method is further applied to a Mode I fracture test without recorded load-bearing capacity data. By mapping the crack length identified from the image, namely 36.89 mm, to the phase-field evolution sequence, the load-bearing capacity of the member at this stage is estimated to be 74.4% of the peak load. The results indicate that the crack geometry extracted from images can be correlated with phase-field crack evolution, thereby supporting preliminary residual load-bearing capacity assessment of plain concrete beams. However, the proposed framework should be regarded as a case-level feasibility study rather than a general structural assessment method. Before broader engineering application, further validation using synchronized crack image sequences, crack mouth opening displacement (CMOD) measurements, and load records is required. Full article

The production process of hot-rolled seamless steel tubes often needs secondary heating for tempering treatment, resulting in high energy consumption and low production efficiency. Controlled cooling technology has been introduced into the production process. By controlling the self-tempering temperature, the residual temperature of the steel tubes is used to realize self-tempering, so as to achieve the purposes of energy savings, emissions reduction, cost reduction, and efficiency increase. This study investigated the evolution of temperature and stress fields in a seamless steel tube during temperature-controlled quenching. The size of the steel tube is 140 mm × 20 mm × 200 mm. The outer-wall cooling intensity and the self-tempering temperature were selected as the main variables, and the other process parameters remained constant. The temperature distribution and stress variation curves under different cooling intensities were obtained. The results showed that the greater the cooling intensity of the outer wall, the higher the temperature recovery of the outer wall. Under the fixed cooling intensity, the lower the self-tempering temperature, the lower the return temperature. In the thickness direction of the steel tube, there is a stress distribution of “internal tension and external compression”. Moreover, the greater the cooling intensity, the lower the self-tempering temperature and the greater the residual stress. Full article

Background: Acute kidney injury (AKI) is one of the more serious complications following cardiac surgery, consistently linked to prolonged mechanical ventilation and higher in-hospital mortality. This study examined whether preoperative anaemia and impaired renal function are associated with AKI and death in a high-risk cardiac-surgery cohort requiring extended postoperative ICU monitoring and how these associations behave after adjustment for procedure type and intraoperative variables. Methods: In this single-centre retrospective cohort study, we screened 950 patients admitted to a cardiothoracic ICU between January 2018 and January 2024. After standard exclusion criteria and an audit of operative records, 553 cardiac-surgery patients formed the principal analysis cohort. AKI was defined by KDIGO criteria using serial postoperative serum-creatinine measurements during the first 7 days. Multivariable logistic regression for AKI and in-hospital mortality was built sequentially: Model A (baseline only); Model B (+procedure type); and Model C (+intraoperative variables: aortic cross-clamp time, intraoperative RBC units, and intraoperative inotrope use). Calibration was assessed by the Hosmer–Lemeshow test. Total cardiopulmonary bypass duration was not separately captured in the institutional database and is disclosed as a limitation. Results: AKI occurred in 174 of 553 patients (31.5%), and in-hospital mortality was 16.6% (92/553). Patients with AKI were older (median 77 vs. 68 years, p < 0.001), with lower preoperative haemoglobin (11.4 vs. 12.3 g/dL, p < 0.001) and lower eGFR (38.1 vs. 63.7 mL/min/1.73 m², p < 0.001). The aortic cross-clamp time was shorter in AKI patients (56 vs. 70 min, p = 0.043), a counterintuitive finding likely reflecting residual confounding by case-mix and procedure selection rather than a protective operative effect. In the fully adjusted multivariable model, the haemoglobin–AKI association attenuated and was no longer independently significant (OR 0.89 per 1 g/dL, 95% CI 0.73–1.08, p = 0.24), while intraoperative RBC transfusion emerged as an independent predictor (OR 1.12 per unit, p = 0.046). For mortality, AKI remained an independent predictor after full adjustment for procedure type and intraoperative variables (OR 7.14, 95% CI 1.45–35.13, p = 0.016), with cross-clamp time (OR 1.30 per 10 min, p = 0.010) and intraoperative RBC units (OR 1.48 per unit, p < 0.001) also independently associated. Both fully adjusted models showed acceptable calibration (Hosmer–Lemeshow p = 0.48 for AKI, p = 0.56 for mortality). Conclusions: In cardiac-surgery patients with a prolonged ICU stay, AKI is independently associated with in-hospital mortality even after adjustment for operative variables. The univariable association between preoperative haemoglobin and AKI is attenuated after adjustment for procedure type and intraoperative transfusion exposure, suggesting confounding or mediation by operative and case-mix factors rather than an independent direct association. The contribution of this analysis is aetiological/analytical rather than predictive (modest discrimination, AUROC 0.67 for AKI), and findings should be interpreted within the selected high-risk ICU ≥ 72 h population. Full article