Browse Articles

Case-matched comparison of combined phacoemulsification with ab-interno trabeculectomy via Kahook dual blade and trabectome in a Caucasian population

Abstract The purpose of this retrospective study was to compare the outcomes after combined phacoemulsification and ab-interno trabeculectomy via Kahook Dual Blade (KDB) and Trabectome, being represented in two groups of patients of Caucasian ethnicity with matched baseline criteria. We included 60 eyes of 49 participants being treated for cataract, of which 30 eyes underwent additional ab-interno trabeculectomy via KDB (Kahook group) and 30 eyes received additional Trabectome surgery (Trabectome group). For this comparative analysis, the Kahook group and Trabectome group were matched at a 1:1-ratio, based on the following criteria: preoperative IOP, maximum known preoperative IOP, preoperative medication score, cup/disc-ratio, follow-up time, best-corrected visual acuity and age. Successful surgery was defined by three scores: IOP at longest follow-up < 21 mmHg (Score A) or < 18 mmHg (Score B) without re-surgery and an IOP reduction > 20% or IOP ≤ 15 mmHg without re-surgery and an IOP reduction ≥ 40% (Score C). Furthermore, we compared postoperative IOP, as well as medication score, and side effects between both groups. Both surgical techniques led to a relative IOP reduction of 29% within their respective groups. Specifically, preoperative IOP decreased from 19.5 ± 5.0 mmHg to 13.8 ± 3.9 mmHg in the Kahook group, and from 19.8 ± 4.5 mmHg to 14.0 ± 3.9 mmHg in the Trabectome group during an average follow-up period of 23–24 months. There was no statistical significant difference noted. Both the KDB and Trabectome yielded similar success rates, according to Score A (67% vs. 70%), Score B (63% vs. 67%) and Score C (33% vs. 23%). There were no severe side effects notes in either group. In conclusion, the KDB and Trabectome showed similar IOP-lowering properties and safety profiles within our two matched groups of Caucasian patients.

Prevalence of heart failure with preserved ejection fraction in patients with ischemia and non-obstructive coronary arteries

Deployment of a machine learning-based predictive system for childhood diarrhea in Sub-Saharan Africa

Abstract Diarrhea remains a leading cause of child mortality in Sub-Saharan Africa, necessitating advanced predictive tools for early intervention. Despite the growing adoption of machine learning in healthcare, gaps persist in deploying models as scalable, real-world solutions. This study developed an end-to-end machine learning framework to predict diarrhea among children under five in SSA, integrating rigorous model development with Flask-based deployment for practical use. Using nationally representative Demographic and Health Surveys (DHS) data from 27 SSA countries (2016-2024), we preprocessed data (handling missing values, feature selection, and SMOTE for class imbalance), trained a Random Forest classifier (optimized via RandomizedSearchCV), and deployed the model as a RESTful API with Flask. The final model demonstrated strong predictive power, with 79.6% accuracy and a particularly high recall of 84.1%, meaning it is exceptionally effective at identifying true diarrhea cases. Most importantly, the model is no longer just a research output; it is a deployed, interactive system ready for practical application. This work successfully demonstrates a complete pipeline from data to deployment, offering a tangible solution that can aid public health decision-making. We have proven that it is possible to close the gap between machine learning research and real-world implementation. To build on this foundation, future work should focus on enhancing the model’s interpretability for health workers, adopting more scalable deployment technologies like FastAPI and Docker, and conducting rigorous field validation with community stakeholders to ensure these tools truly meet the needs of those they are designed to serve.

Investigation of deformation characteristics of gate and earth–rock dam systems under deep overburden conditions

High-efficiency predictive torque control of induction motors in PV water pumping using GTO-optimized PI controller

A study on the efficacy and safety of fecal microbiota transplantation as an adjunctive therapy for treating depressive episodes

SV-TransFusion for LiDAR 3D object detection with Sparse Voxel–Query Interaction

Integrating EfficientNetV2 with guided filopic diffusion for enhanced rice leaf disease recognition

Spinal disease image segmentation technology integrating U-ResNet and shape-aware attention

Separating tectonic and climate signals in Holocene sea-level records using marine terraces in central Chile

Abstract Field observations of past sea-level variations are needed to validate models predicting future sea-level rise. Along tectonically active coasts, separating tectonic and non-tectonic sea-level components is challenging as both have similar amplitudes but necessary to decipher sea-level histories driven by climate forcing. Here, we present a new framework to decipher Holocene sea-level changes using marine terraces–geomorphic features formed by wave erosion of bedrock–mapped with high-resolution LiDAR data and numerical modelling. Applied to 266 sites along 500 km of central Chilean coast, we found that Holocene terrace elevations linearly correlate with Late Pleistocene terrace elevations, evidencing steady-state tectonics over the past 125,000 years. This proof of steady-state uplift allows subtracting tectonic components from Holocene elevations using uplift rates from Pleistocene terraces. We find that during the mid-Holocene, sea level reached 3.18 ± 0.15 m above modern elevation, only 0.33 m below glacial isostatic model predictions with 2·10 20  Pa·s mantle viscosity. We validated this relationship by reproducing Holocene terrace elevations using a landscape evolution model and glacial isostatic sea-level curves. Our results suggest that accounting for millennial-scale vertical land motion rates that average many seismic cycles may improve future relative sea-level change projections, highlighting the potential of rocky-shore geomorphology for sea-level research along tectonically active coastlines.

Unsegmented marine annelids as biomechanical models for soft robotics

In-situ ion irradiation investigations on MBE grown Sb thin films on sapphire

Halide-assisted Al-doped graded shells for emission tunability and photostability in CdSe NPLs

Knowledge and practice of surgical site infection prevention and associated factors among nurses working in public hospitals of Sodo town, Wolaita Zone, Southern Ethiopia

Improving elderly-oriented transportation in rural areas through a case study of Zhenglu Town

Multidisciplinary approaches to lithological discrimination and structural mapping for mineral resource assessment

Abstract Understanding the geological and structural controls on mineralization is globally important for sustainable resource exploration and management. This study investigates the Gardan Ophiolitic Mélange (GOM) and the Shait Granite Complex (SGC) in the Wadi Shait area, Eastern Desert, Egypt, using an integrated, multidisciplinary approach. The GOM comprises basal metasediments, metabasalt slices, and schistose hornblende-bearing metagabbros, representing a tectonically imbricated, low-grade metamorphosed unit. Sentinel-2 satellite imagery, analysed using false colour composites (FCCs), principal component analysis (PCA), band ratios, and textural correlation, effectively discriminated lithological units and structural patterns. These results were validated and extended using high-resolution aeromagnetic data. Interpretation of aeromagnetic, employing advanced edge-detection filters such as the improved horizontal tilt derivative (impTDX) and STDR, together with classical techniques including the first vertical derivative (FVD), horizontal gradient magnitude (HGM), and tilt derivative (TDR), as well as three-dimensional (3D) Euler deconvolution and 3D magnetic modeling, revealed a network of NW–SE, NE–SW, N–S, and E–W trending faults at depths of 124–782 m. Within the SGC, NW-trending shear zones indicate late orogenic extensional exhumation associated with Najd fault-related tectonics. These structures govern the distribution of gold and radioactive (K, U, Th) mineralization. The results highlight the effectiveness of integrating remote sensing and aeromagnetic techniques for resolving lithological complexity, subsurface architecture, and mineral potential in structurally complex terranes worldwide.

A selective machine learning algorithm for severe periodontitis labeling from questionnaire data

Suppression of LTBP1 enhances the sensitivity of bladder cancer to cisplatin

Exosome-mediated delivery of miRNA-1290 inhibitor enhances JNK-dependent tissue-resident memory T cell immunity in prostate cancer

PADP: progressive and adaptive data pruning for efficient incremental learning

Abstract Data pruning is a key technique for reducing training costs and improving model performance. However, most existing methods rely on fixed pruning rates or single metrics, which are largely designed for static training settings, making them unsuitable for incremental learning, where data distributions and model states change dynamically. To address this, we propose , a progressive and adaptive data pruning method for incremental learning, which dynamically evaluates sample difficulty and its difficulty changes during training to enable adaptive sample selection for each incremental learning task. Specifically, we propose two metrics, the instant difficulty score and the difficulty variation score. The former evaluates the learning difficulty of a sample, while the latter evaluates the variation in difficulty over a training interval. These two metrics are combined to guide pruning decisions. To prevent certain classes from being completely removed, we also introduce a class-balance retention mechanism. Experimental results show that outperforms existing data selection methods on CIFAR-100 and Tiny-ImageNet, generalizes across multiple incremental learning frameworks, and still maintains or exceeds the original accuracy even when training time is reduced by up to 52.90% compared to using the full dataset, demonstrating its effectiveness and practical value.