Iğdır University

Recent advances in Kolmogorov–Arnold Networks (KANs) have shown strong potential for functional representation learning; however, existing formulations remain limited in their ability to provide explicit geometric interpretability, curvature control, and deformation-aware adaptation for biomedical image analysis. This paper introduces q-FunKAN, a geometry-aware Functional Kolmogorov–Arnold Network that integrates learnable Lupaş q-Bézier inner functions and q-Hermite spectral parameterizations within a unified geometric–spectral framework for biomedical image enhancement and segmentation. The central novelty of the proposed model lies in embedding a learnable deformation parameter q into the KAN functional space, enabling curvature-adaptive modulation of local anatomical structures while preserving global smoothness through spectral regularization.The proposed framework combines three complementary mechanisms: Bézier-based control-point parameterization for interpretable local deformation, q-Hermite spectral expansion for stable global representation, and topology-preserving deformation regularization for anatomically plausible enhancement and segmentation. This design allows q-FunKAN to explicitly balance local boundary precision and global structural consistency, providing a transparent alternative to highly parameterized convolutional and transformer-based models.Extensive experiments on five benchmark MRI datasets, including BRATS 2021, CHAOS, fastMRI, IXI, and a controlled synthetic phantom dataset, demonstrate the effectiveness of the proposed framework. Compared with strong convolutional, transformer-based, restoration-oriented, and KAN-based baselines, q-FunKAN achieves consistent improvements in image fidelity, perceptual quality, and segmentation accuracy, including gains of up to (+1.3) dB PSNR, (+1.0%) Dice, and (-0.004) LPIPS over leading competing models. Ablation studies further confirm that learnable q-adaptivity, Bézier geometric modeling, Hermite spectral regularization, and Jacobian-based topology preservation make complementary contributions to performance and stability. Qualitative analyses show sharper anatomical boundaries, reduced artifacts, and interpretable q-heatmaps aligned with curvature-sensitive regions.By bridging q-calculus, geometric approximation theory, and functional neural representation learning, q-FunKAN establishes a mathematically grounded, interpretable, and geometry-aware framework for biomedical image enhancement and segmentation.

Efficient enrichment of eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA) from natural sources remains a major challenge for sustainable ω-3 fatty acid production. This study reports the immobilization of Rhizomucor miehei lipase (RML) on multi-walled carbon nanotubes (MWCNTs) using three distinct coupling chemistries, including genipin (MWCNT/Gen@RML), glutaraldehyde (MWCNT/Glu@RML), and glyoxyl (MWCNT/Gly@RML). The resulting nanostructured biocatalysts were systematically evaluated for the selective enrichment of docosahexaenoic acid (DHA) and eicosapentaenoic acid (EPA) from commercial fish oil. The immobilized RML derivatives were characterized using FTIR, SEM, SEM-EDS, and TGA analysis. The maximum immobilized protein amounts were approximately 8.4, 8.1, and 8.6 mg g⁻1 support for MWCNT/Gen@RML, MWCNT/Glu@RML, and MWCNT/Gly@RML, respectively, when 10 mg of protein was initially loaded per gram of support. The optimal pH was 7.5 for free RML and all immobilized RML derivatives, and the optimal temperatures were 45 °C for free RML, 55 °C for MWCNT/Glu@RML, and 60 °C for MWCNT/Gen@RML and MWCNT/Gly@RML. Thermal stability improved markedly for all immobilized derivatives, increasing by approximately 34.6, 25.5, and 44.4 fold for MWCNT/Gen@RML, MWCNT/Glu@RML, and MWCNT/Gly@RML, respectively at 60 °C. Kinetic analysis indicated that MWCNT/Glu@RML achieved the highest catalytic efficiency (kcat/Km) of 19.2 mM⁻1 min⁻1, while MWCNT/Gen@RML exhibited superior reusability, retaining 75

This study investigates the relationship between wind speed and atmospheric constituents (nitrogen dioxide (NO2), ozone (O3), carbon dioxide (CO2), and aerosol index (AI)) in the Eastern Anatolia Region of Türkiye. The analysis focuses on how topographically constrained basins and low wind speeds influence the accumulation of atmospheric constituents. Pixel-based seasonal analyses were conducted using TROPOMI satellite data (2019–2024) and TerraClimate wind speed data. The methodological framework integrates Pearson correlation, time-lag analysis, and K-means clustering to quantify both instantaneous and delayed relationships between wind speed and atmospheric variables. Results indicate that NO2 and CO2 concentrations increased by 40–60

This study investigates the feasibility of using calcined pumice as a sustainable precursor for geopolymer production. Natural pumice was calcined at different temperatures (600, 750, and 900 degrees C) and durations (1, 2, and 4 h). The effects of calcination were evaluated through color change, particle size distribution, scanning electron microscopy, energy-dispersive X-ray spectroscopy, Fourier transform infrared spectroscopy, and X-ray diffraction. The results showed that calcination induced structural and mineralogical modifications in pumice, including increased disorder in the aluminosilicate network and partial recrystallization, which enhanced its reactivity. Consequently, geopolymer mortars produced with calcined pumice exhibited significantly improved compressive strength, with the highest strength of 53.5 MPa obtained for the sample calcined at 750 degrees C for 1 h, corresponding to an 84.5% increase compared to the mortar produced with raw pumice. In addition, calcination at 600 degrees C and 900 degrees C significantly improved water resistance. Considering mechanical performance, durability-related properties, and energy efficiency together, the calcination condition of 600 degrees C for 2 h was identified as the optimum treatment. These findings demonstrate that calcined pumice is a promising and sustainable precursor for geopolymer production.

Seed priming with bio-based agents is an effective approach to improving plant performance under environmental stress. This study evaluated the effects of seed priming with Arthrospira platensis on wheat growth under salt stress. The treatments included T0: control, T1: A. platensis seed priming, T2: NaCl (100 mM), and T3: A. platensis seed priming combined with NaCl (100 mM). In the greenhouse experiment, the effects of the treatments on oxidative stress markers (H2O2, MDA), antioxidant enzyme activities (SOD, CAT), proline, relative water content (RWC), chlorophyll content, and plant height were determined during the seedling (S1), tillering (S2), stem elongation (S3), and heading (S4) stages. Furthermore, the impact of the treatments on phenological development stages and morphological characteristics were also assessed. The results showed that A. platensis improved early seedling development, chlorophyll content and RWC. Notably, the regulatory effect of A. platensis on the accumulation of H2O2, SOD, CAT, proline, and MDA was maintained across the S1-S4 stages under salt stress conditions. Compared to T0 and T1, salt stress (T2) caused an overall reduction in phenological development day durations, whereas the effect of A. platensis (T3) was not significant. Moreover, A. platensis had a significant positive effect on tiller number, spike height, number of grains per spike, and thousand grain weights, even under salt stress conditions (T1, T3). Overall, A. platensis emerges as an effective and viable bio-based priming agent for supporting sustainable agriculture and stress tolerance.