Search Results (363)

Reliable motor-current-based bearing diagnosis requires evaluation on unseen physical bearings and operating conditions. This study uses the Paderborn University benchmark, acquired from a 425 W permanent-magnet synchronous motor (PMSM) test rig, to evaluate time–frequency deep transfer learning under strict bearing-level grouping. Four representations—continuous wavelet transform (CWT), short-time Fourier transform (STFT), wavelet synchrosqueezed transform (WSST), and Fourier synchrosqueezed transform (FSST)—are combined with pretrained CNN backbones across four binary targets: aged-only A/B and artificial-plus-aged C/D, with mixed-fault bearings excluded/included within each pair. The workflow includes pooled-condition candidate discovery, exploratory Main-split leave-one-operating-condition-out (LOCO) screening, and a retrospective multi-split LOCO audit. The audit contains 288 crossed condition–split–seed evaluations. Because pooled test summaries and Main-split LOCO results informed later stages, these evaluations provide descriptive robustness evidence rather than an independent post-selection test. Target A achieved the highest all-split mean balanced accuracy (0.736 for CWT–EfficientNetB0). The pairs for Targets B and C were near-ties, and the Target D ordering reversed when Main was excluded. Across the eight audited candidates, mean sensitivity ranged from 0.618 to 0.948, whereas specificity ranged from 0.092 to 0.564. Target D combined approximately 0.89 sensitivity with an approximately 0.90 false-alarm rate. Thus, operating condition, fault-class composition, bearing split, and error-cost priorities all affect model interpretation. A matched current-domain baseline audit added 336 evaluations using handcrafted-feature RBF–SVM and Random Forest models and a compact raw-current 1D-CNN. The results show that instability is broader than the TF–CNN pipeline but is not uniform across model families: TF candidates were clearly stronger for Targets A and C, feature-based models were stronger for Target B, and Target D remained mixed and protocol-sensitive. A complementary bearing-level source-group analysis quantified six healthy/fault-source categories across the 37 current features; among the eight features with the largest mean between-group variance fraction, only spectral entropy and dominant power fraction preserved the same mixed-versus-non-mixed contrast direction across all four operating conditions. An additional matched Target D reference held the binary target, physical-bearing split, held-out condition, seed, fourth-order FSST representation, ResNet-50 backbone, and training settings fixed while changing the sensing channel. Across 24 matched runs, vibration showed descriptively higher mean balanced accuracy (0.642 vs. 0.503) and specificity (0.476 vs. 0.146), while sensitivity was slightly lower (0.807 vs. 0.859); the comparison does not establish universal modality superiority. Pooled-condition performance is useful for candidate discovery, but credible condition-generalization claims require explicit separation of exploratory selection and confirmatory testing. The numerical findings are specific to the evaluated PMSM benchmark and do not establish universal performance across electric-machine types. Full article

Magnetic shape memory alloys based on the Ni–Mn–Ga system are of strategic interest for aerospace and robotics applications due to their ability to respond to both thermal and magnetic stimuli. However, the NASA Shape Memory Materials Database a key resource for the community exhibits significant gaps in functional parameters, with up to 93.7% of records missing critical properties such as the Curie temperature, and over 88% lacking complete magnetic data. To address this limitation, this study proposes a data imputation strategy based on a stacking ensemble comprising twelve machine learning models (LGBM, XGBoost, CatBoost, GradientBoosting, RandomForest, MLP, BayesianRidge, KNN, SVR, GPR, MICE, and AutoEncoder), optimized via Optuna and evaluated using ten random seeds with 10 repetitions each. The approach was applied to reconstruct missing entries in NASA’s database. For heat treatment 1, the method achieved coefficients of determination ($R^2$) of 0.95 for duration (h) and 0.88 for temperature (°C), respectively. For the phase transformation temperatures (Mf, Ms, As, and Af), the method yielded $R^2$ values of 0.83, 0.82, 0.79, and 0.80, respectively. Magnetic properties saturation magnetization and maximum magnetic field were imputed with an $R^2$ of 0.92. In contrast, the Curie temperature exhibited limited predictive performance ($R^2$ = 0.15–0.35), primarily due to insufficient data availability. Overall, the proposed methodology integrates machine learning based imputation with physically supported constraints, providing a viable alternative to enhance the completeness and utility of materials databases. Full article

Agrochemicals are used worldwide in food production, but their use varies between countries due to the damage observed to nature and human health. Allelopathy is the primary pathway of chemical communication in plants, interfering with biome development through stimulation and/or inhibition mechanisms. Therefore, this study aimed to assess the biological activities of the ethanol extract (EE) and fractions of S. fluminensis leaves on monocotyledonous and eudicotyledonous models. The EE was obtained by percolation with ethanol, and the hexane (HEXF), dichloromethane (DCMF), ethyl acetate (EAF), and hydroethanol (HEF) fractions were obtained by liquid–liquid partition. The phytochemical characterization was performed by ¹H nuclear magnetic resonance (NMR). The allelopathic activity was evaluated on Allium cepa (onion) and Lactuca sativa (lettuce) seeds. The cytotoxic, genotoxic, and antigenotoxic effects on A. cepa meristematic cells were analyzed in vitro. Aliphatic compounds, saponins, and flavonoids derived from quercetin and kaempferol were characterized in the samples. All samples decreased the vigor, germination rate, and germination speed index (GSI) of A. cepa seeds. In contrast, they did not alter the vigor and viability of L. sativa seeds, but decreased the GSI, except for HEF. The samples inhibited the epicotyl and root growth of A. cepa and L. sativa, except HEXF, which stimulated the growth of epicotyls (750 µg/mL) and roots (750 and 1000 µg/mL). The cytotoxic assays showed that the EE and HEXF had cytotoxic action at low concentrations, and no sample showed a genotoxic effect. The following samples exhibited an antigenotoxic effect after pretreatment with atrazine (ATZ): EE (125 and 750 µg/mL), HEXF (750 and 1000 µg/mL), DCMF (125, 250, and 1000 µg/mL), EAF (at all tested concentrations), and HEF (125, 250, and 750 µg/mL). Furthermore, the EAF at 125 and 500 µg/mL and HEF at 125 µg/mL demonstrated the potential to reverse genetic damage induced by glyphosate (GLY). Full article

Broadband antenna, frequency and channel selection requires efficient prioritisation of finite candidate configurations, yet strong surrogate performance within a simplified analytical model does not ensure that the learned ordering will transfer to another electromagnetic representation. This study developed a validation-aware surrogate-shortlisting framework using a controlled breast-mimetic benchmark comprising 120 scenarios and 80 antenna–frequency–channel candidates per scenario. Surrogate models were developed using grouped scenario-level validation, and the model and five-candidate shortlist policy were frozen before external evaluation. Random forest produced the lowest-regret analytical-domain shortlist and remained stable across model-initialisation seeds. The frozen ranking was then challenged on held-out scenarios using a separately implemented restricted two-dimensional transverse-magnetic finite-difference time-domain model. Although numerical-reference checks supported shortlist-level use of the operational grid, candidate ordering did not transfer reliably: optimum inclusion, shortlist agreement and rank association were weak, although a qualified candidate within 2 mm of the finite-library FDTD optimum was retained in 58.3% of cases. Transfer failure varied by frequency band and channel family, while incomplete alignment between the analytical and FDTD candidate libraries prevented attribution of the discrepancy to electromagnetic-model shift alone. The framework therefore positions analytical surrogates as auditable shortlisting tools that reduce downstream candidate-level assessment while retaining independent electromagnetic evaluation before final design selection. Full article

Mechanical speed sensors increase cost and may reduce the reliability of permanent magnet synchronous motor (PMSM) drives under harsh conditions. This paper proposes a validation-protected physics-consistent probabilistic neural estimator for sensorless PMSM speed estimation using only online-deployable signals: the previous estimated speed, measured d/q-axis currents, and commanded d/q-axis voltages. A multi-output probabilistic network predicts the speed distribution and auxiliary residual-compensation variables. Mechanical consistency, electrical consistency, and regularization losses are imposed during training, while a validation-protected rule selects, for each random seed, the checkpoint with the lower validation RMSE from the paired baseline and physics-trained candidates. Experiments use a frozen multi-seed protocol covering locked holdout evaluation, independent comparison, and disturbance tests. Across Datasets 8–11, the frozen predictive distributions yielded Gaussian NLL values from 3.190 to 3.239, 100% empirical coverage of the nominal 95% prediction intervals, and mean interval widths of approximately 35.9 rad/s, indicating conservative rather than well-calibrated uncertainty. On locked Dataset 7, Physics-safe reduced the mean RMSE from 3.415 to 3.277 and the inter-seed standard deviation from 0.290 to 0.052. Results on Datasets 8–11 show that the method is not universally mean-error optimal; its recurring advantage is lower inter-seed variability and more reproducible training outcomes. A local sensitivity analysis on Dataset 4 confirmed seed- and loss-weight-dependent physics-training outcomes, supporting the need for validation protection without implying globally optimal loss weights. On the specified desktop CPU using ONNX Runtime, the complete recursive estimator step required 40.154 microseconds, below the adopted 100-microsecond sampling interval, supporting estimator-level computational feasibility. Full article

Background: Radiotracer- and magnetic seed-based targeted axillary staging techniques are increasingly used after neoadjuvant chemotherapy (NAC) in patients with initially node-positive breast cancer who convert to clinical node-negative (ycN0) status. However, these technologies remain unavailable in many centres worldwide. We evaluated the effectiveness of radiotracer-free axillary staging using dual-tracer sentinel lymph node biopsy (SLNB) with indocyanine green (ICG) and methylene blue, with particular focus on the clinical implications of clipped node non-retrieval. Methods: This retrospective single-centre cohort study included 49 patients with biopsy-proven cN1 breast cancer who achieved ycN0 status following NAC and underwent dual-tracer SLNB using ICG and methylene blue between 2021 and 2024. Clipped node retrieval rates, tracer-specific detection patterns, recurrence outcomes, and disease-free survival (DFS) were analysed. Clinical outcomes, including recurrence and disease-free survival, were compared between patients with successful and unsuccessful clipped node retrieval. Results: Clipped node retrieval was achieved in 41 of 49 patients (83.7%). ICG outperformed methylene blue for clipped node identification (80.5% vs. 61.0%) and was the sole detecting tracer in 39.0% of cases. Using methylene blue alone would have reduced the retrieval rate to 51.0% (p = 0.001). ICG identified the clipped node in all six patients with axillary micrometastases (ypN1mi). Clipped node non-retrieval occurred in eight patients, all of whom had negative SLNB findings and did not undergo completion axillary lymph node dissection. Recurrence rates were similar between the retrieval and non-retrieval groups (17.1% vs. 12.5%, p = 1.000), and no isolated axillary recurrences were observed. Baseline tumour size was the only independent predictor of disease-free survival in multivariable analysis (HR 2.05, 95% CI 1.20–3.51; p = 0.008). Conclusions: Radiotracer-free axillary staging using ICG and methylene blue achieved a high clipped node retrieval rate in patients with ycN0 breast cancer after neoadjuvant chemotherapy. ICG significantly improved clipped node identification compared with methylene blue alone and identified the clipped node in all patients with residual nodal micrometastatic disease. Among patients with negative SLNB findings, clipped node non-retrieval was not associated with higher recurrence rates or worse disease-free survival during follow-up. However, given the small number of clipped node non-retrieval cases (n = 8) and recurrence events (n = 8), these findings should be interpreted with caution and require confirmation in larger prospective studies. These findings support the feasibility of a radiotracer-free dual-tracer approach in settings where radiotracer-based techniques are not available. Prospective multicentre studies are warranted to validate both the oncological performance of radiotracer-free axillary staging and the clinical implications of clipped node non-retrieval. Full article

Standard pushers for the relativistic equations of motion of a charged particle in an electromagnetic field—Boris, Vay, Higuera–Cary—do not, in general, preserve the full symplectic structure of the underlying Hamiltonian system, while high-order non-symplectic schemes such as Runge–Kutta accumulate secular error over long times. We propose a two-stage, symmetry-preserving framework (SP-PINN) for the 3+1-dimensional relativistic dynamics of a charged particle in a prescribed field, including a focused Gaussian laser pulse, that pairs a physics-informed neural network with an explicit symplectic integrator: the network learns a surrogate relativistic Hamiltonian, while the integrator—which is not itself learned—advances it. In Stage 1, an unsupervised physics-informed neural network learns the surrogate from the covariant equations of motion using a Lorentz-invariant loss that enforces the mass-shell constraint $\mathcal{H}=m{c}^2\gamma$; in Stage 2, the surrogate is advanced with an explicit symplectic map built on Tao’s extended phase space, valid for the non-separable relativistic Hamiltonian. To isolate the geometric integrator from neural-network approximation error, every benchmark figure advances the analytic relativistic Hamiltonian through Stage 2, the learned Stage-1 surrogate being assessed separately. We benchmark against the Boris pusher and Runge–Kutta on three core test problems (adding the Higuera–Cary pusher in the symplecticity diagnostic), supplemented by plane-wave, ensemble, and pulse-family studies, and we measure the first Poincaré–Cartan loop invariant directly as a quantitative diagnostic of symplecticity. The magnetic-field test illustrates the contrast between bounded and secular error growth: Runge–Kutta drifts secularly, the Boris pusher conserves the invariants to machine precision as a volume-preserving gyro-integrator, and the symplectic map keeps the error bounded for all time; on a non-integrable magnetic trap, where no exact volume-preserving rotation exists, the symplectic map alone keeps the energy error bounded. The learned surrogate is the current accuracy bottleneck—not yet competitive with the conventional pushers for the static cases—but for the demanding laser case, a vector-potential light-cone reformulation reduces this surrogate error to $(3.0\pm 0.1)\times {10}^{-4}$ (three seeds) and yields learned trajectories that remain phase-coherent over essentially the whole interaction. The framework targets laser–plasma acceleration, synchrotron-radiation modeling, and particle tracking. Full article

The propagation of native plants that support pollinators is a relevant strategy for strengthening seed banks, pollinator gardens, and local restoration efforts. This study evaluated whether pre-sowing magnetic exposure modifies germination, early survival, and initial seedling growth in two Salvia species with ecological and biocultural value from Michoacán, Mexico. A total of 1000 seeds, equally distributed between both species, were analyzed. The seeds were assigned to a non-exposed control group and magnetic field exposure treatments at 83.94, 166.94, 292.30, 305.40, 350.70, or 375.10 mT for 5, 10, 15, or 20 min. Germination, survival, time to radicle emergence, and seedling height on day 13 were recorded. Although magnetic exposure did not significantly increase germination, seedlings from exposed seeds were taller on day 13 than those in the control group. Furthermore, higher seed mass increased germination probability. These results suggest that magnetic exposure cannot be considered a direct strategy to increase germination in these species but could be explored as a complementary tool to promote early seedling growth in native plants intended for conservation, environmental education, and restoration. Full article

Foot-worn electromagnetic harvesters are driven by irregular rigid-body motion, while their response is limited by mechanical stroke, coil geometry, mounting direction, and the electrical interface. This paper presents a reproducible numerical design study of a finite-stroke magnetic quasi-zero-stiffness (QZS) moving-magnet harvester. Two public three-axis foot-IMU records are processed with stated gyroscope-bias estimation, six-axis attitude estimation, gravity removal, residual-offset correction, filtering, and angular-acceleration calculation. Three explicit axes are used in the design screen, and the selected candidate is then evaluated over a 62-direction spherical grid. Rigid-body angular-acceleration and centripetal terms are included for specified sensor-to-harvester offsets. Two normalized magnetic force laws are compared. The electrical model uses position-dependent flux linkage, explicit series connection and polarity of coil sections, winding-derived resistance, and a position-dependent electromagnetic reaction force. A fixed-seed random screen evaluates 720 geometry-constrained candidates. The highest-ranked nominal candidate is a 150 mm external foot-worn module with a 40.6 g moving mass, a 30 mm hard half-stroke, 1649 turns in two series sections, and a 25.27 mm coil outer diameter. Across 72 design-screen cases formed from 12 five-second windows, three mounting axes, and two magnetic laws, this candidate remained hard-stroke- and design-stroke-safe. Its conditional ideal load-side power had a 10th percentile of 1.38 mW and a median of 2.03 mW. In the 62-direction check, all 1488 cases remained hard-stroke-safe; two opposite directions each produced one design-stroke exceedance, with a maximum displacement of 24.15 mm. Re-ranking all 30 Stage-2 candidates under coupling and magnetic-stiffness changes retained the long geometry family, although a 30% coupling reduction changed the highest-ranked candidate from 600 to 632. Soft-stop sensitivity, equation-level consistency, and multi-case Runge–Kutta convergence are also reported. The results support finite-stroke design screening, but they do not constitute prototype, finite-element, or delivered-power validation. Full article

This study used black soldier fly protein (BSFP) as a base material and added 0%, 1%, 2%, 3%, 4%, and 5% of grape seed anthocyanidins (GSAs) to prepare composite gels. Through the combined use of low-field nuclear magnetic resonance, Fourier transform infrared spectroscopy, scanning electron microscopy, and rheometry, the relationships among GSA dosage (0–3%), gel structural properties (secondary protein conformation, water status, and microscopic morphology), and rheological printability were systematically evaluated. It was found that the better GSA content fell within 1–3%, and under this condition the extrusion-type 3D printing performance of the composite gels was significantly enhanced. At a 3% addition amount, the proportion of disordered conformations decreased (random coiling decreased from 15.93% to 15.46%), the ordered structure increased (β-sheet increased from 35.25% to 35.43%), and deformation resistance was enhanced. Low-field nuclear magnetic resonance showed an increase in the proportion of non-flowing water and an increase in physical constraints. Scanning electron microscopy showed a reduction in pore size and a thickening of pore walls, forming a denser 3D network. Rheologic analysis indicated that 3% GSA reached the maximum zero-shear viscosity (η₀) and that the storage modulus (G′) and loss modulus (G″) were higher in the experimental group than those in the control group. Printing fidelity increased from 45.73% in the control group to 60.08% in the 1% group, 62.14% in the 2% group, and 71.05% in the 3% group (p < 0.05). The 3–5% groups (fidelity: 71.05–75.66%) all achieved hollow cylindrical printing without collapse and had excellent self-supporting performance. However, excessive addition (4–5%) caused excess GSA to adsorb onto the protein skeleton surface, reducing the apparent viscosity and damaging the printing performance. Based on all the indicators, the composite gel with 3% GSA achieved the best balance between printability and structural integrity. Our research offers a new idea for using flavonoid compounds to improve the 3D printing performance of insect protein gels. The prepared composite gels can be used as food printing inks and applied to personalized nutrition customization, functional food development, and sustainable protein alternative product fields. Full article

Citrus fruits are a significant source of flavonoids. Of all the citrus fruits, Citrus paradisi (grapefruit) presents the highest concentration of the flavonoid naringin, a compound offering a variety of human health benefits and applications in the pharmaceutical, food, and cosmetic industries. Commonly, when a citrus fruit is consumed, the peel and seeds are discarded, resulting in approximately 50% waste, making the potential use of citrus waste in order to reduce environmental impact a research priority. The present study used fresh grapefruit albedo to extract naringin via eco-friendly methods, such as ultrasound-assisted extraction (UAE) and microwave-assisted extraction (MAE), which were compared against the conventional reflux extraction procedure. Furthermore, the presence of naringin was confirmed by nuclear magnetic resonance (NMR) spectroscopy, while naringin content was determined via HPLC-DAD analysis. The results obtained show that the pink grapefruit variety was the optimal source for extracting the flavonoid of interest, producing the highest content (3.41 g/kg), followed by the red (2.47 g/kg) and white (1.70 g/kg) varieties. The UAE method was observed to reduce the extraction time significantly, to only 10 min, which is up to 30-and -fold less than the extraction times obtained using conventional (5 h) and MAE (40 min) methods, respectively. These results prove the usefulness of UAE as a simple, fast, efficient, and eco-friendly method for extracting naringin from fresh grapefruit albedo, via the use of a green solvent such as ethanol. In addition, the present study is the first to conduct a comparative analysis of naringin content in the three main grapefruit varieties grown in Mexico. Full article

Background/Objectives: Epilepsy is a chronic brain disease characterized by recurrent epileptic seizures. It affects roughly 50 million people worldwide and around one third of the patients have drug-resistant epilepsy (DRE). The current study aimed to find differences in the whole-brain functional connectivity (FC) in patients with DRE compared to patients with well-controlled epilepsy (WCE) and healthy controls (HCs). Methods: This explorative, cross-sectional study included 92 participants (n_DRE = 30; n_WCE = 30; n_HC = 32) who underwent resting-state functional magnetic resonance imaging (fMRI). The CONN Toolbox was used to process and analyze the FC changes among the three groups. Results: There was a statistically significant increase of the FC between the left lateral prefrontal cortex, left inferior temporal gyrus (temporo-occipital), left lobules IV and V of the cerebellum and multiple cortical and subcortical structures in patients with DRE as opposed to WCE and HC. On the other hand, decreased FC was observed between three seeds (the posterior cingulate cortex, precuneus cortex, the right planum polare) and different frontal, temporal and occipital regions. Interestingly, the right nucleus accumbens (r_NAc) showed increased FC with the inferior frontal gyrus in DRE compared to WCE, whereas the r_NAc-left precentral gyrus FC was reduced in DRE as opposed to HC. Conclusions: The acquired information offers valuable insights into the neuronal networks associated with DRE. These data could be used for advancing diagnostic accuracy and future therapeutic strategies. Full article

Conventional water treatment relies on sequential flocculation and disinfection, which inflates infrastructure costs and heightens the risk of disinfection byproduct formation. Here, we report a magnetic lignin-based flocculant (LS-DMC-AM@Fe₃O₄) that integrates flocculation, sterilization, and rapid magnetic separation within a single material. The composite was synthesized by thermally initiated graft copolymerization of methacryloyloxyethyl trimethylammonium chloride (DMC) and acrylamide (AM) onto sodium lignosulfonate (LS), followed by incorporation of Fe₃O₄ nanoparticles (NPs) at 15 wt% loading; the product exhibited a saturation magnetization of 12.8 emu g⁻¹. LS-DMC-AM@Fe₃O₄ achieved 98.2% kaolin turbidity removal at 1 mg L⁻¹ and 98.6% E. coli removal at 8 mg L⁻¹, and displayed a markedly broader effective dosage window than its non-magnetic analog. We attribute this broadened window to Fe₃O₄-enhanced membrane disruption, which liberates anionic intracellular contents that buffer excess cationic charge and thereby suppress restabilization. The bactericidal efficiency reached 90% at 18 mg L⁻¹, 1.6-fold higher than LS-DMC-AM, governed by a synergistic dual mechanism: quaternary ammonium contact-killing coupled with Fe₃O₄ NP-induced intracellular reactive oxygen species (ROS) accumulation. Under an external magnetic field, flocs underwent rapid phase separation and displayed enhanced shear-regrowth capacity (E. coli floc recovery factor: 53% vs. 26%); Fe₃O₄ NPs were recovered at >95% efficiency over two cycles. Despite higher unit production costs, LS-DMC-AM@Fe₃O₄ delivers competitive per-unit-volume treatment economics through its ultralow effective dosage and magnetic seed recyclability. These results establish a viable strategy for engineering multifunctional, recyclable flocculants from industrial lignin waste. Full article

Accurate, automated delineation of adult diffuse gliomas from multi-parametric magnetic resonance imaging (mpMRI) is central to quantitative neuro-oncology. Volumetric 3D networks dominate the BraTS leaderboard but require expensive GPUs, long training cycles, and provide diminishing returns relative to their compute budget. Slice-wise 2D models, by contrast, discard inter-slice context that is informative for thin tumor rims and small enhancing foci. We introduce GDNet, a 2.5D multimodal MRI segmentation framework for adult glioma evaluated on the BraTS 2024 cohort. GDNet consumes a stack of three adjacent axial slices from the four standard BraTS modalities (T1, T1ce, T2, FLAIR) as a 12-channel input to a compact U-shaped encoder–decoder with Group Normalization and predicts whole tumor (WT), tumor core (TC), and enhancing tumor (ET) masks for the central slice. The training pipeline pairs the 2.5D backbone with: (i) Exponential Moving Average (EMA) of model weights with decay 0.999, (ii) mixed tumor-aware slice sampling (p_tumor = 0.50), (iii) a compound Cross-Entropy + Soft-Dice loss, and (iv) AdamW with warm-up plus cosine annealing under Automatic Mixed Precision. We performed a systematic, step-by-step ablation covering a 2D baseline, EMA + mixed sampling, tumor-centered crop fine-tuning, a GDNet-inspired architectural integration, a region-aware loss, 3-slice and 5-slice 2.5D inputs, and connected-component post-processing, and we report multi-seed results to quantify reproducibility. On the held-out BraTS 2024 test partition, the final 3-slice 2.5D GDNet achieved positive-only Dice scores of 0.791 ± 0.000 (WT), 0.736 ± 0.003 (TC), 0.654 ± 0.004 (ET), and a mean foreground positive-only Dice of 0.820 ± 0.000 across seeds; the all-slice mean foreground Dice exceeded 0.927 ± 0.000. Validation positive-only scores were 0.805 ± 0.002 (WT), 0.757 ± 0.004 (TC), 0.683 ± 0.009 (ET). The inter-seed standard deviation was small for every region (≤0.01 Dice points), indicating low inter-seed variance across the two seeds evaluated; with only two seeds, we regard this as preliminary evidence of training stability rather than a strong reproducibility claim. The ablation isolated EMA + mixed tumor sampling and the 2.5D context window as the dominant sources of improvement; notably, a GDNet-style architectural integration with a region-aware loss did not outperform the simpler 2.5D U-Net on positive-only WT/TC/ET, and light post-processing improved only all-slice Dice. A failure-mode audit found that the residual catastrophic predictions are concentrated on a small minority of diffuse, infiltrative tumors with mass effect. Conclusions: Carefully engineered training strategies, tumor-aware sampling, EMA stabilization, and a modest 2.5D context window recover a substantial fraction of the accuracy of much heavier 3D networks at a fraction of the compute, are reproducible across seeds, and outperform a heavier GDNet-inspired architectural variant on the same data. GDNet is therefore a practical and, pending external validation, potentially clinically deployable framework for multimodal glioma segmentation on workstation-class GPU hardware. Full article

Background: Precise intraoperative localisation of small colorectal tumours during laparoscopic surgery remains challenging due to absent tactile feedback and subserosal tumour location. Current standard methods, particularly India ink tattooing, demonstrate 15–30% failure rates for lesions less than 10 mm, leading to prolonged operative times, incomplete resections, and re-operations. Multiple emerging technologies promise improved localisation, yet comparative evidence remains fragmented. Objective: To map and characterise the current landscape of intraoperative marking and identification technologies for small colorectal tumour localisation during laparoscopic surgery, with emphasis on radiofrequency-based methods and alternative approaches, and to identify evidence gaps guiding future research. Methods: Following PRISMA-ScR guidelines, we systematically searched PubMed, Web of Science, and Scopus databases from January 2000 through December 2025 for studies evaluating tumour localisation technologies in colorectal cancer surgery, including primary tumour localisation during laparoscopic colectomy and localisation of colorectal liver metastases during hepatic surgery, or transferable anatomical applications with documented translational potential to colorectal surgery. Two independent reviewers screened all records, with discrepancies resolved through discussion and a third senior reviewer consulted for unresolved disagreements; data were extracted on technical performance, safety, feasibility, cost-effectiveness, usability, innovation potential, and evidence quality. Results: We included 89 studies comprising 18 colorectal-specific articles and 71 transferable/GI-adjacent studies. Detection success rates ranged from 71% to 100% across modalities. Near-infrared fluorescence with indocyanine green demonstrated the strongest clinical evidence with 75–100% detection across eight colorectal studies encompassing 2134 procedures and seamless workflow integration. Radiofrequency identification systems achieved 91.9–99% detection in feasibility studies with promising tissue penetration of 15–35 mm but limited colorectal validation. Electromagnetic navigation excelled in rigid organs with 85–98% success but showed degraded performance in mobile bowel at 71–75%. Critical evidence gaps included absent head-to-head comparative trials, non-standardised outcome metrics limiting cross-study comparability, and limited long-term safety data with only 14 studies providing follow-up exceeding six months. Conclusions: ICG fluorescence represents the most clinically mature technology identified, representing a priority candidate for colorectal-specific validation in challenging localisation scenarios. RFID systems demonstrate promising characteristics justifying prioritised research investment through adequately powered comparative trials. Future research must emphasise consortium-based comparative effectiveness studies, standardised outcome metrics, and integration with robotic and AI-assisted surgical platforms to accelerate clinical translation. Full article