17315 research outputs found
Sort by
Localized drug delivery via tramadol-loaded chitosan/hydroxyapatite scaffolds incorporating mesoporous SiO2-HA particles
No full textThis study investigates chitosan/hydroxyapatite (CS/HA) composite scaffolds integrated with mesoporous SiO2-HA particles for drug delivery applications. Scaffolds were fabricated using freeze-drying, resulting in a highly porous structure with interconnected pores. Tramadol was chosen to examine the encapsulation efficiency, release performance, and release kinetics of the samples. Morphological images confirmed that adding HA and mesoporous particles to the CS scaffold resulted in the formation of smaller pores with a more uniform geometry. The presence of mesoporous SiO2-HA particles in the scaffolds also significantly reduced the release of Ca ions due to their strong silane bonds. Moreover, this unique porous structure provided a drug loading capacity of 93 %, compared to 68 % in non-composite scaffolds. The composite scaffolds maintained complete release capability, with about ∼98 % of tramadol released from the CS/HA scaffolds incorporated with 2 wt% mesoporous SiO2-HA during 336 h. Kinetic investigations indicated the best fit with the Higuchi model as the drug release mechanism. Finally, the effect of tramadol-loaded scaffolds on cellular growth was thoroughly examined using various methods. The performance of tramadol-loaded scaffolds as a painkiller on osteosarcoma SAOS-2 cells demonstrated impressive biocompatibility. According to the findings, CS/HA scaffolds containing mesoporous SiO2-HA particles are strongly recommended as effective drug delivery systems for biomedical applications, particularly in bone tissue engineering
Multi-layer domain generalization for the semantic segmentation of optical remote sensing images
No full textSupervised learning is the dominant training paradigm of semantic segmentation. However, it assumes that the deployment (or target) data follow the same distribution as the training data. And unfortunately this assumption is often violated in practice, leading to the infamous domain shift, coupled with degraded segmentation performances. Domain generalization addresses this issue, by exploiting the source (or training) domains, so as to maximize performance with unseen target data. This paper focuses on domain generalization for the semantic segmentation of optical remote sensing images through an adversarial strategy and investigates the concurrent use of feature maps from multiple layers, and proposes enforcing progressive domain invariance from earlier to latter network layers. The proposed approach is validated with the FLAIR dataset, where it achieves superior performance w.r.t. state-of-the-art studies
Dynamic analysis of OOK modulators using eye diagrams to assess data rate and output matching
No full textThis paper presents a high-speed On-Off Keying (OOK) modulator in 65 nm CMOS, evaluated at data rates up to 50 Gbps. Eye diagrams are generated via a Hilbert transform from transient simulation data, enabling direct measurement of rise/fall times and eye openings at the modulator output. For 50 Gbps operation, the well-matched design achieves a rise time of 13.59 ps and a fall time of 17.95 ps, whereas the loosely matched design exhibits 25.34 ps and 28.00 ps, respectively. To assess the impact of communication-chain bandwidth, band-pass filters with 200 GHz, 150 GHz, and 100 GHz bandwidths - chosen based on covering the main lobe and varying extents of the side lobes - were introduced. As the bandwidth decreases, rise times for both matching conditions converge, reaching 35.47 ps (well-matched) and 34.72 ps (loosely-matched) at 100 GHz, suggesting bandwidth constraints dominate over matching. However, the fall time difference persists, with 21.44 ps and 24.24 ps at 100 GHz for well-matched and loosely matched cases, respectively, underscoring the importance of output matching. Additionally, the normalized eye height narrows more significantly in the loosely matched case
Geometry-induced asymmetric level coupling
No full textTailoring energy levels in quantum systems via Hamiltonian control parameters is essential for designing quantum thermodynamic devices and materials. However, conventional approaches to manipulating finite-size quantum systems, such as tuning external fields or system size, typically lead to uniform shifts across the spectrum, limiting the scope of spectral engineering. A recently introduced technique, known as the size-invariant shape transformation, overcomes this limitation by introducing a new control parameter that deforms the potential landscape without altering the system's size parameters, thereby enabling nonuniform scaling of energy levels. This new degree of freedom-referred to as the shape parameter-gives rise to quantum shape effects in the thermodynamics of confined systems, which are conceptually distinct from quantum size effects. Here, we explore the fundamental limits of nonuniform level scaling in the spectra by asking: what is the minimal quantum system in which such behavior can arise? We demonstrate that even a two-level system can exhibit the thermodynamic consequences of quantum shape effects, including spontaneous transitions into lower-entropy states, a phenomenon absent in classical thermodynamics for noninteracting systems. We identify the spectral origin of these unconventional thermodynamic behaviors as geometry-induced asymmetric level coupling, in which the ground-state energy and energy gap respond in opposite ways to changes in a shape parameter. This asymmetry naturally extends to many-level systems, where the thermally averaged energy spacing and ground-state energy evolve in opposite directions. To characterize unconventional thermodynamic behaviors, we construct thermodynamic spontaneity maps, identifying regions of energy-driven and entropy-driven spontaneous processes in ground-state energy versus energy gap space. These effects emerge under quasistatic, isothermal changes of a shape degree of freedom and illustrate how the confinement geometry alone can enable unconventional thermodynamic behaviors that are otherwise exclusive to interacting or open systems. We argue that any scaling-invariant local parameter transformation that induces asymmetric-level coupling can be used to engineer similar responses, making this a broadly applicable framework. Our results deepen the theoretical foundations of the quantum shape effect and introduce a route to spectral gap control, with potential applications in isolating computational subspaces within quantum information platforms
Palpation characteristics of an instrumented virtual cricothyroidotomy simulator
No full textCricothyroidotomy (CCT) is a critical, life-saving procedure requiring the identification of key neck landmarks through palpation. Interactive virtual simulation offers a promising, cost-effective approach to CCT training with high visual realism. However, developing the palpation skills necessary for CCT requires a haptic interface with tactile sensitivity comparable to human fingers. Such interfaces are often represented by plastic partial mannequins, which require further adaptation to integrate into virtual environments. This study introduces an instrumented physical palpation interface for CCT, integrated into a virtual surgical simulator, and tested on 10 surgeons who practiced the procedure over a training period. Data on haptic interactions collected during the training was analyzed to evaluate participants’ palpation skills and explore their force modulation strategies about landmark identification scores. Our findings suggest that trainees become more precise in their exploration over time, apply greater normal forces around target areas. Initial landmark identification performance influences adjustments in the overall applied pressure
PoseViTNet: multi-scene absolute pose regression using vision transformers
No full textAccurate camera pose estimation is crucial for autonomous driving and vehicle networking. Traditional pipelines based on geometric models and feature matching struggle in dynamic, featureless environments which are common in many environments. Inspired by the success of vision transformers (ViT), our approach uses a ViT backbone with an attention-based mask to extract a global image descriptor, which is then passed through fully connected layers for pose regression. The multi-headed self-attention in ViT helps the model learn scene layouts and focus on relevant features. We introduce an attention mask to improve performance in challenging scenes, especially dynamic or featureless ones. We compare three backbones: ViT (multi-headed self-attention throughout), ConViT (self-attention in the last two layers, gated positional self-attention elsewhere), and ResNet (pure convolution). We evaluate our model on two commonly used benchmarks for outdoor and indoor localization and we show that our model which uses ViT backbone achieves the state of the art results for both indoor and outdoor multi-scene absolute localization benchmarks
Thienothiophene and single-wall carbon nanotube-based hybrid materials: design, photophysical properties and the construction of high-performance supercapacitors
No full textSupercapacitors are widely accepted to be highly promising for energy storage due to their high capacitance and power density with super-long cycling stability. In addition, flexible and binder-free nanomaterials play a crucial role in supercapacitor devices and systems. Herein, we present thienothiophene (TT) and single-wall carbon nanotube (SWCNT)-based two hybrid materials, possessing triphenylamine (TPA), thiophene (Th) and EDOT moieties, i.e.TT-Th-TPA-SWCNT and TT-EDOT-TPA-SWCNT, as highly efficient supercapacitors with flexible and free-standing properties. The nanohybrids were obtained by noncovalent modifications of SWCNTs without using any binding agents. Their hybrid electrodes displayed remarkable supercapacitor performances and energy storage properties with an excellent power density of 10 000 W kg−1 at 20 A g−1, a maximum energy density of 5.19 ± 0.13 Wh kg−1 at 0.1 A g−1 and a maximum specific capacitance of 158 F g−1 at 1 mV s−1. Regarding the GCD results, 10 000 cycle stability was achieved with a coulombic efficiency of over 95%. These findings highlight the potential of TT and SWCNT-based hybrid materials as advanced electrodes in energy storage applications
Harnessing the photothermal properties of table olives: a near-infrared light strategy for enhancing the shelf-life of low-salt naturally fermented black olives
No full textThis study introduced a novel, non-chemical strategy for enhancing the microbial safety of low-salt, naturally fermented black table olives by harnessing their intrinsic light-to-heat conversion properties. Spectroscopic analysis revealed broad absorbance in the visible and near-infrared (NIR) regions, confirming that the olives acted as light-to-heat converters due to their rich phenolic and pigment content. When exposed to NIR light sources olive samples reached surface temperatures up to 80 °C within 7 min, providing the conditions required for localized photothermal inactivation of surface microorganisms. Artificial contamination experiments with Clostridium butyricum at an initial inoculum of ∼108 CFU/mL showed a ∼ 4-log reduction after a 4-min NIR light treatment, with microbial counts reduced below the detection limit (<1 log CFU/g) in some samples. Listeria monocytogenes viability decreased by ∼2.5 log units when irradiated on olive surfaces (∼106 CFU/g), but no reduction was observed when the bacterial suspension alone was irradiated, confirming that inactivation occurred through an olive-mediated photothermal mechanism. SEM analysis revealed membrane disruption in treated cells, supporting heat-induced rather than non-thermal damage. Physicochemical analyses confirmed that the NIR light treatment did not significantly alter key quality attributes, including pH, color, firmness, dry matter, oil content, acidity, and total phenolic content. Sensory evaluation showed no significant differences in bitterness, acidity, firmness, fibrousness, or crunchiness, with only a minor reduction in perceived saltiness. A six-month ambient storage study (22–25 °C) revealed that a single pre-treatment with NIR light reduced total viable bacterial and mold counts by up to 2-log units compared to untreated controls, while preserving all physicochemical and sensory characteristics. These results demonstrated that NIR light irradiation is an effective, clean-label intervention capable of reducing microbial risks and maintaining product quality in low-sodium table olive formulations
Just kidding? Exploring the role of traditional versus counter-traditional gender role jokes on gender identity threat
No full textWhile traditional gender roles have been examined in the context of online communication, less is known about the implications of encountering counter-traditional gender roles (e.g., depicting men as caring and women as independent) on social media. We investigated participants' perceived gender identity threat upon exposure to traditional versus counter-traditional gender role jokes that targeted either men or women. An online experiment (N = 265) using a 2 (content: traditional versus counter-traditional gender roles) by 2 (joke's target gender: men versus women) by 2 (participant's gender: man versus woman) mixed design demonstrated that overall, jokes targeting women elicited greater identity threat and women perceived greater identity threat than men. Moreover, the three-way interactions showed that women, in particular, perceived greater identity threat from traditional gender role jokes targeting women. The current study highlights the damaging role of the spread of traditional gender roles through humor, particularly for women's gender identities
Going forward-forward in distributed deep learning
Get PDFWe introduce a new approach in distributed learning, build- ing on Hinton’s Forward-Forward (FF) algorithm to speed up the train- ing of neural networks in distributed environments without losing ac- curacy. Unlike traditional methods that rely on forward and backward passes, the FF algorithm employs a dual forward pass strategy, eliminat- ing the dependency among layers required during the backpropagation period, which prevents efficient parallelization of the training process. Although the original FF algorithm focused on its ability to match the performance of the backpropagation algorithm, this work aims to re- duce the training time with pipeline parallelism. We propose three novel pipelined FF algorithms that speed up training 3.75 times on the MNIST dataset while maintaining accuracy when training a four-layer network with four compute nodes. These results show that FF is highly paral- lelizable and its potential in large-scale distributed/federated systems to enable faster training for larger and more complex models