The peripheral nerves of the human hand play a crucial role in sensorimotor function. However, their complex anatomy, multiscale biomechanics, and electrically coupled behavior make them difficult to model computationally. Recent advances in medical imaging, finite element modeling, and data-driven methods have enabled the development of increasingly realistic models of nerve mechanics and signal conduction. The results obtained from modeling nerve mechanics and electrical conduction can provide useful insight into the underlying mechanisms of several neuropathies such as carpal tunnel syndrome (CTS), diabetic neuropathy, and traumatic nerve injury. This systematic review examined key modeling strategies, including the use of anatomical reconstruction, finite element analysis (FEA) for stress and strain analysis, computer simulations of the electrophysiological activity of nerves to model nerve conduction, multi-physics analyses, and artificial intelligence (AI)-assisted modeling approaches. A systematic literature search was conducted in PubMed, IEEE Xplore, Scopus, and Web of Science, covering studies published up to December 2025 to identify studies related to computational and AI-assisted modeling of the human hand’s peripheral nerves. After removing duplicates, screening, and eligibility assessment, 50 studies were included. The literature was synthesized across four thematic areas: anatomical reconstruction and imaging-based models, biomechanical and finite element modeling, electrophysiological and conduction models, and AI-assisted data-driven approaches. Comparative analysis reveals substantial variability in the modeling assumptions, validation strategies, and clinical relevance, with limited consensus on standardized evaluation metrics. Persistent gaps include the inadequate representation of nerve branching networks, insufficient patient-specific connective tissue modeling, and weak experimental or clinical validation. Despite advances in peripheral nerve modeling, no comprehensive review exists focusing specifically on anatomically accurate computational modeling of the human hand’s nerves. Future work should improve the alignment of imaging and computational pipelines, develop hybrid electro-mechanical and AI-enhanced frameworks, and enhance translation into surgical planning and neuroprosthetics.

Drug classification is a key task in medical decision-making, and this process can be supported through appropriate selection of drugs that fit patient attributes and medical history. Currently, state-of-the-art solutions for this problem tend to make use of attribute-based representation and may overlook the relational structure that could exist in the data points. This paper seeks to examine a machine learning algorithm for multiclass drug classification, where the target variable is the prescribed type of drug, using a graph-based feature extraction method. The proposed method can be used thorough data preprocessing strategies, including handling imbalance and removal of outliers. Evaluation of the method requires a 10-fold cross-validation strategy to achieve a genuine and impartial evaluation. In each training set, a similarity graph is created and graph-based attributes such as degree centrality measures and clustering coefficient are extracted by using three different distance measures to consider graph relationships between samples. In the case of the testing data, graph attributes are computed by supplanting a weighted k-nearest neighbor strategy, wherein there is absolute prevention of information leakage. The graph attributes obtained along with other attributes are used to train different machine learning classifiers. The results show that the performance across all distance measures plays a great role for such classifications.

Preeclampsia is a pregnancy-associated disorder characterized by hypertension and multisystem maternal complications. Soluble fms-like tyrosine kinase-1 (sFLT-1) is widely used to establish experimental animal models, since elevated circulating sFLT-1 reproducibly induces endothelial dysfunction, hypertension, proteinuria, and glomerular injury. These findings indicate that angiogenic imbalance can generate the maternal syndrome of the disease. However, preeclampsia originates from abnormal placentation, whereas sFLT-1 primarily mediates the maternal systemic response. To clarify the biological meaning of these models, published sFLT-1 animal models were evaluated from four aspects: species differences, source of expression, placenta specificity, and isoform variation. The analysis showed that systemic sFLT-1 overexpression models reliably reproduce maternal vascular and renal manifestations but lack primary placental pathology. In contrast, placenta-restricted expression improves physiological relevance but still bypasses upstream mechanisms of trophoblast invasion failure. Therefore, sFLT-1 models do not represent the entire disease process and instead correspond to a specific pathological stage. These models are suitable for studying endothelial injury, renal pathology, and angiogenic-targeted interventions but are limited in investigating disease initiation, immune maladaptation, and clinical heterogeneity. Recognizing these boundaries can improve model selection and interpretation in experimental preeclampsia research.

This paper describes a novel diabetes prediction framework underpinned by a neural network architecture which features a parameterized activation function, called the Self-Learning Activation Linear Unit (SLALU). The proposed diabetes prediction model is composed of multiple self-learning activation linear units with parameters, and this SLALU can fit most of the traditional activation functions, including but not limited to the Rectified Linear Unit (ReLU) and the Parametric Rectified Linear Unit (PReLU). This self-adjusting activation approach provides a flexible bending ability, thereby outperforming the bending capabilities of traditional activation functions and reducing the intricacy involved in their manual selection. The design of our neuron model improves the network’s representational capacity and fitting ability. Deployed in diabetes prediction, our advanced network model demonstrated superior performance across a suite of metrics, thereby recording an accuracy of 98.18%, thus outperforming several existing models.

Peri-implantitis is a prevalent complication in dental implantology, often resulting from bacterial contamination on implant surfaces. Researchers have explored diode lasers as a promising alternative for bacterial decontamination. In this study, we evaluated the effectiveness of the 810 nm diode laser in eliminating Staphylococcus aureus from titanium implant surfaces. This investigation was designed as a controlled, standardized in vitro laboratory experiment. A total of 95 titanium implants were divided into the following groups: (1) A negative control group (n = 5): New integrity implants; (2) a positive control group (n = 6): Implants contaminated with a S. aureus suspension and left untreated; and (3) four intervention groups (n = 21 for each group): Implants contaminated and then treated with the 810 nm diode laser at varying power settings (1 W, 1.5 W, 2 W, 2.5 W) for 3 minutes. Bacterial load was quantified using colony-forming unit analysis, while surface roughness and morphology were assessed with a surface roughness tester and scanning electron microscopy (SEM) after laser treatment. The results showed that all intervention groups demonstrated significant reductions in bacterial load compared to the positive control group, with bacterial reduction increasing with higher laser power. The 2.5 W settings proved most effective in decontamination, although none of the laser settings completely eradicated the bacteria. Laser treatment did not result in significant changes in implant surface roughness, suggesting its safety for use in surface decontamination without compromising implant integrity. This technology may represent a safer, more effective approach to implant surface decontamination than traditional methods.

Athletes performing in complex environments rely on multisensory integration to maintain dynamic postural stability, with visual input playing a critical role in movement regulation and environmental perception. This reliance may be particularly pronounced in athletes with chronic ankle instability (CAI), who often exhibit proprioceptive deficits, impaired neuromuscular control, and ligamentous laxity. In this narrative review, we aimed to examine how visual distraction and dual-task load influence postural control and compensatory movement strategies in athletes with CAI while discussing potential implications for assessment and rehabilitation. Relevant English-language literature was identified through searches of PubMed, Scopus, and Web of Science using keywords related to chronic ankle instability, visual disturbance, blurred vision, dual-task interference, and postural control. Evidence suggested that visual disturbance may impair environmental perception and dynamic stability, whereas uncorrected refractive errors may further exacerbate postural control deficits. Athletes with CAI frequently demonstrate compensatory neuromuscular adaptations, including increased proximal muscle activation and altered landing biomechanics, which may increase mechanical loading on the knee and hip joints. Under dual-task conditions, increased attentional and cognitive demands may further impair sensorimotor processing, reaction time, and dynamic postural regulation, leading CAI athletes to adopt hip-dominant stabilization strategies. Collectively, these findings highlight the importance of integrating visual, cognitive, and proprioceptive assessments into rehabilitation and injury prevention programs. In addition, dual-task and sport-specific training paradigms may help improve dynamic stability, neuromuscular control, and injury resilience in athletes with CAI.