Surface electromyography (sEMG)-based hand gesture recognition has been widely applied in prosthetic control and rehabilitation engineering; however, classification performance may be overestimated depending on the data partitioning strategy. Therefore, this study analyzed the hand gesture classification performance and misclassification characteristics of a one-dimensional convolutional neural network (1D CNN) under a repetition-based data partitioning scheme designed to minimize information leakage. Experiments were conducted using Exercise A data from subjects S1-S10 of the Ninapro DB1 dataset. The sEMG signals were segmented into 300 ms windows, and the model was trained using class imbalance handling, light data augmentation, and segment-level probability smoothing. In addition, repetition-level three-fold validation and an ensemble model comparison were performed to evaluate the model's generalization performance. The proposed 1D CNN achieved an average accuracy of 87.13 ± 2.17% and a Macro-F1 score of 86.66 ± 2.15%. In contrast, the ensemble model combining the 1D CNN and SE CNN yielded only a marginal improvement of approximately 0.05 percentage points despite increased model complexity. Repetition-level three-fold validation resulted in an average accuracy of 82.81 ± 5.48%, indicating that classification performance varied depending on the selected repetition combinations. Class-wise analysis further revealed that most misclassifications occurred among thumb adduction, abduction, and flexion movements. These results demonstrate that a 1D CNN combined with probability smoothing can provide stable hand gesture classification performance under a repetition-based evaluation protocol and highlight the importance of appropriate data partitioning and class-wise misclassification analysis for reliable performance evaluation in sEMG-based hand gesture recognition.

Purpose: This study aimed to experimentally determine how the actual volume of a single droplet changes with increasing flow rate in the drip nozzle of a medical infusion set. Methods: An automated experimental system was developed to measure droplet volume over a flow-rate range of approximately 15-2,600 mL/h using a 20 drops/mL macrodrip infusion set. For each flow condition, the actual flow rate and mean droplet volume were determined from the time and number of droplets required to deliver 100 mL of distilled water. High-speed imaging at 960 frames/s was used to characterize droplet formation, necking, and pinch-off behavior. Polynomial regression and leave-one-out cross-validation were performed to derive and evaluate an empirical relationship between flow rate and droplet volume. Results: Droplet volume increased with flow rate, reaching approximately 125% of the initial value near 1,500 mL/h, and subsequently decreased at higher flow rates. The fourth-order empirical model showed a high goodness of fit (R²=0.9987) and a LOOCV-RMSE of 0.00352 mL/drop. Conclusions: Droplet volume varied nonlinearly with infusion flow rate, indicating that a constant droplet-volume assumption may cause errors in drip-based flow estimation. The proposed empirical model quantitatively describes this variation under the tested conditions and may provide a basis for improving infusion monitoring. Further validation with different fluids, infusion sets, and clinical conditions is required before broader application.

In this study, a simple and low-cost fabrication method for a sub-millifluidic droplet-generation device is presented, utilizing 3D-printed stamps and thermoplastic polycaprolactone (PCL). The proposed process leverages the low softening temperature of PCL (55~70℃) to form sub-millifluidic channel and sealing layers via hot embossing, followed by direct PCL-to-PCL bonding using hot air, without the need for cleanroom facilities, photolithography, or external adhesive materials. This represents a key advancement over previous PCL microfluidic approaches that relied on SU-8 molds and adhesive film sealing. The fabricated PCL sub-millifluidic channels showed comparable cross-sectional geometry to PDMS channels, with average width and height of 1000 ± 10 ㎛ and 500 ± 20 ㎛, respectively. These dimensions were intentionally selected to enable visual observation without optical microscopy and to ensure compatibility with passive negative-pressure driving using disposable syringes, targeting educational and resource-limited applications. Bonding quality was critically dependent on hot-air exposure time, with stable channel sealing achieved within an optimal window of 30~50 s. Contact angle measurements (101.9 ± 3.93°, CV = 3.86%) confirmed the inherent hydrophobicity of PCL surfaces, enabling droplet generation without surface treatment. Droplets were successfully generated in a T-junction configuration with an average diameter of 526 ± 73 ㎛ (CV = 13.9%), representing acceptable uniformity for educational and basic research purposes under passive driving conditions. These results demonstrate the functional feasibility of 3D-printed-stamp-based PCL hot embossing as a proof-of-concept approach for sub-millifluidic droplet-generation devices, suggesting its potential applicability to point-of-care diagnostics, educational settings, and resource-limited research environments.

Hyaluronic acid (HA) homeostasis is essential for maintaining the extracellular matrix and hydration properties of human skin, but its transcriptomic relationship with cellular senescence remains incompletely characterized. In this study, we propose an HA-PPI-informed transcriptomic framework to investigate whether HA homeostasis genes are coupled with SASP-oriented senescence activity in human skin. First, a STRING-based protein-protein interaction (PPI) network was constructed using HA synthesis, degradation, receptor, signaling, and aging/senescence-related genes. This network suggested that HA-associated genes, including CD44, HAS2, and HYAL1, are topologically connected with inflammatory and senescence-related hubs such as IL6, IL1B, TNF, MMPs, and TLR-related signaling nodes. Second, public aged skin transcriptome datasets were analyzed to evaluate whether HA-related genes show aging-associated transcriptional changes. CD44 showed significant downregulation in aged epidermis, while HYAL1 showed a statistically robust but moderate effect-size signal. Third, GTEx skin transcriptomes were used to test whether an HA-PPI-derived gene panel is associated with SenMayo-based SASP-oriented senescence activity. To reduce the influence of broad transcriptomic covariance, HA-PPI genes and SenMayo genes were excluded from PCA input, and residualized SenMayo scores were generated after adjusting for global expression principal components. The HA-PPI panel retained stronger performance than random gene panels after covariance adjustment, supporting a population-level transcriptomic coupling between HA homeostasis and SASP-oriented senescence. HYAL1 and CD44 emerged as convergent candidates across PPI topology, aged skin expression analysis, and covariance-adjusted senescence modeling, warranting experimental validation in skin senescence models.

Recently, an increasing number of attempts have been made to objectively evaluate individual activity rhythms using wearable devices and accelerometer data, and temporal patterns of physical activity are garnering attention as biosignals capable of reflecting mental health status. The associations between step-count-based circadian rhythm indices and quality of life, stress, and depression were analyzed using accelerometer data from the 2014-2017 Korea National Health and Nutrition Examination Survey (KNHANES) for 1,036 individuals who met the valid wear-time criteria. Mesor, Amplitude, Acrophase, M10, L8, IS, IV, and average daily steps were calculated, followed by the application of adjusted regression analyses and a machine learning-based regression model. The results indicated that higher Mesor, Amplitude, and M10 were associated with a higher quality of life and a lower risk of depression. Furthermore, higher M10 and Mesor exhibited a protective association against perceived stress, whereas a delayed Acrophase was associated with a higher risk of stress. Based on the machine learning regression model and SHAP analysis, Mesor, IV, M10, and L8 were identified as major predictor variables. These findings suggest that step-count-based circadian rhythm indicators have the potential to be utilized as objective digital biomarkers reflecting declines in mental health and quality of life.

Zein-based discoidal polymeric particles (DPPs) have emerged as promising drug delivery carriers owing to their excellent biocompatibility, unique geometry, and favorable physicochemical properties. In this study, fucoidancoated doxorubicin (DOX)-loaded Zein DPPs (Fuc/DOX-Zein DPPs) were developed as a potential drug delivery platform for lung cancer therapy. The formulation conditions for DOX loading and fucoidan coating were optimized, and the physicochemical characteristics, colloidal stability, pH-responsive drug release, cellular uptake, and in vitro anticancer activity were systematically evaluated. The optimal formulation was obtained with a DOX feed amount of 10 mg and a fucoidan concentration of 1 mg/mL, producing particles with a low polydispersity index, a highly negative surface charge, and a stable discoidal morphology with an average diameter of approximately 2.4 ㎛. Fuc/DOX-Zein DPPs exhibited accelerated DOX release under acidic conditions while maintaining physicochemical stability comparable to that of uncoated DOX-Zein DPPs under physiological conditions, indicating that fucoidan surface modification did not alter the structural integrity or release behavior of the particles. Blank Zein DPPs exhibited negligible cytotoxicity toward LLC-1 cells, whereas all DOX-containing formulations showed concentration-dependent anticancer activity. Compared with uncoated DOX-Zein DPPs, Fuc/DOX-Zein DPPs exhibited enhanced cytotoxicity and increased intracellular DOX uptake, suggesting that fucoidan surface modification may enhance the biological performance of Zein DPPs, potentially through interactions associated with the affinity of fucoidan for P-selectin. This putative P-selectin-mediated targeting mechanism, however, was not directly verified in the present study. Confocal microscopy further confirmed the intracellular delivery of DOX in cells treated with both DOX-Zein DPPs and Fuc/DOX-Zein DPPs. Overall, fucoidan surface modification preserved the discoidal morphology, physicochemical stability, and pH-responsive drug release behavior of DOX-loaded Zein DPPs while improving their in vitro biological performance. These findings suggest that Fuc/DOX-Zein DPPs represent a promising discoidal drug delivery platform for lung cancer therapy. Further in vivo studies are warranted to determine whether fucoidan surface modification enhances tumor-targeting efficiency and therapeutic efficacy.

In Korea, the medical device approval certificate serves both as legal evidence of regulatory approval and as a means of managing detailed technical information on approved devices. This study investigated medical device approval systems in the United States, the European Union (EU), Japan, and Korea and compared approval-related documents, their information elements, and corresponding post-market management systems. The regulatory and post-market functions of individual information elements were comparatively evaluated based on relevant laws, regulations, and guidelines, including the U.S. Federal Food, Drug, and Cosmetic Act (FD&C Act), the EU Medical Device Regulation (MDR), Japanese regulations and guidelines issued by the Ministry of Health, Labour and Welfare (MHLW) and the Pharmaceuticals and Medical Devices Agency (PMDA), and Korean medical device regulations. In the U.S., legally effective authorization or clearance documents are separated from documents containing detailed technical information, while summaries of safety and effectiveness information supporting regulatory decisions are separately provided. In the EU, no government-issued approval certificate equivalent to the Korean certificate exists; instead, Notified Bodies (NBs) issue certificates for devices requiring their involvement in conformity assessment under the MDR. In Japan, approval documents are relatively concise, with regulatory information supplemented by review reports, electronic instructions for use (eIFU). In contrast, Korea incorporates extensive regulatory and technical information directly into the approval certificate. Korea should therefore consider simplifying approval certificates while maintaining their essential legal functions and improving information accessibility, administrative efficiency, transparency, and alignment with international regulatory frameworks.

This study compared reimbursement systems for digital therapeutics (DTx) in South Korea, the United States, and Germany to identify strategies for improving reimbursement for mental disorder DTx in South Korea. A literature review was conducted using academic publications, legislation, government documents, policy reports, and payer documents. Qualitative comparative analysis examined reimbursement pathways, reimbursement structures, pricing mechanisms, patient cost-sharing, and transparency of reimbursement-related information. SWOT analysis and the TOWS matrix were used to develop policy strategies. All three countries had reimbursement pathways for DTx but differed substantially in their reimbursement structures. South Korea operated multiple market-entry pathways with National Health Insurance coverage for physician consultation and treatment outcome evaluation fees. The United States used a decentralized reimbursement system, while Germany adopted a value-based reimbursement model through the DiGA Fast-Track pathway. SWOT analysis identified major barriers in South Korea, including unclear reimbursement listing criteria, high patient cost-sharing, limited value-based reimbursement, and insufficient transparency. The TOWS matrix suggested introducing value-based reimbursement, reducing patient cost-sharing, and establishing a real-world data-based performance management system. Improving the reimbursement system is essential for expanding access to DTx for mental disorders and provides strategic directions for future reimbursement policy development in South Korea.