Purpose: There are concerns that social costs will content and and national competitiveness will decline due to stress, depression, etc. To solve these problems, this study focused on the uniqueness of forest sounds. We extracted fractal dimensions from forest sounds and studied the relaxation effect through tactile manipulation. To this end, Materials and Methods: we developed an algorithm that collects forest sounds and calculates fractal dimensions in real time. In particular, we devised a method to find self-similarity, a characteristic of fractals, in forest sounds. We extracted forest sounds containing fractals and made them into one content and applied two forest sounds with different fractal dimensions to 20 people and analyzed the effects on brain waves. Considering that auditory thresholds increase with age, we experimented by replacing forest sounds with tactile sensations. Results: The real-time algorithm for finding fractal dimensions and self-similarity worked well. Through this program, we confirmed that forest sounds have fractal characteristics and self-similarity. The results of applying forest sounds with a fractal dimension of 1.46 to subjects showed significant results (p=0.009). Conclusion: Through this study, we found a scientific method that helps psychological relaxation by utilizing forest sounds. That is, we found the possibility of improving stress by utilizing forest sounds not only in forest environments but also in general office spaces. Since the fractal dimension of forest sounds can vary depending on conditions such as forest size, season, and altitude, further research is needed.

The activation of polygenic risk score (PRS) in precision medicine has raised concerns about privacy preservation. Current encryption methods that delete or modify variant information may reduce prediction accuracy. The purpose of this study was to evaluate the feasibility of PRS-based disease susceptibility prediction in a fully homomorphic encryption (FHE) environment while assessing computational trade-offs. We conducted disease susceptibility prediction using synthetic public genome-wide association study (GWAS) data on coronary artery disease (CAD) and applied inverse probability weighting, k-fold cross-validation, and linkage disequilibrium (LD) shrinkage to enhance polygenic risk score stability and accuracy. Four machine learning models-PRS only logit, ridge regression, linear support vector machine, and radial basis function support vector machine-were trained and tested on fully encrypted genomic and clinical data, with performance evaluated using area under the curve (AUC), mean absolute error (MAE), and turnaround time. A total of 1,401 samples and 404,663 SNPs were used. Ridge regression balanced prediction accuracy and computational efficiency (AUC = 0.7631, MAE = 0.4375, Turnaround Time = 1,343.65s). RBF SVM achieved the highest AUC (0.7792, MAE = 0.3120) but required the longest time (1,983.57s). PRS Only Logit had the shortest time (279.27s) but the lowest AUC (0.6339, MAE = 0.4162). These results suggest ridge regression may have the best balance between prediction accuracy and computational efficiency, making it the most practical choice in an FHE environment.

This paper proposes an efficient and reconfigurable deep neural network architecture for multi-label classification of cardiovascular diseases (CVDs) based on electrocardiogram (ECG) signals for on-device application. It employs conditional and iterative multi-stage binary classifiers to replace the conventional single-stage multi-label classifier that can't simply reconfigure structure without redesigning hardware such as GPUs or tensor processing cores if the classifier needs to support more labels. The proposed architecture preprocesses multi-labeled disease data into a binary classification-friendly format to train the binary classifier in each stage and computes the optimal decision thresholds at each stage using the Youden index. When classifying diseases, only one binary classifier will be iteratively reused by loading different neural network model parameters for each disease classification. In this architecture, new disease classification can be easily implemented by saving new parameters in the system memory and increasing the iteration number programmably without redesigning the hardware. Based on the evaluation with PTB-XL-ECG dataset, the proposed model demonstrated that it can achieve the better Hamming Loss (0.1365) and macro AUC (0.9255) as the ResNet50-based multi-label classifier with over 1000x fewer parameters and scalability. In addition, we confirm that it can further reduce the Hamming Loss by 0.006 compared to identical AUC multi-binary classifiers by placing normal signal classification at the first stage and skipping disease classifications stage.

Cardiovascular disease (CVD) is a leading cause of death worldwide, necessitating electrocardiogram (ECG) monitoring for its early detection and appropriate treatment. Particularly for hospitalized patients, continuous real-time ECG monitoring can enhance treatment efficacy. With advancements in technology, IoT devices such as wearable sensors are being integrated for Real-Time ECG monitoring system. Centralized systems enable medical staff to monitor multiple patients simultaneously, thereby improving the efficiency of decision support. We developed a system in which a self-scheduling-based parallel processing and adaptive batching method is employed to process real-time ECG data from multiple patients in a single-system. In this system, independently operating algorithm modules dynamically retrieve patient data for analysis through self-scheduling. Different patient data are processed, and the batch size is flexibly adjusted based on the volume of the data, enabling real-time data processing. The proposed system was evaluated in two aspects: real-time processing performance and algorithm performance. The proposed system could enable real-time processing for up to 240 patients, achieving an average completion time of 2.788 s, which was shorter than the input interval of 3-s. The performance of the algorithm was evaluated in a cohort of 80 hospitalized patients who underwent radiofrequency catheter ablation, demonstrating arrhythmia detection performance with a sensitivity of 99.429% and a positive predictive value of 91.238%. This system can efficiently and rapidly process data within a single-system, even when utilizing multiple wearable sensors, allowing for the low-cost monitoring of a large number of patients.

The purpose of this study is to develop a portable device and algorithm based on multimodal imaging for early diagnosis of breast cancer and to verify its effectiveness. In order to overcome the limitations of existing imaging diagnostic methods, this study designed the Tactile Sensation and Diffuse Optical Imaging System (TSDOIS) that combines tactile sensing image (TSI) and diffuse optical image (DOI). This system is designed to measure the physical and physiological characteristics of the lesion simultaneously, and increased accessibility and usability through miniaturization and mobile interlocking device implementation. The TSI acquisition module extracts information on the elasticity, depth, and size of the lesion through a total internal reflection (TIR)-based elastic waveguide, and the DOI acquisition module detects lesions with high hemoglobin concentration based on the red light (638 nm) reflectance reduction pattern. As a result of the phantom experiment, it was confirmed that the TSI and DOI contrast were significantly increased in high stiffness, surface layer, large size, and high hemoglobin nodules. This study suggests that small diagnostic equipment based on multimodal imaging can present new possibilities for early diagnosis of breast cancer. In the future, it is expected that the scope of use of this system can be expanded through performance verification based on clinical data and application of machine learning-based automatic analysis algorithms.

Individual-specific steady-state visual evoked potential (SSVEP)-based brain-computer interfaces (BCIs) rely on individual data calibration, achieving high performance. However, existing methods use generalized channels and task-related subspaces, which limit their potential and lead to suboptimal solutions. To address this, we propose an adaptive selection strategy called AdapTRCA for developing a purely individual-specific SSVEP-BCI speller. AdapTRCA optimizes both channel and subspace selection. The channel selection process constrains sparse learning by spatial distance to identify the optimal subject-specific channels, while subspace selection adaptively determines the optimal number of subject-specific task-related subspaces by maximizing profile likelihood. Extensive experiments on two publicly available datasets with 40 classes show that AdapTRCA selects more meaningful channel subsets and determines the proper number of task-related subspaces compared to traditional methods. Moreover, when integrated with advanced calibration-based SSVEP decoding methods, AdapTRCA fully leverages the potential of individual-specific SSVEP-BCI. In conclusion, AdapTRCA enhances the performance of user-specific SSVEP-BCI spellers, promoting their practical application.

Digital staining (DS) has solved the problems of traditional histological staining (HS), which is complicated, time-consuming, and costly. However, differences in staining protocols (e.g., solution concentration, staining time, scanner used, patient variability) have not completely overcome the differences in staining color and quality. To solve this problem, we used a DS method utilizing domain transformation in unstained images. This study proposes a single-domain transformation staining method combining unsupervised and supervised learning, as well as an unsupervised learning-based multi-domain transformation staining method. The data were constructed using fourier ptychographic microscopy and bright-field microscopy images and optimized for training through five preprocessing steps. Afterward, we compared the proposed methods with the existing unsupervised staining method, CycleGAN. Prior to evaluation, the data were further processed through postprocessing to facilitate analysis. The visual evaluation showed that the proposed method effectively distinguished the background compared to the control group and prevented the over-staining issue that occurs during the HS process. Quantitative evaluation using a cell nucleus detection model demonstrated that the proposed model showed improvements of more than 7.96% and 10.81% in the Dice and Jaccard indices, respectively, compared to the CycleGAN model. This indicates that the proposed model achieves digital staining that closely resembles the traditional Hematoxylin and Eosin staining method, and it suggests that supervised learning can be utilized even in environments without paired data, and unstained images can be used in multi-staining research.

This study examines the regulatory frameworks for companion diagnostic medical devices (CDx) in selected advanced countries and compares them with South Korea's current system. By analyzing global CDx regulatory principles, we identify regulatory hurdles and administrative challenges that may hinder the long-term growth of the CDx industry. A literature review was conducted using official websites of medical device and healthcare regulatory agencies, legal requirements, guidelines published by regulatory authorities, and relevant academic and industrial reports. The collected data were analyzed to compare CDx regulatory policies across different countries and evaluate their alignment with South Korea's CDx regulatory framework. The results revealed that advanced regulatory systems, such as the Single-Site Assay (SSA-CDx) framework in the U.S. and the Co-Review System in Japan and Australia, facilitate faster CDx approval and integration with targeted therapeutics. In contrast, South Korea lacks a structured regulatory framework for laboratory-developed test (LDT)-based CDx and does not mandate concurrent review of CDx and relevant therapeutics, resulting in delays in market entry and clinical adoption. To enhance global competitiveness, South Korea should establish a structured approval process with clear criteria for LDT-based CDx, implement a Co-Review System of therapeutics and CDx, streamline the reimbursement process, and develop a realtime online CDx database. These measures will help align South Korea's regulatory framework with international standards, foster industry growth, and improve patient access to precision medicine.

Astrocytes play a critical role in maintaining brain homeostasis and exhibit temperature sensitivity that is not yet fully understood. However, current chambers for studying thermal responses in Astrocytes are limited by inadequate temperature control and real-time observation capabilities. In this study, we developed a cost-effective PVC-gel heater-based thermal chamber that enables precise temperature modulation within physiological temperature range (37℃ to 42℃ with an error of ±1℃) and real-time imaging of Astrocytes. The PVC-gel heater-based thermal chamber ensures uniform temperature distribution (+1.5℃) due to its low thermal diffusivity (α=0.087×10^-6m²/s) and shows high transmittance (T >90% across the entire visible range) for in-situ observation. We validated this chamber with microscopy-compatible thermal stability tests and live-dead assays to confirm biocompatibility under various temperature conditions 37℃, 40℃ and 42℃. Using this chamber, we investigated the temperature sensitivity of Astrocytes by analyzing temperature-sensitive genes and inflammation-related gene expression. Our findings revealed significant changes in gene expression, indicating Astrocytes response to thermal stress by activating temperature-sensitive pathways and inflammatory responses. The PVC-gel heater-based thermal chamber offers a robust and effective tool for studying Astrocyte thermosensitivity and can be broadly applied to investigate temperature-induced cellular mechanisms.

Alzheimer's disease is an irreversible neuro degenerative disease, and associated declines in cognitive function have a significant effect on daily life. Methods such as mini-mental state examination and Montreal cognitive assessment, the existing diagnostic methods for Alzheimer's disease, take a long time and have poor objective assessment score accuracy. This paper presents a novel method that assesses the brain's cognitive state by utilizing eye-tracking technology to measure abnormal fixational eye movement patterns of microsaccades, which are involuntary eye movements, and changes in pupil size using eye-tracking technology in nervous system diseases. The experiment results confirmed that the number of microsaccades occurrences on the x and y axes gradually increased as aging progressed. In addition, it was confirmed that the number of microsaccades occurrences increased as cognitive function scores declined, even among seniors in the same age groups. In the case of pupil size, it was observed that the amplitude of pupil size variation decreases with increasing age, along with a reduction in both constriction and dilation rates. This paper proves the correlation between aging and the brain's cognitive functions and verifies that declines in the brain's cognitive functions affect microsaccades and changes in pupil size.

Vitiligo is a chronic autoimmune skin disorder characterized by depigmented patches caused by the loss of melanocytes. While not life-threatening, facial lesions can severely impact patients' psychological well-being. Conventional assessments rely on 2D images, which inherently fail to capture the true spatial distribution of lesions due to limitations in perspective, curvature distortion, and occlusion along the facial contours. This study presents a novel 3D-based framework for analyzing vitiligo lesion distribution on the face using multi-view images and a 3D Morphable Model (3DMM). Facial meshes are reconstructed from frontal, left, and right images, and vitiligo masks are aligned to the 3D surface using UV mapping. By projecting lesion information onto a fully reconstructed 3D facial surface, the framework overcomes the structural blind spots of 2D analysis-capturing curvature, depth, and areas along the sides of the face that are otherwise distorted or hidden from view. A comparative analysis of 2D and 3D methods using patient data revealed that the 3D approach consistently estimated larger lesion areas, particularly in regions where 2D projections underestimate due to view-dependent loss or distortion. This 3D framework significantly enhances the anatomical accuracy and reliability of vitiligo assessment, providing a powerful tool for clinical evaluation, longitudinal tracking, and therapeutic outcome measurement.

Although Euclidean Distance (ED) has limitations in fully capturing the inherent similarity between signals, it has demonstrated higher accuracy in personal identification than Dynamic Time Warping (DTW) when applied in Electroencephalogram (EEG) signal-based authentication systems. In this study, we aim to compare the performance of ED, DTW, and DTW-Normalization (DTW-N) algorithms in assessing EEG signal similarity. Furthermore, this study evaluates the effects of normalization on similarity measurement across different channels, participants, and signal counts. EEG data were collected from ten participants during speech tasks with auditory stimuli, and all 32 EEG channels were analyzed. The α is an indicator used to quantitatively evaluate the difference between signals from the same subject and different subjects; a higher value indicates a greater difference in signal similarity. DTW-N achieved the highest α values compared to ED and DTW. Across all channels, DTW-N showed the highest α values, with the FC1 channel having the highest average α_DTW-N value of 3.4110 × 10^-2. Additionally, α for participants 3 and 9 reached 4.7225 × 10^-5, approximately 55.79% higher than the DTW-N mean, while α for participants 7 and 8 was the lowest at 4.722× 10^-5. As the number of signals increased, the α values decreased. The DTW-N algorithm effectively addressed temporal distortion and amplitude variations in EEG signals, making it highly effective for distinguishing individuals based on EEG patterns. Future research will explore optimal representative metrics for EEG data and enhance individual identification performance using DTW-N-based classification models.

While ultrasound examination is widely used for breast tumor diagnosis due to its non-invasive and economic advantages, classification accuracy is inherently affected by ultrasound image quality and diagnostic subjectivity. Recent advancements in deep learning have introduced automated analysis methods for classification of medical ultrasound images, achieving high accuracy. However, the performance of these methods critically depends on the availability of large-scale datasets. This paper proposes the classification method using curriculum learning with B-mode partial images to achieve high classification performance with limited datasets. By starting learning from small partial images centered on tumor regions defined by mask images and progressively expanding to include wider contextual areas, the proposed method enables the model to progressively learn ultrasound image characteristics. To validate the proposed method, we conducted experimental evaluation using 200 breast ultrasound images from the OASBUD dataset. Experimental results demonstrated that curriculum learning significantly improved classification accuracy compared to original image learning and data augmentation.

With rapid technological advancements, many countries have introduced regulatory programs to facilitate the market entry of innovative medical devices. This study investigates the designation systems for innovative medical devices in the United States, China, and South Korea, focusing on the period from 2020 to 2024, with particular attention to devices incorporating artificial intelligence (AI) technologies. The findings reveal that all three countries have implemented institutional frameworks to support the development and priority review of designated innovative medical devices. In particular, South Korea has actively promoted the early market entry of AI-based medical devices through the introduction of the Integrated Review Program in October 2022. Furthermore, analysis of designation trends in the digital healthcare sector shows that such devices account for 21% of the total designations in the United States, 9% in China, and 74% in South Korea, indicating South Korea's concentrated efforts in fostering AI-based medical device development. Based on these findings, this study examines the operational characteristics of South Korea's Integrated Review Program and proposes improvement strategies to enhance its effectiveness and promote greater utilization of regulatory benefits.

In Korea, the submission of clinical evidence is required as supporting evidence to demonstrate the safety and effectiveness of medical devices. This study proposes a procedure for determining whether medical devices require clinical evidence submission, aiming to assist manufacturers in assessing its necessity. The criteria for determination of medical devices for clinical evidence submission in Korea were established through an investigation of relevant regulations. A category reflecting detailed determination criteria was established, flowchart was proposed to determine the subjects for clinical evidence submission. The proposed procedure in this study is expected to determine the subject of submission of clinical evidence systematic and clearly, for manufacturing, import, approval, and certification medical devices to determine clinical evidence submission requirements.