The maximum a posteriori (MAP) approach has played an important role in developing model-based iterative reconstruction methods for positron emission tomography (PET). In this approach, the objective function consists of the likelihood and prior terms, with the balance between these two terms controlled by a positive smoothing parameter. When the prior term takes a quadratic form as a regularizer, the objective function can be easily optimized using the method of iterated conditional modes (ICM). However, the quadratic regularizer has the disadvantage of being unable to preserve the edges of the reconstructed image. In this study, we propose a new method for enhancing edges even with a quadratic regularizer within the ICM framework by adaptively fine-tuning the smoothing parameter. To adjust the smoothing parameter, we use a method that transforms the histogram, obtained from the pixel-wise standard deviation of the estimated image in the previous iteration, into a mapping function for assigning an optimal value of the smoothing parameter for each pixel. The simulation results demonstrate that the proposed method improves the reconstruction accuracy in terms of the percentage error as well as the bias calculated from independent noise trials by efficiently enhancing the edges.

This study examines the regulatory landscape and impact of implementing Good Clinical Laboratory Practice (GCLP) in clinical trial sample analysis across the United States, Europe, China, and Korea. Clinical trials are essential for drug development, and reliable sample analysis plays a critical role in ensuring data integrity. GCLP integrates the principles of Good Laboratory Practice (GLP) and Good Clinical Practice (GCP) to maintain quality and reliability in clinical laboratory data. In particular, this research highlights how GCLP helps standardize processes and reduce discrepancies in multinational clinical trials. The comparative analysis of regulatory frameworks and the real-world effects of GCLP adoption in these countries provides insights into improving Korea's clinical trial infrastructure. The study suggests that enhancing GCLP-compliant infrastructure, expanding personnel training, and establishing central labs are crucial steps for Korea to strengthen its global competitiveness. These measures are vital to ensuring data consistency and increasing the success rates of clinical trials in a rapidly evolving international regulatory environment.

The purpose of this study is to investigate bacterial contamination on mammography equipment(face shields, compression Paddles and detectors, handles) due to physical contact. Given the high infection risk from skin surface bacteria during breast cancer screenings, we measured contamination levels before and after disinfection, identifying detected bacterial strains. Equipment from two medical facilities was examined, focusing on specified parts. Samples were collected pre-disinfection, post-disinfection(before and after daily examinations), and analyzed via plate culture. Results highlighted significant bacterial presence, especially pre-disinfection and from airborne bacteria on equipment surfaces. compression Paddles and Detectors areas showed higher contamination, emphasizing the need for focused disinfection. Disinfection is crucial before mammography to prevent infections from airborne bacteria. Harmful bacteria were notably detected in the compression Paddles and Detectors area, necessitating focused disinfection of the compression Paddles and Detectors part. Additionally, the face shields and handles area, being at high risk for infections, also showed the presence of harmful bacteria, indicating the need for disinfection in that area as well. Alcohol swabs proved more effective than sprays for airborne bacteria removal, with the most bacteria detected post-examination without disinfection. This supports alcohol-based disinfection's critical role. We propose that removing airborne bacteria is essential, mammography equipment should be disinfected focusing on certain areas, and disinfection should be performed with alcohol swabs before, during, and after examinations.

This study investigates efficient artifact removal methods based on adaptive filtering to improve the performance of steady-state visual evoked potential (SSVEP)-based brain-computer interface (BCI) systems. We utilized a custom-designed helmet equipped with a head-mounted display (HMD) to provide visual stimuli and electrodes to measure electroencephalography (EEG) signals from the occipital region as well as bio-signals from facial regions. We compared various filtering methods including band-pass filtering (BPF), normalized least mean square (NLMS), and recursive least square (RLS) adaptive filtering, and evaluated their performance in terms of classification accuracy using canonical correlation analysis (CCA) to identify the target frequency of the SSVEP. Results showed that the combination of BPF and NLMS adaptive filtering achieved the highest classification accuracy of 97.19%. Moreover, the choice of reference signal channels was important to the performance, showing electrodes placed below the eyes being more effective than those above. In conclusion, these findings demonstrate that incorporating efficient adaptive filtering methods and selecting appropriate reference channels can improve the accuracy and reliability of SSVEP-BCI systems.

Involuntary tongue movement and tremor involve isolated tongue motion without affecting other body parts, often linked to neuropsychiatric and overall health conditions. Unlike general object tracking, which relies on distinct color boundaries, tracking tongue movement requires detailed observation due to its similar color areas. This study aims to find an effective method to monitor tongue tremors and movements, including a magnification technique. The researcher captured 10 coherent images of the tongue at 0.1-second intervals, presenting a challenge due to the lack of distinct feature points like corners. The Farnebäck optical flow (FOF) method, which calculates optical flow for each pixel, proved more effective for this purpose. To illustrate detailed tongue movements, two approaches were considered: using image up-sampling to acquire optical flow and obtaining optical flow at the original resolution before up-sampling the vector for high-resolution optical flow. The FOF method successfully detected tongue movements and their magnitude across all positions, regardless of movement size. The first method of image up-sampling introduced errors, creating artifacts at boundaries and altering the optical flow direction. In contrast, the second method of vector up-sampling was aligned with feature changes, such as tongue cracks and spots, avoiding errors on curved or flat surfaces. In conclusion, the 2D vector up-sampling method, based on FOF, was found to be effective for detecting detailed tongue movements, proving to be a reliable technique for monitoring involuntary tongue tremors.

People of the same chronological age can show different signs of aging. Biological age is an indicator of the degree of biological aging of the body excluding diseases. Perceived age is highly correlated with biological age, which reflects health assessment, and is often used as a clinical indicator of aging. However, there is a lack of objective methods to quantify perceived age. Therefore, this study aimed to propose a novel perceived age estimation algorithm. The proposed algorithm consists of two steps. First, the initial perceived age is predicted from a facial image using a convolutional neural network (CNN) ensemble model. In the second step, the final perceived age is estimated by applying a regression algorithm to the predicted gender, predicted BMI, texture features extracted from the facial image, and the predicted perceived age obtained in the first step. Better performance results were obtained by averaging models generated from various base regression models. The averaged model of CatBoost and LightGBM showed a mean absolute error of 2.9614. The proposed method can be used as a health care model to promote self-care.

As the market size of medical devices grows and the spread of information technology, including artificial intelligence, interest in software medical devices is growing. Many researchers are seeking to commercialize their software R&D achievements through licensing, and regulatory agencies are responding to this demand by establishing a separate group of reviews for software medical devices and managing review procedures based on the IEC 62304 standard. However, due to the differences between the process in the R&D and the medical device approval procedure, it is challenging for individual researchers to effectively respond to the approval process. This study provides guidelines for the documentation that should be elicited for authorization review in the software R&D process. By analyzing the requirements of the U.S. Food and Drug Administration (FDA) and the Ministry of Food and Drug Safety (MFDS) for medical device approval, we provide guidelines for responding to both U.S. and Korean regulatory agencies. We facilitate the preparation of researchers for licensing by presenting the requisite licensing documents at each of the six stages of research and development.

Unlike previous classification-based approaches that discretize continuous emotional states, this study proposes a multi-output regression model for predicting emotional valence and arousal using electroencephalography (EEG) signals. This method captures the continuous and complex nature of emotions. We compared three approaches: single-output regression, which predicts valence and arousal separately, and multi-output regression and multi-output regression using chain structure, both of which predict valence and arousal simultaneously. The single-output approach, though simple and easy to analyze, showed limitations in handling complex emotional states and required two model parameters, leading to inefficiencies. However, the multi-output approach effectively handled both emotional dimensions simultaneously, offering a more efficient and streamlined solution while maintaining high accuracy. Both LGGNet and CCNN models demonstrated the ability to accurately predict emotional states in the valencearousal space, closely matching actual emotional states. This indicates that the regression models can effectively capture complex emotions. Future work will explore more advanced network architectures to further improve emotion recognition accuracy.

This study proposes a deep learning-based system to overcome the challenges of degraded image quality due to limited lighting and small size of camera in endoscopic surgery, particularly addressing the issue of smoke obscuring the surgeon's view. The image enhancement algorithm consists of dark area improvement, blood vessel emphasis, and noise reduction, implemented through filtering and histogram equalization in the Lab color space. For smoke removal, a Generative Adversarial Network-based model was employed, using synthetic data generated by adding artificial smoke via Perlin noise to the original images. The results demonstrate the effectiveness of the proposed system, with significant improvements in visual clarity, particularly in dark areas and blood vessel visibility, as evidenced in qualitative and quantitative analyses. Test results demonstrated that the proposed system achieved superior image quality compared to existing smoke-specific removal models, with Peak Signal-to-Noise Ratio and Structural Similarity Index Measure indices of 23.17 and 0.8802, respectively. These outcomes are expected to significantly contribute to improving the efficacy and safety of endoscopic surgeries.