• Journal of the Korea Society of Computer and Information
• /
• v.26 no.1
• /
• pp.27-35
• /
• 2021

In this paper, we propose a new stacking ensemble framework for deep learning models which reflects the distribution of label embeddings. Our ensemble framework consists of two phases: training the baseline deep learning classifier, and training the sub-classifiers based on the clustering results of label embeddings. Our framework aims to divide a multi-class classification problem into small sub-problems based on the clustering results. The clustering is conducted on the label embeddings obtained from the weight of the last layer of the baseline classifier. After clustering, sub-classifiers are constructed to classify the sub-classes in each cluster. From the experimental results, we found that the label embeddings well reflect the relationships between classification labels, and our ensemble framework can improve the classification performance on a CIFAR 100 dataset.

• Journal of Neurogastroenterology and Motility
• /
• v.25 no.3
• /
• pp.387-393
• /
• 2019

Background/Aims Delayed gastric emptying (GE) is associated with high morbidity and mortality in subjects with diabetes. The aim of this study is to investigate associations between GE time and the major cardiovascular events (coronary heart diseases and ischemic stroke) in diabetic subjects with upper gastrointestinal (UGI) symptoms. Methods Among 259 subjects with chronic UGI symptoms who underwent gastric emptying study (GES) over 13 years, 122 diabetic subjects without gastric surgery and/or rapid GE were enrolled in this study. We also gathered data about baseline demographics, clinical characteristics, estimated GE half-time (GE T½) and incidence of cardiovascular events following GES. Results The mean age of subjects was 64.0 ± 17.4 years. There were 86 women and 104 subjects with type 2 diabetes. There were 52 (42.6%) subjects with normal GE, 50 (41.0.%) subjects with mild delayed GE, and 20 (16.4%) subjects with marked delayed GE. During follow-up (median, 207 weeks), cardiovascular events occurred in 7 (13.5%) subjects with normal GE, 4 (8.0%) subjects with mild delayed GE and 7 (35.0%) subjects with marked GE (P = 0.015). Univariate analysis showed that GE T½ was significantly associated with incidence of cardiovascular events (crude OR, 1.74; 95% CI, 1.12-2.69; P = 0.014). In a multivariate model, association between GE T½ and incidence of cardiovascular events remained statistically significant after adjustment for baseline characteristics and comorbidities (adjusted OR, 1.94; 95% CI, 1.21-3.12; P = 0.006). Conclusion A delay of GE was associated with an increased incidence of cardiovascular events in diabetic subjects with chronic UGI symptoms.

• Journal of Dental Anesthesia and Pain Medicine
• /
• v.24 no.4
• /
• pp.273-283
• /
• 2024

Background: This study evaluated the effect of remimazolam and propofol on changes in autonomic nerve activity caused by surgical stimulation during orthognathic surgery, using power spectrum analysis of blood pressure variability (BPV) and heart rate variability (HRV), and their respective associations with cardiovascular fluctuations. Methods: A total of 34 patients undergoing Le Fort I osteotomy were randomized to the remimazolam (Group R, 17 cases) or propofol (Group P, 17 cases) groups. Observables included the low-frequency component of BPV (BPV LF; index of vasomotor sympathetic nerve activity), high-frequency component of HRV (HRV HF; index of parasympathetic nerve activity), balance index of the low- and high-frequency components of HRV (HRV LF/HF; index of sympathetic nerve activity), heart rate (HR), and systolic blood pressure (SBP). Four observations were made: (1) baseline, (2) immediately before down-fracture, (3) down-fracture, and (4) 5 min after down-fracture. Data from each observation period were compared using a two-way analysis of variance with a mixed model. A Bonferroni multiple comparison test was performed in the absence of any interaction. One-way analysis of variance followed by Tukey's multiple comparisons test was performed when a significant interaction was observed between time and group, with P < 0.05 indicating statistical significance. Results: Evaluation of autonomic nerve activity in comparison with baseline during down-fracture showed a significant increase in BPV LF (P < 0.001), an increasing trend in HRV LF/HF in Group P, and an increasing trend in HRV HF in Group R. There were no significant differences in HR or SBP between the two groups. Conclusion: During down-fracture of Le Fort I osteotomy, sympathetic nerve activity was predominant with propofol anesthesia, and parasympathetic nerve activity was predominant with remimazolam anesthesia.

• Ocean and Polar Research
• /
• v.34 no.2
• /
• pp.239-251
• /
• 2012

The food web dynamics in a coastal ecosystem of Korea were predicted with Ecosim, a trophic flow model, under various scenarios of primary productivity due to ocean warming and ocean acidification. Changes in primary productivity were obtained from an earth system model 2.1 under A1B scenario of IPCC $CO_2$ emission and replaced for forcing functions on the phytoplankton group during the period between 2020 and 2100. Impacts of ocean acidification on species were represented in the model for gastropoda, bivalvia, echinodermata, crustacean and cephalopoda groups with effect sizes of conservative, medium and large. The model results show that the total biomass of invertebrate and fish groups decreases 5%, 11~28% and 14~27%, respectively, depending on primary productivity, ocean acidification and combined effects. In particular, the blenny group shows zero biomass at 2080. The zooplankton group shows a sudden increase at the same time, and finally reaches twice the baseline at 2100. On the other hand, the ecosystem attributes of the mean trophic level of the ecosystem, Shannon's H and Kempton's Q indexes show a similar reduction pattern to biomass change, indicating that total biomass, biodiversity and evenness shrink dynamically by impacts of climate change. It is expected from the model results that, after obtaining more information on climate change impacts on the species level, this study will be helpful for further investigation of the food web dynamics in the open seas around Korea.

• The Korean Journal of Pain
• /
• v.32 no.4
• /
• pp.264-270
• /
• 2019

Background: To develop a rabbit epidural steroid injection (ESI) model for analyzing steroid retention in the tissue, and to assess the difference in steroid retention in the model according to the location and time elapsed after ESI. Methods: Fluoroscopy-guided ESI was performed using the interlaminar approach between the lowest two lumbar segments in 13 female New Zealand white rabbits. Four rabbits were allocated to each of three different groups according to the time of sacrifice: 3, 7, and 15 days post-ESI; the remaining rabbit was sacrificed immediately post-ESI to obtain baseline data. After sacrifice, two segments were harvested: the lowest two lumbar vertebrae and another two lumbar vertebrae immediately above these. The residual steroid amount (RSA) and residual steroid concentration (RSC) in the collected spinal columns were analyzed. A linear mixed model was used to compare RSAs and RSCs between the injected and adjacent segments, and among the number of days until sacrifice; P < 0.05 was considered statistically significant. Results: Both RSA and RSC of the injected segment were significantly higher than those of the adjacent segment (P < 0.001, both). The RSA and RSC significantly decreased over time (P = 0.009 and P = 0.016, respectively). Conclusions: The developed rabbit ESI model verified that significantly more steroid was retained at the injected segment than at the adjacent segment and the residual steroid decreased over time. This model could be useful not only for comparing current steroid medications, but also for developing new, better steroid formulations.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.1
• /
• pp.16-37
• /
• 2022

Accurate liver segment segmentation based on radiological images is indispensable for the preoperative analysis of liver tumor resection surgery. However, most of the existing segmentation methods are not feasible to be used directly for this task due to the challenge of exact edge prediction with some tiny and slender vessels as its clinical segmentation criterion. To address this problem, we propose a novel deep learning based segmentation model, called Boundary-Aware Dual Attention Liver Segment Segmentation Model (BADA). This model can improve the segmentation accuracy of liver segments with enhancing the edges including the vessels serving as segment boundaries. In our model, the dual gated attention is proposed, which composes of a spatial attention module and a semantic attention module. The spatial attention module enhances the weights of key edge regions by concerning about the salient intensity changes, while the semantic attention amplifies the contribution of filters that can extract more discriminative feature information by weighting the significant convolution channels. Simultaneously, we build a dataset of liver segments including 59 clinic cases with dynamically contrast enhanced MRI(Magnetic Resonance Imaging) of portal vein stage, which annotated by several professional radiologists. Comparing with several state-of-the-art methods and baseline segmentation methods, we achieve the best results on this clinic liver segment segmentation dataset, where Mean Dice, Mean Sensitivity and Mean Positive Predicted Value reach 89.01%, 87.71% and 90.67%, respectively.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.17 no.7
• /
• pp.1841-1857
• /
• 2023

CenterNet, a novel object detection algorithm without anchor based on key points, regards the object as a single center point for prediction and directly regresses the object's height and width. However, because the objects have different sizes, directly regressing their height and width will make the model difficult to converge and lose the intrinsic relationship between object's width and height, thereby reducing the stability of the model and the consistency of prediction accuracy. For this problem, we proposed an algorithm based on the regression of the diagonal half-length and the center angle, which significantly compresses the solution space of the regression components and enhances the intrinsic relationship between the decoded components. First, encode the object's width and height into the diagonal half-length and the center angle, where the center angle is the angle between the diagonal and the vertical centreline. Secondly, the predicted diagonal half-length and center angle are decoded into two length components. Finally, the position of the object bounding box can be accurately obtained by combining the corresponding center point coordinates. Experiments show that, when using CenterNet as the improved baseline and resnet50 as the Backbone, the improved model achieved 81.6% and 79.7% mAP on the VOC 2007 and 2012 test sets, respectively. When using Hourglass-104 as the Backbone, the improved model achieved 43.3% mAP on the COCO 2017 test sets. Compared with CenterNet, the improved model has a faster convergence rate and significantly improved the stability and prediction accuracy.

• Asian Pacific Journal of Cancer Prevention
• /
• v.14 no.3
• /
• pp.2059-2065
• /
• 2013

This goal of this research was to evaluate the cost-effectiveness of the National Cancer Screening Program (NCSP) for breast cancer in the Republic of Korea from a government expenditure perspective. In 2002-2003 (baseline), a total of 8,724,860 women aged 40 years or over were invited to attend breast cancer screening by the NCSP. Those who attended were identified using the NCSP database, and women were divided into two groups, women who attended screening at baseline (screened group) and those who did not (non-screened group). Breast cancer diagnosis in both groups at baseline, and during 5-year follow-up was identified using the Korean Central Cancer Registry. The effectiveness of the NCSP for breast cancer was estimated by comparing 5-year survival and life years saved (LYS) between the screened and the unscreened groups, measured using mortality data from the Korean National Health Insurance Corporation and the National Health Statistical Office. Direct screening costs, indirect screening costs, and productivity costs were considered in different combinations in the model. When all three of these costs were considered together, the incremental cost to save one life year of a breast cancer patient was 42,305,000 Korean Won (KW) (1 USD=1,088 KW) for the screened group compared to the non-screened group. In sensitivity analyses, reducing the false-positive rate of the screening program by half was the most cost-effective (incremental cost-effectiveness ratio, ICER=30,110,852 KW/LYS) strategy. When the upper age limit for screening was set at 70 years, it became more cost-effective (ICER=39,641,823 KW/LYS) than when no upper age limit was set. The NCSP for breast cancer in Korea seems to be accepted as cost-effective as ICER estimates were around the Gross Domestic Product. However, cost-effectiveness could be further improved by increasing the sensitivity of breast cancer screening and by setting appropriate age limits.

• The Journal of Korean Academy of Prosthodontics
• /
• v.19 no.1
• /
• pp.39-45
• /
• 1981

Distal-extension removable partial dentures have long been implicated in the increase in mobility and the destruction of the supporting structures of the primary abutment teeth. Various clasping systems have traditionally been used to retain distal extension removable partial dentures, and other designs have been proposed to minimize torquing forces on the abutment teeth. Most recent studies investigating the effects of removable partial dentures on abutment teeth have been performed in it laboratory setting. Results obtained from in vitro research have given dentists insight into removable partial denture design, but laboratory test model cannot be constructed that simulates actual functional or parafunctiona1 movements and forces. The purpose of this study was to clinically evaluate the degree of tooth mobility produced by two clasping systems (suprabulge type and infrabulge type) used for distal extension removable partial dentures. Akers clasp and R.P.I. system were selected for the evaluation, and four patients required a distal extension removable partial denture on the mandibular arch were selected for participation in the study. Two partial dentures were constructed in the same condition expect the design of clasp. All abutments in the study were mandibular first or second premolars. Measurements of mobility were made with a research tool designed by $M\"{u}hlemann$. This instrument, periodontometer, measures tooth mobility in the mouth by means of a dial gauge accurated to 0.01mm when the tooth is stressed with a force meter. Lingual and buccal deflection of abutment tooth was measured using buccal and lingual pressure. The amount of force applied was 500gm. Tooth mobility tests were made at four key stages; 1. Before insertion of the first removable partial denture, baseline mobility was establsihed. 2. After wearing of the first prosthesis, measurement was made at weekly intervals for 4 weeks. 3. The removable partial denture was then taken from the patient, and tooth mobility was measured again at weekly intervals until the patient's established baseline mobility had returned. 4. The second prosthesis of different clasp design was worn for a month and evaluated in the same manner as the first. The sequence of placement of clasping system was alternated between patients. The following results were obtained from this study; 1. The mobility of abutment tooth increased during the initial stage of wear and returned to baseline mobility after removal of removable partial dentures. 2. The mobility of abutment tooth showed no difference between Akers clasp and I-bar clasp during the 4-week test period. 3. All teeth tested showed greater mobility toward the buccal than the lingual direction.

• Journal of dental hygiene science
• /
• v.15 no.2
• /
• pp.190-195
• /
• 2015

The objective of this study is to estimate the relationship between amalgam removal and urinary mercury levels. To measure urinary mercury concentration, urine of participants was collected at baseline, immediately, 24 hours and 48 hours after removal of amalgam restorations. The statistical analysis was performed using IBM SPSS Statistics ver. 20.0. The overall mean urinary mercury concentration at baseline, immediately, 24 hours and 48 hours after removal of amalgam restorations was 2.77, 2.75, 2.95 and $4.00{\mu}g/g$ creatinine, respectively. Logistic regression model shows that the gender leads to increased odds of high urinary mercury concentration in children (odds ratio, OR=1.99), even after adjusting for high amalgam surfaces (OR=1.23) and fish consumption (OR=1.26) at the baseline. Our findings suggest that mercury exposure from dental amalgam adversely impact health and therefore are a health risk.