• Journal of IKEEE
• /
• v.26 no.2
• /
• pp.176-185
• /
• 2022

Data augmentation is general approach to solve overfitting of machine learning models by applying various data transformations and distortions to dataset. However, when data augmentation is applied in GAN-based model, which is deep learning image generation model, data transformation and distortion are reflected in the generated image, then the generated image quality decrease. To prevent this problem called augmentation leak, we propose a scheme that can prevent augmentation leak regardless of the type and number of augmentations. Specifically, we analyze the conditions of augmentation leak occurrence by type and implement auxiliary augmentation task classifier that can prevent augmentation leak. Through experiments, we show that the proposed technique prevents augmentation leak in the GAN model, and as a result improves the quality of the generated image. We also demonstrate the superiority of the proposed scheme through ablation study and comparison with other representative augmentation leak prevention technique.

• Korean Journal of Veterinary Research
• /
• v.62 no.3
• /
• pp.26.1-26.7
• /
• 2022

This study was aimed to evaluate the prophylactic anti-glaucoma effect of topical 5% betaxolol (BTX) and 2% dorzolamide (DRZ) on the second eye in dogs with unilateral primary angle closure glaucoma (PACG). Medical records of 60 dogs with unilateral PACG who received prophylactic anti-glaucoma eyedrops in the second eye, from 2016 to 2021, were reviewed. The prophylactic effects of BTX were maintained on 28/60 (46.7%) eyes until last visit and BTX failure was observed on median 510 (range, 53-1,927) days in 32/60 (53.3%) eyes. After DRZ instillation in BTX failure eyes, the prophylactic effects were extended at median 610 (range, 157-2,270) days in 21/32 (65.6%) eyes. DRZ failure eyes (17/21, 81.0%) eyes required chemical ablation or surgical intervention due to uncontrolled intraocular pressure. The duration of prophylactic effects was decreased with aging (R² = 0.334, p = 0.006). The predominant breeds were Shih-Tzu (41.9%) and American Cocker Spaniel (30.6%) with no significant differences in survival curves (p = 0.210). The average prophylactic effects of BTX persisted more than 1.5 year and could be selected the first prophylactic eye drop in unilateral PACG. Also, early surgical intervention should be considered in prophylactic medications failure cases.

• Journal of Chest Surgery
• /
• v.55 no.5
• /
• pp.364-377
• /
• 2022

An anatomical understanding of the atrial myocardium is crucial for surgeons and interventionists who treat atrial arrhythmias. We reviewed the anatomy of the inter-nodal and intra-atrial conduction systems. The anterior inter-nodal route (#1) arises from the sinus node and runs through the ventral wall of the atrial chambers. The major branch of route #1 approaches the atrioventricular node from the anterior aspect. Other branches of route #1 are Bachmann's bundle and a vestibular branch around the tricuspid valve. The middle inter-nodal route (#2) begins with a broad span of fibers at the sinus venarum and extends to the superior limbus of the oval fossa. The major branch of route #2 joins with the branch of route #1 at the anterior part of the atrioventricular node. The posterior inter-nodal route (#3) is at the terminal crest and gives rise to many branches at the pectinate muscles of the right atrium and then approaches the posterior atrioventricular node after joining with the vestibular branch of route #1. The branches of the left part of Bachmann's bundle and the branches of the second inter-nodal route form a thin myocardial network at the posterior wall of the left atrium. These anatomical structures could be categorized into major routes and side branches. There are 9 or more anatomical circles in the atrial chambers that could be structural sites for macro re-entry. The implications of normal and abnormal structures of the myocardium for the pathogenesis and treatment of atrial arrhythmias are discussed.

• Archives of Plastic Surgery
• /
• v.48 no.4
• /
• pp.347-352
• /
• 2021

Background Mastectomy leaves unsightly scarring, which can be distressing to patients. Laser therapy for scar prevention has been consistently emphasized in recent studies showing that several types of lasers, including fractional ablation lasers, are effective for reducing scar formation. Nonetheless, there are few studies evaluating the therapeutic efficacy of ablative CO₂ fractional lasers (ACFLs). Methods This study had a randomized, comparative, prospective, split-scar design with blinded evaluation of mastectomy scars. Fifteen patients with mastectomy scars were treated using an ACFL. Half of each scar was randomized to "A," while the other side was allocated to group "B." Laser treatment was conducted randomly. Scars were assessed using digital photographs of the scar and Vancouver scar scale (VSS) scores. Histological assessments were also done. Results The mean VSS scores were 2.20±1.28 for the treatment side and 2.96±1.40 for the control side. There was a significant difference in the VSS score between the treatment side and the control side (P=0.002). The mean visual analog scale (VAS) scores were 4.13±1.36 for the treatment side and 4.67±1.53 for the control side. There was a significant difference in VAS score between the treatment side and the control side (P=0.02). Conclusions This study demonstrated that early scar treatment using an ACFL significantly improved the clinical results of the treatment compared to the untreated scar, and this difference was associated with patient satisfaction.

• Korean Journal of Exercise Nutrition
• /
• v.25 no.4
• /
• pp.54-58
• /
• 2021

[Purpose] Deleted in breast cancer 1 (DBC1) ablation causes obesity, and stearoyl-CoA desaturase 1 (SCD1) induces the biosynthesis of monounsaturated fatty acids. This study examined whether voluntary wheel running (VWR) alters SCD-1 and DBC1 protein levels in the liver of leptin-deficient ob/ob mice. [Methods] Twenty-five Ob/Ob mice were divided into two groups (ob/ob-Sed and ob/ob-Ex). The expression of DBC1 and SCD1 in the mouse liver was determined using western blotting. [Results] After 10 weeks, VWR significantly reduced body weight without affecting the fatty acid synthase and CD36 protein levels. The average daily running distance was 4.0±1.0 km/day. This improvement was associated with changes in the hepatic SCD1 and DBC1 levels. Hepatic SCD-1 protein levels increased significantly, and DBC1 protein levels decreased in ob/ob-Sed animals. On the other hand, VWR inhibited the obesity-induced increase in SCD1 expression and impaired the obesity-induced decrease in DBC1 expression in the liver of leptin-deficient ob/ob mice. [Conclusion] This is the first study showing that VWR has strong effects on hepatic SCD1 and DBC1 in ob/ob mice, and provides key insights into the effects of exercise on obesity.

• Smart Structures and Systems
• /
• v.29 no.1
• /
• pp.29-39
• /
• 2022

Crack detection plays an important role in the maintenance and protection of steel box girder of bridges. However, since the cracks only occupy an extremely small region of the high-resolution images captured from actual conditions, the existing methods cannot deal with this kind of image effectively. To solve this problem, this paper proposed a novel three-stage method based on deep learning technology and morphology operations. The training set and test set used in this paper are composed of 360 images (4928 × 3264 pixels) in steel girder box. The first stage of the proposed model converted high-resolution images into sub-images by using patch-based method and located the region of cracks by CBAM ResNet-50 model. The Recall reaches 0.95 on the test set. The second stage of our method uses the Attention U-Net model to get the accurate geometric edges of cracks based on results in the first stage. The IoU of the segmentation model implemented in this stage attains 0.48. In the third stage of the model, we remove the wrong-predicted isolated points in the predicted results through dilate operation and outlier elimination algorithm. The IoU of test set ascends to 0.70 after this stage. Ablation experiments are conducted to optimize the parameters and further promote the accuracy of the proposed method. The result shows that: (1) the best patch size of sub-images is 1024 × 1024. (2) the CBAM ResNet-50 and the Attention U-Net achieved the best results in the first and the second stage, respectively. (3) Pre-training the model of the first two stages can improve the IoU by 2.9%. In general, our method is of great significance for crack detection.

• The Korean Journal of Pain
• /
• v.36 no.4
• /
• pp.465-472
• /
• 2023

Background: Percutaneous radiofrequency thermocoagulation (RFTC) has been widely utilized in the management of trigeminal neuralgia. Despite using image guidance, accurate needle positioning into the target area still remains a critical element for achieving a successful outcome. This study was performed to precisely clarify the anatomical information required to ensure that the electrode tip is placed on the sensory component of the mandibular nerve (MN) at the foramen ovale (FO) level. Methods: The study used 50 hemi-half heads from 26 South Korean adult cadavers. Results: The cross-sectioned anterior and posterior divisions of the MN at the FO level could be distinguished based on an irregular boundary and color difference. The anterior division was clearly brighter than the posterior one. The anterior division of the MN at the FO level was located at the whole anterior (38.0%), anteromedial (6.0%), anterior center (8.0%), and anterolateral (22.0%) parts. The posterior division was often located at the whole posterior or posterolateral parts of the MN at the FO level. The anterior divisions covered the whole MN except for the medial half of the posterolateral part in the overwrapped images of the cross-sectional areas of the MN at the FO level. The cross-sectional areas of the anterior divisions were similar in males and females, whereas those of the posterior divisions were significantly larger in males (P = 0.004). Conclusions: The obtained anatomical information is expected to help physicians reduce unwanted side effects after percutaneous RFTC within the FO for the MN.

• Medical Lasers
• /
• v.8 no.1
• /
• pp.7-12
• /
• 2019

Background and Objectives Skin and soft tissue defects can be treated according to a range of strategies, such as local flap, skin graft, biological dressing, or free flap. On the other hand, free tissue transfer usually leaves a distinct scar with an inconsistency of color or hypertrophy. This problem is highlighted if the defect is located on the face, which could have devastating effects on a patient's psychosocial health. Materials and Methods The authors used an erbium : yttrium-aluminum-garnet (Er:YAG) laser to resurface the free flap skin and match the color with the surrounding facial skin. This study evaluated the effectiveness of laser skin resurfacing on the harmonious color matching of transferred flap. Patients who had undergone laser resurfacing on facial flap skin between January 2014 and December 2018 were reviewed retrospectively. An ablative 2,940-nm fractional Er:YAG laser treatment was delivered to the entire flap skin at 21 J/cm² with the treatment end-point of pinpoint bleeding. Several months later, the clinical photographs were analyzed. The L^*a^*b^* color co-ordinates of both the flap and surrounding normal skin were measured using Adobe Photoshop. The L^*a^*b^* color difference (ΔE) for the scar and normal surrounding skin were calculated using the following equation: ${\Delta}E=\sqrt{({\Delta}L)^2+({\Delta}a)^2+({\Delta}b)^2}$ Results All five patients were satisfied with the more natural appearance of the flaps. The ΔE values decreased significantly from the pre-treatment mean value of 19.64 to the post-treatment mean value of 11.39 (Wilcoxon signed-rank test, p = 0.043). Conclusion Ablative laser resurfacing can improve the aesthetic outcome of free tissue transfer on the face.

• Medical Lasers
• /
• v.8 no.1
• /
• pp.13-18
• /
• 2019

Background and Objectives High-intensity focused ultrasound (HIFU) can produce small zones of thermal damage. A HIFU procedure is non-invasive and it can achieve rejuvenation of facial skin. Fractional CO₂ laser resurfacing delivers thermal damage to the pixilated columnar zone of the skin and so evoke collagen remodeling, the same as HIFU. In many cases, the patients who want rejuvenation with HIFU are also good candidates for cutaneous photorejuvenation such as can be accomplished via fractional CO₂ resurfacing. If patients are treated in a single session by remodeling both the superficial and deep compartments of skin by using both modalities, then improvement in rhytides and tightening of sagging skin will optimize the aesthetic result. Materials and Methods Between May 2014 and January 2018, a total of 44 patients were treated with combination HIFU and fractional CO₂ laser resurfacing according to our protocol. First, the HIFU was applied to the entire face with an average of 300 treatment lines. Immediately after HIFU treatment, the ultrasound gel was washed off and then fractional CO₂ laser resurfacing was performed. We evaluated the patients using 4-point grading scales. The clinician examined the skin for evidence of complications after the completion of treatment. Results All the patients' skin quality showed improvement. Further. the clinical results after duel modality treatment were substantially better than that after the use of either modality alone. The recovery times and the incidence of adverse events when quickly and consecutively performing both treatments were similar as compared to those with employing stepwise treatment. We encountered no complications whatsoever. Conclusion When compared with stepwise therapy, combination therapy with HIFU and fractional CO₂ resurfacing offers better, safer and more effective clinical results. Thus, for targeting multiple layers of aging facial skin, this combination therapy can be safely performed in a single treatment session.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.17 no.5
• /
• pp.1433-1449
• /
• 2023

In recent years, the rapid development of social networks has led to a rapid increase in the amount of information available on the Internet, which contains a large amount of sensitive information related to pornography, politics, and terrorism. In the aspect of sensitive image detection, the existing machine learning algorithms are confronted with problems such as large model size, long training time, and slow detection speed when auditing and supervising. In order to detect sensitive images more accurately and quickly, this paper proposes a multiclassification sensitive image detection method based on lightweight Convolutional Neural Network. On the basis of the EfficientNet model, this method combines the Ghost Module idea of the GhostNet model and adds the SE channel attention mechanism in the Ghost Module for feature extraction training. The experimental results on the sensitive image data set constructed in this paper show that the accuracy of the proposed method in sensitive information detection is 94.46% higher than that of the similar methods. Then, the model is pruned through an ablation experiment, and the activation function is replaced by Hard-Swish, which reduces the parameters of the original model by 54.67%. Under the condition of ensuring accuracy, the detection time of a single image is reduced from 8.88ms to 6.37ms. The results of the experiment demonstrate that the method put forward has successfully enhanced the precision of identifying multi-class sensitive images, significantly decreased the number of parameters in the model, and achieved higher accuracy than comparable algorithms while using a more lightweight model design.