• Korean Journal of Agricultural Science
• /
• v.47 no.4
• /
• pp.1109-1122
• /
• 2020

The aim of this study was to develop a marbling classification and prediction model using small parts of sirloin images based on a deep learning algorithm, namely, a convolutional neural network (CNN). Samples were purchased from a commercial slaughterhouse in Korea, images for each grade were acquired, and the total images (n = 500) were assigned according to their grade number: 1++, 1+, 1, and both 2 & 3. The image acquisition system consists of a DSLR camera with a polarization filter to remove diffusive reflectance and two light sources (55 W). To correct the distorted original images, a radial correction algorithm was implemented. Color images of sirloins of Hanwoo (mixed with feeder cattle, steer, and calf) were divided and sub-images with image sizes of 161 × 161 were made to train the marbling prediction model. In this study, the convolutional neural network (CNN) has four convolution layers and yields prediction results in accordance with marbling grades (1++, 1+, 1, and 2&3). Every single layer uses a rectified linear unit (ReLU) function as an activation function and max-pooling is used for extracting the edge between fat and muscle and reducing the variance of the data. Prediction accuracy was measured using an accuracy and kappa coefficient from a confusion matrix. We summed the prediction of sub-images and determined the total average prediction accuracy. Training accuracy was 100% and the test accuracy was 86%, indicating comparably good performance using the CNN. This study provides classification potential for predicting the marbling grade using color images and a convolutional neural network algorithm.

• The Korean Journal of Emergency Medical Services
• /
• v.25 no.1
• /
• pp.7-22
• /
• 2021

Purpose: This study aimed to identify the differences in pre-hospital and in-hospital triage (pre-hospital triage and KTAS(Korean Triage and Acuity Scale)) of patients with abdominal pain and the characteristics of high hospitalization probability in the treatment results. Methods: We analyzed 941 people who visited the area C emergency center for 2 years from January 2017 to December 2018. The collected data were analyzed using SPSS 26. Results: Among the transfer hospitals, 84.8% (798) of patients were selected by the 119 rescue service, and the most common diagnosis was simple abdominal pain (46.5%, 438 patients). A total of 50.7% (477) of patients classified as severe pre-hospital cases changed to mild in-hospital cases. There was a difference of 5.3% (50 cases) in cases where patients classified as mild pre-hospital were changed to severe in-hospital cases. The Kappa coefficient did not match with 0.04 (p=.051). Pre-hospital overtriage was 58.2% (548 cases), and 71.2% (670) of patients were discharged from the emergency room as a result of the treatment. Conclusion: The results of this study showed that pre-hospital and in-hospital triage were not consistent. The rates of pre-hospital overtriage were quite high. Most patients with abdominal pain were classified as mild cases, and pre-hospital triage classifiers should be trained to reduce errors in selecting transfer hospitals.

• Korean Journal of Remote Sensing
• /
• v.38 no.1
• /
• pp.111-126
• /
• 2022

Land use and land cover (LULC) mapping is an important factor in geospatial analysis. Although highly precise ground-based LULC monitoring is possible, it is time consuming and costly. Conversely, because the synthetic aperture radar (SAR) sensor is an all-weather sensor with high resolution, it could replace field-based LULC monitoring systems with low cost and less time requirement. Thus, LULC is one of the major areas in SAR applications. We developed a LULC model using only KOMPSAT-5 single co-polarized data and digital elevation model (DEM) data. Twelve HH-polarized images and 18 VV-polarized images were collected, and two HH-polarized images and four VV-polarized images were selected for the model testing. To train the LULC model, we applied the conditional generative adversarial network (cGAN) method. We used U-Net combined with the residual unit (ResUNet) model to generate the cGAN method. When analyzing the training history at 1732 epochs, the ResUNet model showed a maximum overall accuracy (OA) of 93.89 and a Kappa coefficient of 0.91. The model exhibited high performance in the test datasets with an OA greater than 90. The model accurately distinguished water body areas and showed lower accuracy in wetlands than in the other LULC types. The effect of the DEM on the accuracy of LULC was analyzed. When assessing the accuracy with respect to the incidence angle, owing to the radar shadow caused by the side-looking system of the SAR sensor, the OA tended to decrease as the incidence angle increased. This study is the first to use only KOMPSAT-5 single co-polarized data and deep learning methods to demonstrate the possibility of high-performance LULC monitoring. This study contributes to Earth surface monitoring and the development of deep learning approaches using the KOMPSAT-5 data.

• Clinics in Shoulder and Elbow
• /
• v.25 no.3
• /
• pp.182-187
• /
• 2022

Background: This study was designed to evaluate and compare the diagnostic value of magnetic resonance imaging (MRI) and indirect magnetic resonance arthrography (I-MRA) imaging with those of arthroscopy and each other. Methods: This descriptive-analytical study was conducted in 2020. All patients who tested positive for labrum lesions during that year were included in the study. The patients underwent conservative treatment for 6 weeks. In the event of no response to conservative treatment, MRI and I-MRA imaging were conducted, and the patients underwent arthroscopy to determine their ultimate diagnosis and treatment plan. Imaging results were assessed at a 1-week interval by an experienced musculoskeletal radiologist. Image interpretation results and arthroscopy were recorded in the data collection form. Results: Overall, 35 patients comprised the study. Based on the kappa coefficient, the results indicate that the results of both imaging methods are in agreement with the arthroscopic findings, but the I-MRA consensus rate is higher than that of MRI (0.612±0.157 and 0.749±0.101 vs. 0.449±0.160 and 0.603±0.113). The sensitivity, specificity, negative predictive value, positive predictive value, and accuracy of MRI in detecting labrum tears were 77.77%, 75.00%, 91.30%, 50.00%, and 77.14%, respectively, and those of I-MRA were 88.88%, 75.00%, 92.30%, 66.66%, and 85.71%. Conclusions: Here, I-MRA showed higher diagnostic value than MRI for labral tears. Therefore, it is recommended that I-MRA be used instead of MRI if there is an indication for potential labrum lesions.

• Journal of Veterinary Science
• /
• v.22 no.1
• /
• pp.14.1-14.11
• /
• 2021

Background: Quantitation of urine protein is important in dogs with chronic kidney disease. Various analyzers are used to measure urine protein-to-creatinine ratios (UPCR). Objectives: This study aimed to compare the UPCR obtained by three types of analyzers (automated wet chemistry analyzer, in-house dry chemistry analyzer, and dipstick reading device) and investigate whether the differences could affect clinical decision process. Methods: Urine samples were collected from 115 dogs. UPCR values were obtained using three analyzers. Bland-Altman and Passing Bablok tests were used to analyze agreement between the UPCR values. Urine samples were classified as normal or proteinuria based on the UPCR values obtained by each analyzer and concordance in the classification evaluated with Cohen's kappa coefficient. Results: Passing and Bablok regression showed that there were proportional as well as constant difference between UPCR values obtained by a dipstick reading device and those obtained by the other analyzers. The concordance in the classification of proteinuria was very high (κ = 0.82) between the automated wet chemistry analyzer and in-house dry chemistry analyzer, while the dipstick reading device showed moderate concordance with the automated wet chemistry analyzer (κ = 0.52) and in-house dry chemistry analyzer (κ = 0.53). Conclusions: Although the urine dipstick test is simple and a widely used point-of-care test, our results indicate that UPCR values obtained by the dipstick test are not appropriate for clinical use. Inter-instrumental variability may affect clinical decision process based on UPCR values and should be emphasized in veterinary practice.

• Journal of Dental Anesthesia and Pain Medicine
• /
• v.23 no.1
• /
• pp.9-17
• /
• 2023

Background: Dental evaluation and protection are important for preventing traumatic dental injuries when patients are under general anesthesia. The objective of the present study was to develop a questionnaire based on dentition-related risk factors that could serve as a valuable tool for dental evaluation and documentation. Methods: We developed a questionnaire for dental evaluation before administration of general anesthesia, investigated the association between patient-and-dentist responses and mouthguard fabrication, and assessed response agreement between 100 patients. Results: Protective mouthguards were fabricated for 27 patients who were identified as having a high risk of dental injury. There was a strong association between dentists' responses and mouthguard fabrication, depending on the general oral health status, use of ceramic prosthesis, presence of masticatory pain related to periodontal diseases, gingival edema, and implants (P < 0.05). Response agreement between patients and dentists for items related to dental pain, loss of dental pulp vitality, root canal therapy, dental trauma, aesthetic prosthesis, tooth mobility, and implant prosthesis was high (Cohen's kappa coefficient κ ≥ 0.6). Conclusions: A high agreement was observed between patient-dentist responses and a strong association with mouthguard fabrication for items pertaining to ceramic prosthesis, masticatory pain, and dental implants. Patients with a "yes" response to these items are recommended to undergo a dental evaluation and use a dental protective device while under general anesthesia.

• Journal of Korean Society of Rural Planning
• /
• v.28 no.1
• /
• pp.57-69
• /
• 2022

Crop classification is very important for estimating crop yield and figuring out accurate cultivation area. The purpose of this study is to classify crops harvested in fall in Idam-ri, Goesan-gun, Chungcheongbuk-do by using unmanned aerial vehicle (UAV) images and support vector machine (SVM) model. The study proceeded in the order of image acquisition, variable extraction, model building, and evaluation. First, RGB and multispectral image were acquired on September 13, 2021. Independent variables which were applied to Farm-Map, consisted gray level co-occurrence matrix (GLCM)-based texture characteristics by using RGB images, and multispectral reflectance data. The crop classification model was built using texture characteristics and reflectance data, and finally, accuracy evaluation was performed using the error matrix. As a result of the study, the classification model consisted of four types to compare the classification accuracy according to the combination of independent variables. The result of four types of model analysis, recursive feature elimination (RFE) model showed the highest accuracy with an overall accuracy (OA) of 88.64%, Kappa coefficient of 0.84. UAV-based RGB and multispectral images effectively classified cabbage, rice and soybean when the SVM model was applied. The results of this study provided capacity usefully in classifying crops using single-period images. These technologies are expected to improve the accuracy and efficiency of crop cultivation area surveys by supplementing additional data learning, and to provide basic data for estimating crop yields.

• Korean Journal of Remote Sensing
• /
• v.40 no.1
• /
• pp.57-69
• /
• 2024

Crop classification plays a vitalrole in monitoring agricultural landscapes and enhancing food production. In this study, we explore the effectiveness of Long Short-Term Memory (LSTM) models for crop classification, focusing on distinguishing between apple and rice crops. The aim wasto overcome the challenges associatedwith finding phenology-based classification thresholds by utilizing LSTM to capture the entire Normalized Difference Vegetation Index (NDVI)trend. Our methodology involvestraining the LSTM model using a reference site and applying it to three separate three test sites. Firstly, we generated 25 NDVI imagesfrom the Sentinel-2A data. Aftersegmenting study areas, we calculated the mean NDVI values for each segment. For the reference area, employed a training approach utilizing the NDVI trend line. This trend line served as the basis for training our crop classification model. Following the training phase, we applied the trained model to three separate test sites. The results demonstrated a high overall accuracy of 0.92 and a kappa coefficient of 0.85 for the reference site. The overall accuracies for the test sites were also favorable, ranging from 0.88 to 0.92, indicating successful classification outcomes. We also found that certain phenological metrics can be less effective in crop classification therefore limitations of relying solely on phenological map thresholds and emphasizes the challenges in detecting phenology in real-time, particularly in the early stages of crops. Our study demonstrates the potential of LSTM models in crop classification tasks, showcasing their ability to capture temporal dependencies and analyze timeseriesremote sensing data.While limitations exist in capturing specific phenological events, the integration of alternative approaches holds promise for enhancing classification accuracy. By leveraging advanced techniques and considering the specific challenges of agricultural landscapes, we can continue to refine crop classification models and support agricultural management practices.

• Imaging Science in Dentistry
• /
• v.53 no.4
• /
• pp.375-381
• /
• 2023

Purpose: This study assessed the diagnostic performance of stitched and non-stitched cross-sectional cone-beam computed tomography (CBCT) images of non-displaced ovine mandibular fractures. Materials and Methods: In this ex vivo study, non-displaced fractures were artificially created in 10 ovine mandibles (20 hemi-mandibles) using a hammer. The control group comprised 8 hemi-mandibles. The non-displaced fracture lines were oblique or vertical, <0.5 mm wide, 10-20 mm long, and only in the buccal or lingual cortex. Fracture lines in the ramus and posterior mandible were created to be at the interface or borders of the 2 stitched images. CBCT images were obtained from the specimens with an 80 mm×80 mm field of view before and after fracture induction. OnDemand software (Cybermed, Seoul, Korea) was used for stitching the CBCT images. Four observers evaluated 56 (28 stitched and 28 non-stitched) images to detect fracture lines. The diagnostic performance of stitched and non-stitched images was assessed by calculating the area under the receiver operating characteristic curve (AUC). Sensitivity and specificity values were also calculated (alpha=0.05). Results: The AUC was calculated to be 0.862 and 0.825 for the stitched and non-stitched images, respectively (P=0.747). The sensitivity and specificity were 90% and 75% for the non-stitched images and 85% and 87% for the stitched images, respectively. The inter-observer reliability was shown by a Fleiss kappa coefficient of 0.79, indicating good agreement. Conclusion: No significant difference was found in the diagnostic performance of stitched and non-stitched cross-sectional CBCT images of non-displaced fractures of the ovine mandible.

• Imaging Science in Dentistry
• /
• v.54 no.3
• /
• pp.276-282
• /
• 2024

Purpose: This study aimed to assess the performance of 2-dimensional (2D) imaging with microscopy coils in delineating teeth and periodontal tissues compared with conventional 3-dimensional(3D) imaging on a 3 T magnetic resonance imaging (MRI) unit. Materials and Methods: Twelve healthy participants (4 men and 8 women; mean age: 25.6 years; range: 20-52 years) with no dental symptoms were included. The left mandibular first molars and surrounding periodontal tissues were examined using the following 2 sequences: 2D proton density-weighted (PDw) images and 3D enhanced T1 high-resolution isotropic volume excitation (eTHRIVE) images. Two-dimensional MRI images were taken using a 3 T MRI unit and a 47 mm microscopy coil, while 3D MRI imaging used a 3 T MRI unit and head-neck coil. Oral radiologists assessed dental and periodontal structures using a 4-point Likert scale. Inter- and intra-observer agreement was determined using the weighted kappa coefficient. The Wilcoxon signed-rank test was used to compare 2D-PDw and 3D-eTHRIVE images. Results: Qualitative analysis showed significantly better visualization scores for 2D-PDw imaging than for 3D-eTHRIVE imaging (Wilcoxon signed-rank test). 2D-PDw images provided improved visibility of the tooth, root dental pulp, periodontal ligament, lamina dura, coronal dental pulp, gingiva, and nutrient tract. Inter-observer reliability ranged from moderate agreement to almost perfect agreement, and intra-observer agreement was in a similar range. Conclusion: Two-dimensional-PDw images acquired using a 3 T MRI unit and microscopy coil effectively visualized nearly all aspects of teeth and periodontal tissues.