• International conference on construction engineering and project management
• /
• 2024.07a
• /
• pp.1295-1295
• /
• 2024

As extreme weather events such as heavy snowfall and typhoon become more frequent, climate change significantly impacts across various worldwide industries. With demands for dealing with this phenomenon, continuous achievements in safety diagnosis have been announced for large structures. Conversely, in agricultural infrastructures having lower risk to human life, there is lack of established safety diagnosis methods. However, considering expansion of high-value smart farm, the importance of plastic greenhouse cannot be overlooked. Therefore, this study aimed to develop on-site diagnosis technique for structural safety of steel structure greenhouse. To build an analysis model, we generated point cloud data of on-site greenhouse using a camera with LiDAR sensor. Subsequently, we extracted points corresponding to pipes using a pre-trained semantic segmentation model, achieving a pipe segmentation accuracy of 78.1%. These points were then converted into 3D frame model, with a location coordinate error of 5.4 cm for nine reference points, as measured by an on-site survey. In FEM structural analysis, nonlinearity of pipe connection was reflected. The loads were determined based on expected wind speed and snow depth in Korea. The structural safety of on-site model was diagnosed more vulnerable with 10.3% higher maximum axial stress, compared with standard model. Through this research, we expect the quantitative safety diagnosis of predicting greenhouse collapse risk. In addition, this technique will enable localized reinforcement strategies within the structure.

• Journal of Intelligence and Information Systems
• /
• v.24 no.4
• /
• pp.219-240
• /
• 2018

Sentiment analysis, which is one of the text mining techniques, is a method for extracting subjective content embedded in text documents. Recently, the sentiment analysis methods have been widely used in many fields. As good examples, data-driven surveys are based on analyzing the subjectivity of text data posted by users and market researches are conducted by analyzing users' review posts to quantify users' reputation on a target product. The basic method of sentiment analysis is to use sentiment dictionary (or lexicon), a list of sentiment vocabularies with positive, neutral, or negative semantics. In general, the meaning of many sentiment words is likely to be different across domains. For example, a sentiment word, 'sad' indicates negative meaning in many fields but a movie. In order to perform accurate sentiment analysis, we need to build the sentiment dictionary for a given domain. However, such a method of building the sentiment lexicon is time-consuming and various sentiment vocabularies are not included without the use of general-purpose sentiment lexicon. In order to address this problem, several studies have been carried out to construct the sentiment lexicon suitable for a specific domain based on 'OPEN HANGUL' and 'SentiWordNet', which are general-purpose sentiment lexicons. However, OPEN HANGUL is no longer being serviced and SentiWordNet does not work well because of language difference in the process of converting Korean word into English word. There are restrictions on the use of such general-purpose sentiment lexicons as seed data for building the sentiment lexicon for a specific domain. In this article, we construct 'KNU Korean Sentiment Lexicon (KNU-KSL)', a new general-purpose Korean sentiment dictionary that is more advanced than existing general-purpose lexicons. The proposed dictionary, which is a list of domain-independent sentiment words such as 'thank you', 'worthy', and 'impressed', is built to quickly construct the sentiment dictionary for a target domain. Especially, it constructs sentiment vocabularies by analyzing the glosses contained in Standard Korean Language Dictionary (SKLD) by the following procedures: First, we propose a sentiment classification model based on Bidirectional Long Short-Term Memory (Bi-LSTM). Second, the proposed deep learning model automatically classifies each of glosses to either positive or negative meaning. Third, positive words and phrases are extracted from the glosses classified as positive meaning, while negative words and phrases are extracted from the glosses classified as negative meaning. Our experimental results show that the average accuracy of the proposed sentiment classification model is up to 89.45%. In addition, the sentiment dictionary is more extended using various external sources including SentiWordNet, SenticNet, Emotional Verbs, and Sentiment Lexicon 0603. Furthermore, we add sentiment information about frequently used coined words and emoticons that are used mainly on the Web. The KNU-KSL contains a total of 14,843 sentiment vocabularies, each of which is one of 1-grams, 2-grams, phrases, and sentence patterns. Unlike existing sentiment dictionaries, it is composed of words that are not affected by particular domains. The recent trend on sentiment analysis is to use deep learning technique without sentiment dictionaries. The importance of developing sentiment dictionaries is declined gradually. However, one of recent studies shows that the words in the sentiment dictionary can be used as features of deep learning models, resulting in the sentiment analysis performed with higher accuracy (Teng, Z., 2016). This result indicates that the sentiment dictionary is used not only for sentiment analysis but also as features of deep learning models for improving accuracy. The proposed dictionary can be used as a basic data for constructing the sentiment lexicon of a particular domain and as features of deep learning models. It is also useful to automatically and quickly build large training sets for deep learning models.

• Journal of Intelligence and Information Systems
• /
• v.25 no.1
• /
• pp.163-177
• /
• 2019

As smartphones are getting widely used, human activity recognition (HAR) tasks for recognizing personal activities of smartphone users with multimodal data have been actively studied recently. The research area is expanding from the recognition of the simple body movement of an individual user to the recognition of low-level behavior and high-level behavior. However, HAR tasks for recognizing interaction behavior with other people, such as whether the user is accompanying or communicating with someone else, have gotten less attention so far. And previous research for recognizing interaction behavior has usually depended on audio, Bluetooth, and Wi-Fi sensors, which are vulnerable to privacy issues and require much time to collect enough data. Whereas physical sensors including accelerometer, magnetic field and gyroscope sensors are less vulnerable to privacy issues and can collect a large amount of data within a short time. In this paper, a method for detecting accompanying status based on deep learning model by only using multimodal physical sensor data, such as an accelerometer, magnetic field and gyroscope, was proposed. The accompanying status was defined as a redefinition of a part of the user interaction behavior, including whether the user is accompanying with an acquaintance at a close distance and the user is actively communicating with the acquaintance. A framework based on convolutional neural networks (CNN) and long short-term memory (LSTM) recurrent networks for classifying accompanying and conversation was proposed. First, a data preprocessing method which consists of time synchronization of multimodal data from different physical sensors, data normalization and sequence data generation was introduced. We applied the nearest interpolation to synchronize the time of collected data from different sensors. Normalization was performed for each x, y, z axis value of the sensor data, and the sequence data was generated according to the sliding window method. Then, the sequence data became the input for CNN, where feature maps representing local dependencies of the original sequence are extracted. The CNN consisted of 3 convolutional layers and did not have a pooling layer to maintain the temporal information of the sequence data. Next, LSTM recurrent networks received the feature maps, learned long-term dependencies from them and extracted features. The LSTM recurrent networks consisted of two layers, each with 128 cells. Finally, the extracted features were used for classification by softmax classifier. The loss function of the model was cross entropy function and the weights of the model were randomly initialized on a normal distribution with an average of 0 and a standard deviation of 0.1. The model was trained using adaptive moment estimation (ADAM) optimization algorithm and the mini batch size was set to 128. We applied dropout to input values of the LSTM recurrent networks to prevent overfitting. The initial learning rate was set to 0.001, and it decreased exponentially by 0.99 at the end of each epoch training. An Android smartphone application was developed and released to collect data. We collected smartphone data for a total of 18 subjects. Using the data, the model classified accompanying and conversation by 98.74% and 98.83% accuracy each. Both the F1 score and accuracy of the model were higher than the F1 score and accuracy of the majority vote classifier, support vector machine, and deep recurrent neural network. In the future research, we will focus on more rigorous multimodal sensor data synchronization methods that minimize the time stamp differences. In addition, we will further study transfer learning method that enables transfer of trained models tailored to the training data to the evaluation data that follows a different distribution. It is expected that a model capable of exhibiting robust recognition performance against changes in data that is not considered in the model learning stage will be obtained.

• Korean Journal of Environmental Agriculture
• /
• v.36 no.3
• /
• pp.154-160
• /
• 2017

BACKGROUND:This study was carried out to establish an efficient sample preparation for the simultaneous determination of bisphenols (BPs) in river water samples using gas chromatography-mass spectrometry (GC-MS). Sample preparation was examined with conventional extraction methods, such as solid-phase extraction (SPE) and liquid-liquid extraction (LLE), and their efficiency was compared with validation results, including linearity of calibration curve, method detection limit (MDL), limit of quantification (LOQ), accuracy, and precision. METHODS AND RESULTS:The BPs (bisphenol A, BPA; bisphenol B, BPB; bisphenol C, BPC; bisphenol E, BPE; bisphenol F, BPF; bisphenol S, BPS) were analyzed using GC-MS. The range of MDLs by SPE and LLE methods was $0.0005{\sim}0.0234{\mu}g/L$ and $0.0037{\sim}0.2034{\mu}g/L$, and that of LOQs was $0.0015{\sim}0.0744{\mu}g/L$ and $0.0117{\sim}0.6477{\mu}g/L$, respectively. The calibration curve obtained from standard solution of $0.004{\sim}4.0{\mu}g/L$ (SPE) and $0.016{\sim}16{\mu}g/L$ (LLE) showed good linearity with $r^2$ value of 0.9969 over. Accuracy was 93.2~108% and 97.4~120%, and precision was 1.7~4.6% and 0.7~6.5%, respectively. The values of MDL and LOQ resulted from the SPE method were higher than those from the LLE method, particularly those values of BPA were highest among the BPs. Based on the results, the SPE method was applied to determine the BPs in river water samples. Water samples were collected from mainstream, tributary and sewage wastewater treatment plants (SWTPs) in the Yeongsan river basin. The concentration of BPB, BPC, BPE, BPF and BPS were not detected in all sites, whereas BPA was ranged $0.0095{\sim}0.2583{\mu}g/L$, which was $0.0166{\sim}0.0810{\mu}g/L$ for mainstreams, $0.0095{\sim}0.2583{\mu}g/L$ for tributaries, $0.0352{\sim}0.1217{\mu}g/L$ for SWTPs. CONCLUSION: From these results, the SPE method was very effective for the simultaneous determination of BPs in river water samples using GC-MS. We provided that it is a convenient, reliable and sensitive method enough to monitor and understand the fate of the BPs in aquatic ecosystems.

• Investigative Magnetic Resonance Imaging
• /
• v.13 no.2
• /
• pp.152-160
• /
• 2009

Purpose : To intra-individually compare diagnostic accuracy of high-resolution contrast-enhanced magnetic resonance angiography (CE-MRA) with computed tomography angiography (CTA) and digital subtraction angiography (DSA) for the assessment of supraaortic steno-occlusive disease. Materials and Methods : Twenty-eight patients (20 men, 8 women, 53-79 years of age) underwent supraaortic CE-MRA, CTA and DSA. CE-MRA was performed on two 1.5T MR scanners (voxel dimension: $0.66{\times}0.66{\times}1.1$ or $1.2\;mm^3$), and CTA on 64-slice CT scanners (voxel dimension: $0.42{\times}0.42{\times}0.63\;mm^3$). All the three examinations were completed within 40 days (median 19 days; range 1-40 days). Retrospective evaluation and measurement of diameter of 6 extracranial and 9 intracranial arterial segments was done by 2 experienced radiologists. Results: A total of 420 arterial segments were examined by CE-MRA, CTA and DSA. On DSA, 34 stenoocclusive lesions were noted at extracranial (n= 19) and intracranial (n = 15) vessels. For extracranial stenosis greater than 70%, sensitivity, specificity, positive predictive value (PPV) and negative predictive values (NPV) were 94.7%, 98.7%, 90.0% and 99.3% on CE-MRA, and 94.7%, 99.3%, 94.7% and 99.3% on CTA. For intracranial stenosis greater than 50%, sensitivity; specificity, PPV and NPV were 93.3%, 98.3%, 77.8%and 99.6% on CE-MRA, and 86.7%, 97.9%, 72.2% and 99.1 % on CTA, with DSA as the standard of reference. Conclusion : Supraaortic CE-MRA is as reliable as CTA in depicting the arterial stenosis, and is effective in screening of significant stenosis of both extracranial and intracranial arterial stenosis.

• Progress in Medical Physics
• /
• v.17 no.3
• /
• pp.179-186
• /
• 2006

Purpose: Photon beam data of linear accelerators in Korea are collected, analyzed, and a simple method for checking and verifying the dose calculations in a TPS are suggested. Materials and Methods: Photon beam data such as output calibration condition, output factor, wedge factor, percent depth dose, beam profile, and beam quality were collected from 26 institutions in Korea. In order to verify the accuracy of dose calculation, ten sample planning tests were peformed. These Include square, elongated, and blocked fields, wedge fields, off-axis dose calculation, SSD variation. The planned data were compared to that of manual calculations. Results: The average and standard deviation of photon beam quality for 6, 10, and 15 MV were $0.576{\pm}0.005,\;0.632{\pm}0.004,\;and\;0.647{\pm}0.006$, respectively. The output factors of 6 MV photon beam measured at depth of dose maximum for $5{\times}5cm,\;15{\times}15cm,\;20{\times}20cm\;were\;0.944{\pm}0.006,\;1.031{\pm}0.006,\;and\;1.055{\pm}0.007$. For 10 MV photon beam, the values were $0.935{\pm}0.006,\;1.031{\pm}0.007,\;1.054{\pm}0.0005$. The collected data were not enough to calculate average, the output factors for 15MV photon beam with field size of $5{\times}5cm,\;15{\times}15cm,\;20{\times}20cm\;were\;0.941{\pm}0.008,\;1.032{\pm}0.004,\;1.049{\pm}0.014$. There was seven institutions $e{\times}ceeding$ tolerance when monitor unit values calculated from treatment planning system and manually were compared. The measured average MU values for the machines calibrated at SAD setup were 3 MU and 5 MU higher than the machines calibrated at SSD for 6 MV and 10 MV, respectively except the wedge case. When the wedges were inserted, the MU values to deliver 100 cGy to 5 cm depends on manufactures. When the same wedge angle was used, Siemens machine requires more MUs then Varian machine. Conclusion: In this study, photon beam data are collected and analyzed to provide a baseline value for chocking beam data and the accuracy of dose calculation for a treatment planning system.

• The Korean Journal of Nuclear Medicine
• /
• v.35 no.1
• /
• pp.61-68
• /
• 2001

Purpose: The C-14 urea breath test (C-14 UBT) is the most specific noninvasive method to detect Helicobacter (H) pylori infection. We investigated if the C-14 UBT can reflect the presence and degree of H. pylori detected by gastroduodenoscopic biopsies (GBx). Materials and methods: One hundred fifty patients (M:F=83:67, age $48.6{\pm}11.2$ yrs) underwent C-14 UBT, rapid urease test (CLO test) and GBx on the same day. For the C-14 UBT, a single breath sample was collected at 10 minutes after ingestion of C-14 urea (137 KBq) capsule and counting was done in a liquid scintillation counter for 1 minute, and the results were classified as positive (${\geq}200dpm$), Intermediate ($50{\sim}199dpm$) or negative (<50 dpm). The results of CLO tests were classified as positive or negative according to color change. The results of GBx on giemsa stain were graded 0 (normal) to 4 (diffuse) according to the distribution of H. pylori by the Wyatt method. We compared C-14 UBT results with GBx grade as a gold standard. Results: In the assessment of the presence of H. pylori infection, the C-14 UBT global performance yielded sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and accuracy of 92.5%, 88.4%, 97.1%, 88.4% and 91.3%, respectively. However, the CLO test had sensitivity, specificity, PPV, NPV and accuracy of 83.2%, 81.4%, 91.8%, 81.4% and 82.7%, respectively. The quantitative values of the C-14 UBT were $45{\pm}27$ dpm in grade 0, $707{\pm}584dpm$ in grade 1, $1558{\pm}584dpm$ in grade 2, $1851{\pm}604dpm$ in grade 3, and $2719{\pm}892dpm$ in grade 4. A significant correlation (r=0.848, p<0.01) was found between C-14 UBT and the grade of distribution of H. pylori infection on GBx with giemsa stain. Conclusion: We conclude that the C-14 UBT is a highly accurate, simple and noninvasive method for the diagnosis of ongoing H. pylori infection and reflects the degree of bacterial distribution.

• Nuclear Medicine and Molecular Imaging
• /
• v.42 no.3
• /
• pp.229-234
• /
• 2008

Purpose: A urea breath test (UBT) using C-14 or C-13 has been developed for identifying Helicobacter (H) pylori infection on the basis of urease production with release of labeled $CO_2$. We investigated if the C-14 and C-13 UBT have the difference to reflect the presence and degree of H. pylori infection detected by gastro-duodenoscopic biopsies (CBx) in the same patients. Materials and methods: Thirty eight patients (M:F = 28:10, age $53.4{\pm}13.0$ yrs) with upper gastrointestinal symptoms such as indigestion, gastric fullness or pain consecutively underwent C-14 UBT, GBx and C-13 UBT within one week before medications. For the C-14 UBT, a single breath sample was collected at 10 minutes after ingestion of C-14 urea (37 KBq) capsule and counting was done in a liquid scintillation counter for 1 minute, and the results were classified as positive (${\ge}$ 200 dpm), intermediate (50-199 dpm) or negative (50 dpm). For the C-13 UBT, the results were classified as positive (${\ge}2.5\%_{\circ}$) or negative ($<2.5\%_{\circ}$). The results of GBx with Giemsa stain were graded 0 (normal) to 4 (diffuse) according to the distribution of H. pylori by the Wyatt method. We compared C-14 UBT and C-13 UBT results with GBx grade as a gold standard. Results: The prevalence of H. pylori infection by GBx with Giemsa stain was 25/38 (65.8%). In the assessment of the presence of H. pylori infection, the C-14 UBT global performance yielded sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and accuracy of 92.0%, 92.3%, 95.8%, 91.7% and 92.1%, respectively. However, the C-13 UBT had sensitivity, specificity, PPV, NPV and accuracy of 96.0%, 84.6%, 92.3%, 91.7% and 92.1%, respectively. The more significant correlation in C-14 than C-13 UBT (r=0.948 vs r=0.819, p <0.001) was found between the value of UBT and the grade of distribution of H. pylori infection. Conclusion: We conclude that the diagnostic performance between C-14 and C-13 UBT to detect H. pylori infection is not significantly different, but the value of C-14 UBT more significantly reflects the degree of bacterial distribution.

• Analytical Science and Technology
• /
• v.24 no.5
• /
• pp.345-351
• /
• 2011

A liquid chromatography-electrospray ionization tandem mass spectrometry (LC-ESI/MS/MS) method was developed and validated for the determination of finasteride in human serum. Beclomethasone was used as internal standard (IS) and liquid-liquid extraction (LLE) using methyl tert-butyl ether (MTBE) was carried out to isolate analyte. The mass transitions monitored in multiple reaction monitoring (MRM) in positive ion mode were m/z 373.2${\rightarrow}$305.2 for finasteride and m/z 409.3${\rightarrow}$391.2 for IS. Retention times of finasteride and IS were 5.81 and 5.46 min, respectively. The limit of quantitation (LOQ) was 0.1 ng/mL and the calibration curve showed good linearity in the range of 0.1~20.0 ng/mL ($R^2$=0.9997). The intra-day assay precision and accuracy were in the range 6.3~10.6% and 97.3~103.6%, respectively, and the inter-day assay precision and accuracy were in the range 0.8~5.2% and 99.8~102.5%, respectively. The sample extract recovery of the method was 80~83%.

• Journal of Chest Surgery
• /
• v.40 no.5 s.274
• /
• pp.362-368
• /
• 2007

Background: Spontaneous pneumothorax patients with blebs or bullae are considered to be good candidates for operation, and various objective diagnostic modalities have been performed for detection of blebs and bullae. This study was performed to compare the efficacy of thoracoscopic examination with using a minimally invasive 2 mm thoracoscope with high-resolution computed tomography (HRCT) for treating primary spontaneous pneumothorax. Material and Method: From June 2001 to March 2002, 34 patients with spontaneous pneumothorax undewent study with 2 mm video-thoracoscopic examination and HRCT. We regarded a blob larger than 5 mm in diameter as significant. Standard thoracoscopic wedge resection was performed in 18 patients with significant blob via a 2 mm video-thoracoscopic examination. 1 patient incurred bleeding, and the remaining 15 patients were treated with pleural drainage. Result: Multiple or single blob lesions were detected by 2 mm video-thoracoscope in 52.9% (18/34) of the patients with primary pneumothorax. For a total of 19 patients who were operated on, the diagnostic accuracy of the 2 mm video-thoracoscopic examination for bullae and blob was 94.7% (18/19), which was superior to that of HRCT (73.7%, 14/19). At a mean follow-up of $30{\pm}3$ months, no recurrence occurred in both the operative group and the non-operative group. Conclusion: 2 mm video-thoracoscopic examination under local anesthesia has higher diagnostic accuracy than HRCT, and it is a useful alternative for determining the operative indications for spontaneous pneumothorax.