• Journal of Intelligence and Information Systems
• /
• v.23 no.2
• /
• pp.71-88
• /
• 2017

Language models were originally developed for speech recognition and language processing. Using a set of example sentences, a language model predicts the next word or character based on sequential input data. N-gram models have been widely used but this model cannot model the correlation between the input units efficiently since it is a probabilistic model which are based on the frequency of each unit in the training set. Recently, as the deep learning algorithm has been developed, a recurrent neural network (RNN) model and a long short-term memory (LSTM) model have been widely used for the neural language model (Ahn, 2016; Kim et al., 2016; Lee et al., 2016). These models can reflect dependency between the objects that are entered sequentially into the model (Gers and Schmidhuber, 2001; Mikolov et al., 2010; Sundermeyer et al., 2012). In order to learning the neural language model, texts need to be decomposed into words or morphemes. Since, however, a training set of sentences includes a huge number of words or morphemes in general, the size of dictionary is very large and so it increases model complexity. In addition, word-level or morpheme-level models are able to generate vocabularies only which are contained in the training set. Furthermore, with highly morphological languages such as Turkish, Hungarian, Russian, Finnish or Korean, morpheme analyzers have more chance to cause errors in decomposition process (Lankinen et al., 2016). Therefore, this paper proposes a phoneme-level language model for Korean language based on LSTM models. A phoneme such as a vowel or a consonant is the smallest unit that comprises Korean texts. We construct the language model using three or four LSTM layers. Each model was trained using Stochastic Gradient Algorithm and more advanced optimization algorithms such as Adagrad, RMSprop, Adadelta, Adam, Adamax, and Nadam. Simulation study was done with Old Testament texts using a deep learning package Keras based the Theano. After pre-processing the texts, the dataset included 74 of unique characters including vowels, consonants, and punctuation marks. Then we constructed an input vector with 20 consecutive characters and an output with a following 21st character. Finally, total 1,023,411 sets of input-output vectors were included in the dataset and we divided them into training, validation, testsets with proportion 70:15:15. All the simulation were conducted on a system equipped with an Intel Xeon CPU (16 cores) and a NVIDIA GeForce GTX 1080 GPU. We compared the loss function evaluated for the validation set, the perplexity evaluated for the test set, and the time to be taken for training each model. As a result, all the optimization algorithms but the stochastic gradient algorithm showed similar validation loss and perplexity, which are clearly superior to those of the stochastic gradient algorithm. The stochastic gradient algorithm took the longest time to be trained for both 3- and 4-LSTM models. On average, the 4-LSTM layer model took 69% longer training time than the 3-LSTM layer model. However, the validation loss and perplexity were not improved significantly or became even worse for specific conditions. On the other hand, when comparing the automatically generated sentences, the 4-LSTM layer model tended to generate the sentences which are closer to the natural language than the 3-LSTM model. Although there were slight differences in the completeness of the generated sentences between the models, the sentence generation performance was quite satisfactory in any simulation conditions: they generated only legitimate Korean letters and the use of postposition and the conjugation of verbs were almost perfect in the sense of grammar. The results of this study are expected to be widely used for the processing of Korean language in the field of language processing and speech recognition, which are the basis of artificial intelligence systems.

• Journal of dental hygiene science
• /
• v.5 no.2
• /
• pp.51-56
• /
• 2005

This study investigated the medicines prescribed in dental services in order to provide information and materials on dental clinics and hospitals, for the intention of leading the dental circle to make efforts for voluntary improvement, do adequate prescription, and attempt for bench marking through this unfolding of their tendency in continuous prescription behaviors, and leading the people to have a right recognition on adequate medicine use through this information on their medicine-taking behaviors including antibiotics and injections. From the records of outpatient prescriptions of medicines under health insurance over the period of June 1st to September 31th in 2003 in the 34,226 recuperation institutions, antibiotics and injections were analyzed into administration days, prescription frequency, medicine cost per administration day, the number of medicines per prescription, and the number weight of high priced medicines. The findings were as below: 1. Adminstration days of antibiotics was 90.11% in the dental clinics, which was a decrease than the same quarter and the previous quarter of the previous year. While the prescription frequency of antibiotics was 15.5%, higher than the same quarter and the previous quarter of the previous year. In dental hospitals, administration days and prescription frequency of antibiotics were 71.57% and 21.05%, respectively, a little higher than the previous quarter. Compared to other kind of recuperation institutions, dental clinics and hospitals had higher administration days and lower prescription frequency. 2. For injections, adminstration days and prescription frequency in dental clinics were 0.13% and 0.05%, respectively, which were decreases than the same quarter of the previous year. In dental hospitals, adminstration days and prescription frequency were 1.03% and 0.88%, respectively, a little lower than those of the previous quarter. Compared to other kind of recuperation institutions, dental clinics and hospitals were very lower in injection administration days and injection prescription frequency. 3. The number of prescribed medicines was 2.79 in the dental clinics, which was lower than the same quarter of the previous year but higher than the previous quarter. Dental hospitals put 2.67 numbers of medicines per prescription, an increase than the same quarter of the previous year and the previous quarter. Compared to other kind of recuperation institutions, dental clinics and hospitals put smaller number of medicines per prescription. 4. Medicine cost per administration day was 863 won in the dental clinics, which was an increase than the same quarter of the previous year and in the previous quarter. Compared to other kind of recuperation institutions, dental clinics and hospitals had lower medicine cost per administration day. 5. The number weight of high priced medicines was 46.43% in dental clinics, which was an increase than the previous quarter. In dental hospitals, it was 54.05%, so remarkable an increase than the previous quarter. Compared to other kind of hospitals and clinics, dental clinics and hospitals prescribed larger number of high priced medicines. 6. By districts, the frequency of antibiotics prescriptions was the highest in Kwanju and the lowest in Daejeon. The frequency of injection prescriptions was high in all Youngnam districts as was in the second quarter, while low in all the Metropolitan districts. There was a large variation in the prescription frequency to the districts, as the district of the highest prescription frequency had more than 2 times larger frequency than the district of the lowest frequency. Medicine cost per administration day was the highest in Ulsan but the lowest in the north part of Cholla province. The number of medicines per prescription was the largest in Kyonggi province while the smallest in Cheju-do.

• KOREAN JOURNAL OF CROP SCIENCE
• /
• v.39 no.1
• /
• pp.27-37
• /
• 1994

To catch the relative importance of varietal and environmental variation in various grain quality components associated with palatability of cooked rice, grain appearance, milling recovery, several physicochemical properties of milled rice and texture or eating quality of cooked rice for rice materials of five japonica cultivars, produced at four locations of the mid-mountainous and alpine area of Korea in 1989, were evaluated and analyzed the obtained data. Highly significant varietal and locational variations were detected in 1000-grain weight, amylose content, K/Mg ratio, gelatinization temperature, peak viscosity, breakdown and setback viscosities as compared with variety x location interaction variation. Also, marked locational variations were recongnized in milling recovery from rough to brwon rice, alkali digestibility and protein content, and significant varietal variation was caught in stickiness /hardness ratio of cooked rice. The variety x location interaction variation was especially large in quality components of grain appearance and ripening, palatability of cooked rice and consistency viscosity. One thousand kernel weight was heaviest in Jinbuolbyeo and Odaebyeo, and the unfilled grain ratio was lowest in Jinbuolbyeo. Odaebyeo showed slightly' lower ratio of intact and clear milled rice because of more chalky rice kernels compared with other cultivars. Amylose content of Jinbuolbyeo and Sobaegbyeo was about 1% lower than that of others and K/Mg ratio of Odaebyeo was the lowest one among rice materials. Odaebyeo, Sobaegbyeo and Jinbuolbyeo revealed significantly low gelatinization temperature and setback viscosity while high peak and breakdown viscosities. Cholwon rice showed the greatest kernel weight, good grain filling but lowest ratio of intact and clear milled rice while Jinbu rices exhibited the highest milling recovery from rough to brown rice and ratio of sound milled rice. Amylose content of milled rice in Jinbu rices was about 2-3% lower than those in other locations. Protein content of polished rice was about 1% lower in rice materials of middle zone than those of southern part of Korea. K/Mg ratio of milled rice was highest in Jinbu rice and potassium content was slightly higher in the rice materials of middle region than in those of southern region. Alkali digestion value and gelatinization temperature of polished rice was markedly high in Jinbu rices as compared with other locations. Breakdown viscosity was hightest in Chlown rices and next higher with the order of Hwaso>Unbong>Jinbu rices, and setback viscosity was the quite contrary tendency with breakdown. The stickiness /hardness ratio of cooked rice was relatively higher value in Cholwon rices than in the others and the palatability of cooked rice was a little better in Unbong and Cholwon rices than in Jinbu and Hwaso rices, although variety x location interaction variation was large. The rice materials can be classified largely into two groups of Jinbu and the others by the distribution on the plane of 1st and 2nd principal components (about 60% of total informations) contracted from twelve grain quality properties closely associated with eating quality of cooked rice. Also, Jinbu and the other rices were divided into two and three rice groups respectively. Varietal variation of overall rice quality was smallest in Hwaso. The most superior rice group in overall quality evaluation included Odaebyeo produced at Cholwon, Unbong and Hwaso, and Sobaegbyeo grown at Unbong

• Journal of the Korean Society of Food Culture
• /
• v.14 no.4
• /
• pp.371-383
• /
• 1999

This report gave analysis of food demand both in Korea and Japan through introducing the concept of cohort analysis to the conventional demand model. This research was done to clarify the factors which determine food demand of the household. The traits of the new model for demand analysis are to consider and quantify those effects on food demand not only of economic factors such as expenditure and price but also of non-economic factors such as the age and birth cohort of the householder. The results of the analysis can be summarized as follows: 1) The comparison of the item-wise elasticities of food demand demonstrates that the expenditure elasticity is higher in Korea than in Japan and that the expenditure elasticity is -0.1 for cereal and more than 1 for eating-out in both countries. In respect to price elasticity, the absolute values of all the items except alcohol and cooked food are higher in the Korea than in Japan, and especially the price elasticities of beverages, dairy products and fruit are predominantly higher in Japan. In this way, both expenditure and price elasticities of a large number of items are higher in Korea than in Japan, which may be explained from the fact that the level of expenditure is higher in Japan than in Korea. 2) In both of Korea and Japan, as the householder grows older, the expenditure for each item increases and the composition of expenditure changes in such a way that these moves may be regarded as due to the age effect. However, there are both similarities and differences in the details of such moves between Korea and Japan. Those two countries have this trait in common that the young age groups of the householder spend more on dairy products and middle age groups spend more on cake than other age groups. In the Korea, however, there can be seen a certain trend that higher age groups spend more on a large number of items, reflecting the fact that there are more two-generation families in higher age groups. Japan differs from Korea in that expenditure in Japan is diversified, depending upon the age group. For example, in Japan, middle age groups spend more on cake, cereal, high-caloric food like meat and eating-out while older age groups spend more for Japanese-style food like fish/shellfish and vegetable/seaweed, and cooked food. 3) The effect of the birth cohort effect was also demonstrated. The birth cohort effect was introduced under the supposition that the food circumstances under which the householder was born and brought up would determine the current expenditure. Thus, the following was made clear: older generations in both countries placed more emphasis upon stable food in their composition of food consumption; the share of livestock products, oil/fats and externalized food was higher in the food composition of younger generation; differences in food composition among generations were extremely large in Korea while they were relatively small in Japan; and Westernization and externalization of diet made rapid increases simultaneously with generation changes in Korea while they made any gradual increases in Japan during the same time period. 4) The four major factors which impact the long-term change of food demand of the household are expenditure, price, the age of the householder, and the birth cohort of the householder. Investigations were made as to which factor had the largest impact. As a result, it was found that the price effect was the smallest in both countries, and that the relative importance of the factor-by-factor effects differed among the two countries: in Korea the expenditure effect was greater than the effects of age and birth cohort while in Japan the effects of non-economic factors such as the age and birth cohort of householder were greater than those of economic factors such as expenditures.

• The Journal of Korean Society for Radiation Therapy
• /
• v.27 no.2
• /
• pp.131-143
• /
• 2015

Purpose : By comparing a CT fusion plan based on MRI with a plan based on only MRI without CT, we intended to study usefulness of a plan based on only MRI. And furthermore, we intended to realize a realtime MR-IGRT by MRI image without CT scan during the course of simulation, treatment planning, and radiation treatment. Materials and Methods : BBB CT (Brilliance Big Bore CT, 16slice, Philips), Viewray MRIdian system (Viewray, USA) were used for CT & MR simulation and Treatment plan of 11 patients (1 Head and Neck, 5 Breast, 1 Lung, 3 Liver, 1 Prostate). When scanning for treatment, Free Breathing was enacted for Head&Neck, Breast, Prostate and Inhalation Breathing Holding for Lung and Liver. Considering the difference of size between CT and Viewray, the patient's position and devices were in the same condition. Using Viewray MRIdian system, two treatment plans were established. The one was CT fusion treatment plan based on MR image. Another was MR treatment plan including electron density that [ICRU 46] recommend for Lung, Air and Bone. For Head&Neck, Breast and Prostate, IMRT was established and for Lung and Liver, Gating treatment plan was established. PTV's Homogeneity Index(HI) and Conformity Index(CI) were use to estimate the treatment plan. And DVH and dose difference of each PTV and OAR were compared to estimate the treatment plan. Results : Between the two treatment plan, each difference of PTV's HI value is 0.089% (Head&Neck), 0.26% (Breast), 0.67% (Lung), 0.2% (Liver), 0.4% (Prostate) and in case of CI, 0.043% (Head&Neck), 0.84% (Breast), 0.68% (Lung), 0.46% (Liver), 0.3% (Prostate). As showed above, it is on Head&Neck that HI and CI's difference value is smallest. Each difference of average dose on PTV is 0.07 Gy (Head&Neck), 0.29 Gy (Breast), 0.18 Gy (Lung), 0.3 Gy (Liver), 0.18 Gy (Prostate). And by percentage, it is 0.06% (Head&Neck), 0.7% (Breast), 0.29% (Lung), 0.69% (Liver), 0.44% (Prostate). Likewise, All is under 1%. In Head&Neck, average dose difference of each OAR is 0.01~0.12 Gy, 0.04~0.06 Gy in Breast, 0.01~0.21 Gy in Lung, 0.06~0.27 Gy in Liver and 0.02~0.23 Gy in Prostate. Conclusion : PTV's HI, CI dose difference on the Treatment plan using MR image is under 1% and OAR's dose difference is maximum 0.89 Gy as heterogeneous tissue increases when comparing with that fused CT image. Besides, It characterizes excellent contrast in soft tissue. So, radiation therapy using only MR image without CT scan is useful in the part like Head&Neck, partial breast and prostate cancer which has a little difference of heterogeneity.

• Korean Journal of Agricultural Science
• /
• v.4 no.2
• /
• pp.215-228
• /
• 1977

In order to provide good germplasm for developing good corn hybrids, a total of 948 Korean indigenous corn lines were collected from various parts of country and major morphological characterstics of ears were investigated. The results obtained were as follows; 1) Ear Type; From the east-north mountaionus region where more than 80% of total corn production is practiced, cylinder (type I) or similar types to cylinder corn were collected, and from the southern plain region, where rather small scales of corn is grown, cone type (type IV) or similar types to cone were prevalent. 2) Kernel color; In the ear colors of all the indigenous corn lines collected from ten regions, ears with mono color were 54.4%, ears with two color mixed were 39.0% and ears with three or more color mixed were 6.6%. In northern mountainous region, region A and region I, ear color was mostly white or white plus other colors, while in other regions ear color was yellow or yellow plus other colors. 3) Denting; Dent type was only 4.3% of Korean indigenous corn lines collected, and others were flint type. Dent type was collected from northern regions, where foreign corn varieties were introduced and grown. 4) Ear row number; Ear row numbers of indigenous corn lines collected were 12 to 16. There was no significant differences among the ear row numbers in a ear ciassified by regions. However, it was observed that ear row number was closely related to kernel size. For instance, the ears with 24 ear-rows were the smallest in kernel size. 5) Quality of starch; 70.9% of the indigenous corn lines collected were kernels with hard starch. Corn with soft starch was 26.0% and medium type was 3.1%. In region A and region I, where lot of corn is grown, corn with hard starch was more frequently collected. 6) Pop corn and waxy corn; In all the indigenous corn lines collected, popcorn was distributed uniformly through the regions except region I, and waxy corn was found more in the northern mountainous region. 7) Ear length; The mean ear length of indigenous corn lines collected was 13cm. In region A and region I ear length was larger than that in other regions. 8) Ear diameter; The mean ear diameter of indigenous corn lines collected was 3.3cm. In region A and region I ear diameter was larger than that in other regions. 9) Kernel length, kernel width and kernel thickness; The mean kernel length, kernel width and kernel thickness of indigenous corn lines collected were 0.82cm, 0.42cm, and 0.78cm, respectively. The kernel size in the region A and region I was larger than that in other regions. 10) Ear weight; The mean ear weight of indigenous corn lines collected was 58.04gr. Ear weight was remarkably heavier in region A and region I. The heaviest ear weighed 330gr, and the lightest ear weighed 5 gr. 11) Kernel weight of a ear and 100 kernel weight; Kernel weight of a ear and 100 kernel of indigenous corn lines collected were 47.07gr and 15.07gr, respectively. Kernel weights and 100 kernel weights were much heavier in region A and region I than other regions.

• Journal of Intelligence and Information Systems
• /
• v.24 no.2
• /
• pp.59-83
• /
• 2018

With the increasing importance of sentiment analysis to grasp the needs of customers and the public, various types of deep learning models have been actively applied to English texts. In the sentiment analysis of English texts by deep learning, natural language sentences included in training and test datasets are usually converted into sequences of word vectors before being entered into the deep learning models. In this case, word vectors generally refer to vector representations of words obtained through splitting a sentence by space characters. There are several ways to derive word vectors, one of which is Word2Vec used for producing the 300 dimensional Google word vectors from about 100 billion words of Google News data. They have been widely used in the studies of sentiment analysis of reviews from various fields such as restaurants, movies, laptops, cameras, etc. Unlike English, morpheme plays an essential role in sentiment analysis and sentence structure analysis in Korean, which is a typical agglutinative language with developed postpositions and endings. A morpheme can be defined as the smallest meaningful unit of a language, and a word consists of one or more morphemes. For example, for a word '예쁘고', the morphemes are '예쁘(= adjective)' and '고(=connective ending)'. Reflecting the significance of Korean morphemes, it seems reasonable to adopt the morphemes as a basic unit in Korean sentiment analysis. Therefore, in this study, we use 'morpheme vector' as an input to a deep learning model rather than 'word vector' which is mainly used in English text. The morpheme vector refers to a vector representation for the morpheme and can be derived by applying an existent word vector derivation mechanism to the sentences divided into constituent morphemes. By the way, here come some questions as follows. What is the desirable range of POS(Part-Of-Speech) tags when deriving morpheme vectors for improving the classification accuracy of a deep learning model? Is it proper to apply a typical word vector model which primarily relies on the form of words to Korean with a high homonym ratio? Will the text preprocessing such as correcting spelling or spacing errors affect the classification accuracy, especially when drawing morpheme vectors from Korean product reviews with a lot of grammatical mistakes and variations? We seek to find empirical answers to these fundamental issues, which may be encountered first when applying various deep learning models to Korean texts. As a starting point, we summarized these issues as three central research questions as follows. First, which is better effective, to use morpheme vectors from grammatically correct texts of other domain than the analysis target, or to use morpheme vectors from considerably ungrammatical texts of the same domain, as the initial input of a deep learning model? Second, what is an appropriate morpheme vector derivation method for Korean regarding the range of POS tags, homonym, text preprocessing, minimum frequency? Third, can we get a satisfactory level of classification accuracy when applying deep learning to Korean sentiment analysis? As an approach to these research questions, we generate various types of morpheme vectors reflecting the research questions and then compare the classification accuracy through a non-static CNN(Convolutional Neural Network) model taking in the morpheme vectors. As for training and test datasets, Naver Shopping's 17,260 cosmetics product reviews are used. To derive morpheme vectors, we use data from the same domain as the target one and data from other domain; Naver shopping's about 2 million cosmetics product reviews and 520,000 Naver News data arguably corresponding to Google's News data. The six primary sets of morpheme vectors constructed in this study differ in terms of the following three criteria. First, they come from two types of data source; Naver news of high grammatical correctness and Naver shopping's cosmetics product reviews of low grammatical correctness. Second, they are distinguished in the degree of data preprocessing, namely, only splitting sentences or up to additional spelling and spacing corrections after sentence separation. Third, they vary concerning the form of input fed into a word vector model; whether the morphemes themselves are entered into a word vector model or with their POS tags attached. The morpheme vectors further vary depending on the consideration range of POS tags, the minimum frequency of morphemes included, and the random initialization range. All morpheme vectors are derived through CBOW(Continuous Bag-Of-Words) model with the context window 5 and the vector dimension 300. It seems that utilizing the same domain text even with a lower degree of grammatical correctness, performing spelling and spacing corrections as well as sentence splitting, and incorporating morphemes of any POS tags including incomprehensible category lead to the better classification accuracy. The POS tag attachment, which is devised for the high proportion of homonyms in Korean, and the minimum frequency standard for the morpheme to be included seem not to have any definite influence on the classification accuracy.

• Korean Journal of Ecology and Environment
• /
• v.36 no.1 s.102
• /
• pp.21-28
• /
• 2003

Community composition and functional feeding groups of aquatic insects according to stream order were investigated from the Gapyeong Creek, a typical mid-sized Korean stream in Gyeonggi -do, Korea, in April 2000. For field investigations, the main watercourse and three major tributaries of the stream that belong to stream order ll to Vll were divided into reaches (ca. $1{\sim}4$ km in distance). Aquatic insects were sampled from one or two sites each reach (total 30 sites) using a Surber sampler ($50{\times}50$cm, mesh 0.75 mm). As a result of the quantitative samplings(two Surber samplings at riffle and pool/run per site; total $2{\times}30$=60 Surber samplings) and additional qualitative samplings, a total of 164 species of aquatic insects in 103 genera, 54 families, and 8 orders were collected from the stream. Those aquatic insects were composed of Trichoptera (56spp.: 34.1%), Ephemeroptera (43 spp.: 26.2%), Diptera (25spp.: 15.2%), Plecoptera (23 spp.: 14.0%), Coleoptera (6 spp.: 3.7%), Odonata (6 spp.:3.7%), Hemiptera (3 spp.: 1.8%), and Megaloptera (2 spp.: 1.2%); EPT-group (122 spp.:74.4%) or EPT-group plus Diptera (147 spp.: 89.6%) occupied most of the aquatic insect community; relatively larger number of species occurred in the mid-stream reaches (order III-Vl). The quantitative samplings throughout the study sites yielded a total of 26,286 individuals of aquatic insects ($136{\sim}2522$ inds./0.5 $m^2$, mean 906.4inds./0.5 $m^2$) that belongs to Ephemeroptera (11,994 inds.: 45.6%), Diptera (8730 inds.:33.2%), Trichoptera (4123 inds.: 15.7%), Plecoptera (1213 inds.: 4.6%), Coleoptera (204 inds.: 0.8%), Odonata (13 inds.: 0.05%), Megaloptera (5 inds.: 0.02%), and Hemiptera (4inds.: inds.: 0.02%); average number of individuals of aquatic insects increased as the stream order increased: average numbers of individuals of Bllecoptera and Trichoptera decreased and increased, respectively, as the stream order increased. Tolerant species such as Chironomidae spp., Uracanthella rufa and Hydropsychidae spp. were particularly abundant in the down stream reaches (order Vll) . Species diversity indices (H`) and dominance indices (Dl) were relatively higher and lower, respectively, in the mid-stream reaches (order $IV{\sim}VI$). Shredders occupied the smallest partand collector-gatherers were most abundant among the functional feeding groups(FFGs); collector-filterers considerably increased in the down stream reaches (orders Vl and VII); scrappers were relatively evenly distributed throughout the stream reaches: predators were relatively more abundant in the uppermost stream reaches(order ll) . Overall, the characteristics of aquatic insect comminity and FFGs in the Gapyeong Creek are largely similar to those in the normal streams of temperate deciduous forest in the northern hemisphere that is explained by the river confineum concept.

• Progress in Medical Physics
• /
• v.22 no.1
• /
• pp.42-51
• /
• 2011

Nuclear medicine images (SPECT, PET) were widely used tool for assessment of myocardial viability and perfusion. However it had difficult to define accurate myocardial infarct region. The purpose of this study was to investigate methodological approach for automatic measurement of rat myocardial infarct size using polar map with adaptive threshold. Rat myocardial infarction model was induced by ligation of the left circumflex artery. PET images were obtained after intravenous injection of 37 MBq $^{18}F$-FDG. After 60 min uptake, each animal was scanned for 20 min with ECG gating. PET data were reconstructed using ordered subset expectation maximization (OSEM) 2D. To automatically make the myocardial contour and generate polar map, we used QGS software (Cedars-Sinai Medical Center). The reference infarct size was defined by infarction area percentage of the total left myocardium using TTC staining. We used three threshold methods (predefined threshold, Otsu and Multi Gaussian mixture model; MGMM). Predefined threshold method was commonly used in other studies. We applied threshold value form 10% to 90% in step of 10%. Otsu algorithm calculated threshold with the maximum between class variance. MGMM method estimated the distribution of image intensity using multiple Gaussian mixture models (MGMM2, ${\cdots}$ MGMM5) and calculated adaptive threshold. The infarct size in polar map was calculated as the percentage of lower threshold area in polar map from the total polar map area. The measured infarct size using different threshold methods was evaluated by comparison with reference infarct size. The mean difference between with polar map defect size by predefined thresholds (20%, 30%, and 40%) and reference infarct size were $7.04{\pm}3.44%$, $3.87{\pm}2.09%$ and $2.15{\pm}2.07%$, respectively. Otsu verse reference infarct size was $3.56{\pm}4.16%$. MGMM methods verse reference infarct size was $2.29{\pm}1.94%$. The predefined threshold (30%) showed the smallest mean difference with reference infarct size. However, MGMM was more accurate than predefined threshold in under 10% reference infarct size case (MGMM: 0.006%, predefined threshold: 0.59%). In this study, we was to evaluate myocardial infarct size in polar map using multiple Gaussian mixture model. MGMM method was provide adaptive threshold in each subject and will be a useful for automatic measurement of infarct size.

• Korean Journal of Weed Science
• /
• v.9 no.1
• /
• pp.46-55
• /
• 1989

Stomatal variation was observed at the Yeongnam Crop Experiment Station in 1988 using 42 rice cultivars and 30 weed species. The shape, density or size of stomata was varied depending on the species. Two general trends, however, were found that more number of stomata was found at lower leaf epidermis than upper leaf epidermis and stomata number was negatively correlated with stomata size. Aneilema japonica and Portulaca oleracea had the least number of stomata having 17-20 stomata per $m^2$ for upper leaf epidermis and 17-54 stomata for lower leaf epidermis while Polygonum conspicuum had the greatest number of stomata (449 for upper leaf epidermis and 511 for lower leaf epidermis). Soybean, Aeschynomene indica, Ludwigia prostrata and Lactuca indica had the smallest in stomata size while the biggest stomata was found at P. oleracea and A. Japonica that had the least number of stomata. Cyperus species such as C. difformis, C. iria and C. serotinus had no stomata at upper leaf epidermis. The stomata were distributed only at lower leaf epidermis for these species. Potamogeton distinctus, on the other hand, had stomata almost at upper leaf epidermis and thus, hardly found the stomata at lower leaf epidermis. Among rice cultivars, Tongil-type had the greatest number of stomata followed by Indica-type and Japonica-type, in order. Cultivars released after 1960 had more stomata than cultivars released before 1960 for Japonica-type cultivars while stomata size had reversed trend. Jinheung had the least number of stomata (${\fallingdotseq}$ 150 per $mm^2$) while Yushin had the greatest number of stomata (350 for upper and 449 for lower leaf epidermis, respectively) among rice cultivars. Other cultivars having more than 350 stomata per $mm^2$ were Samgangbyeo, Milyang 23, Woonbongbyeo, etc.