• Nuclear Medicine and Molecular Imaging
• /
• v.43 no.5
• /
• pp.459-467
• /
• 2009

Purpose: The maximum likelihood-expectation maximization (ML-EM) is the statistical reconstruction algorithm derived from probabilistic model of the emission and detection processes. Although the ML-EM has many advantages in accuracy and utility, the use of the ML-EM is limited due to the computational burden of iterating processing on a CPU (central processing unit). In this study, we developed a parallel computing technique on GPU (graphic processing unit) for ML-EM algorithm. Materials and Methods: Using Geforce 9800 GTX+ graphic card and CUDA (compute unified device architecture) the projection and backprojection in ML-EM algorithm were parallelized by NVIDIA's technology. The time delay on computations for projection, errors between measured and estimated data and backprojection in an iteration were measured. Total time included the latency in data transmission between RAM and GPU memory. Results: The total computation time of the CPU- and GPU-based ML-EM with 32 iterations were 3.83 and 0.26 see, respectively. In this case, the computing speed was improved about 15 times on GPU. When the number of iterations increased into 1024, the CPU- and GPU-based computing took totally 18 min and 8 see, respectively. The improvement was about 135 times and was caused by delay on CPU-based computing after certain iterations. On the other hand, the GPU-based computation provided very small variation on time delay per iteration due to use of shared memory. Conclusion: The GPU-based parallel computation for ML-EM improved significantly the computing speed and stability. The developed GPU-based ML-EM algorithm could be easily modified for some other imaging geometries.

• Fire Science and Engineering
• /
• v.29 no.5
• /
• pp.104-109
• /
• 2015

The purpose of this study was to compare the ventilatory volume and airway pressure of a facial mask, endotracheal intubation, King tube, and I-gel devices with an Oxylator EM-100 using a RespiTrainer. The data were obtained from July 20 to 21, 2015. Data were analyzed using SPSS WIN 18.0 software. The ventilatory volume for endotracheal intubation was 537 ml (95% CI 530~545 ml), that for the King tube was 502 ml (95% CI 499~506 ml), that for the I-gel was 88 ml (95% CI 485~491 ml), and that for the facial mask was 499 ml (95% CI 496~503 ml). The airway pressure for endotracheal intubation was $11.34cmH_2O$ (95% CI $11.21{\sim}11.41cmH_2O$), that for the King tube was $10.67cmH_2O$ (95% CI $10.60{\sim}10.75cmH_2O$), that for the I-gel was $10.42cmH_2O$ (95% CI $10.35{\sim}10.67cmH_2O$), and that for the facial mask was $10.61cmH_2O$ (95% CI $10.55{\sim}10.68cmH_2O$). As a result, we were able to identify the appropriate ventilatory volume using the Oxylator EM-100.

• Fire Science and Engineering
• /
• v.31 no.2
• /
• pp.113-118
• /
• 2017

The purpose of this study was to compare the ventilatory volume and airway pressure using Oxylator EM-100, MicroVenT CSI-3000, OXY-LIFE II. The data were obtained from February 13 in 2017 andanalyzed using the SPSS WIN 18.0 program. The results of theventilatory volume showed the Oxylator EM-100 551.44 ml (${\pm}18.70$), MicroVenT CSI-3000 527.26 ml (${\pm}17.98$), and OXY-LIFE II 369.46 ml (${\pm}12.30$). The airway pressure showed the Oxylator EM-100 $11.89cmH_2O$ (${\pm}.41$), MicroVenT CSI-3000 $11.66cmH_2O$ (${\pm}.34$), and OXY-LIFE II $8.02cmH_2O$ (${\pm}.25$). This study will provide the basic data for an appropriate ventilation method by an oxygen supply device including Oxylator EM-100, MicroVenT CSI-3000, and OXY-LIFE II.

• Proceedings of the IEEK Conference
• /
• 1999.11a
• /
• pp.569-572
• /
• 1999

In this paper we present a color image segmentation algorithm based on statistical models. A novel deterministic annealing Expectation Maximization(EM) formula is derived to estimate the parameters of the Gaussian Mixture Model(GMM) which represents the multi-colored objects statistically. The experimental results show that the proposed deterministic annealing EM is a global optimal solution for the ML parameter estimation and the image field is segmented efficiently by using the parameter estimates.

• Proceedings of the Ginseng society Conference
• /
• 2002.10a
• /
• pp.317-334
• /
• 2002

Ginsenosides are metabolized (deglycosylated) by intestinal bacteria to active forms after oral administration. 20(S)-Protopanaxadiol $20-O-{\beta}-D-glucopyranoside$ (M1) and 20(S)-protopanaxatriol (M4) are the main intestinal bacterial metabolites (IBMs) of protopanaxadiol- and protopanaxatriol-type glycosides. M1 was selectively accumulated into the liver soon after its intravenous (i.v.) administration to mice, and mostly excreted as bile; however, some M1 was transformed to fatty acid ester (EMl) in the liver. EM1 was isolated from rats in a recovery dose of approximately $24mol\%.$ Structural analysis indicated that EM1 comprised a family of fatty acid mono-esters of M1. Because EM1 was not excreted as bile as Ml was, it was accumulated in the liver longer than M1. The in vitro cytotoxicity of M1 was attenuated by fatty acid esterification, implying that esterification is a detoxification reaction. However, esterified M1 (EM1) inhibited the growth of B16 melanoma more than Ml in vivo. The in vivo antitumor activity paralleled with the pharmacokinetic behavior. In the case of M4, orally administered M4 was absorbed from the small intestine into the mesenteric lymphatics followed by the rapid esterification of M4 with fatty acids and its spreading to other organs in the body and excretion as bile. The administration of M4 prior to tumor injection abrogated the enhanced lung metastasis in the mice pretreated with 2-chloroadenosine more effectively than in those pretreated with anti-asialo GMl. Both EM1 and EM4 did not directly affect tumor growth in vitro, whereas EM1 promoted tumor cell lysis by lymphocytes, particularly non-adherent splenocytes, and EM4 stimulated splenic NK cells to become cytotoxic to tumor cells. Thus, the esterification of IBM with fatty acids potentiated the antitumor activity of parental IBM through delay of the clearance and through immunostimulation. These results suggest that the fatty acid conjugates of IBMs may be the real active principles of ginsenosides in the body.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.54 no.3
• /
• pp.87-93
• /
• 2017

A maximum likelihood (ML) approach is presented for estimating the mean of radar cross section (RCS) of a Swerling III target and its numerical solution methods are discussed. The solution methods are based on an approximate expression for implementing the expectation maximization (EM) algorithm. The methods are evaluated and compared through Monte Carlo simulations in terms of estimation accuracy and computational efficiency to obtain a most efficient method for both Swerling I and Swerling III targets. The methods are also compared with a previously reported method based on heuristics.

• Communications for Statistical Applications and Methods
• /
• v.14 no.3
• /
• pp.517-530
• /
• 2007

We studied a modelling process for unimodal and multimodal circular data by using von Mises and its mixture distribution. In particular we suggested EM algorithm to find ML estimates of the mixture model. Simulation results showed the suggested methods are very accurate. Applications to two kinds of real data sets are also included.

• The Journal of the Korean Society for Microbiology
• /
• v.35 no.2
• /
• pp.171-180
• /
• 2000

Ninety two strains of Streptococcus pyogenes were isolated from patients with pharyngitis, scarlet fever, skin infection, and invasive streptococcal infections in Seoul, Korea from January to December, 1998. All isolates were epidemiologically characterized by T protein serotype, and serum opacity factor (OF) detection to phenotypes. To analyze the genetic relationship, fifty two isolates including 32 erythromycin-clindamycin (Em-Cm) resistant strains, 20 antimicrobial susceptible strains were attempted to the pulsed-field gel electrophoresis (PFGE). T protein serotype showed 16 kinds in distribution including T12 and T4. Among the total isolates, 40 strains (43.5%) belonged to the T12 serotype and twenty strains (21.7%) to T4 serotype. On the other hand, when infection aspect of S. pyogenes isolates were analysed by T serotype distribution, T12 type was predominant for pharyngitidis which contributed to 21 strains (53%) and for skin infection isolates which contributed to 11 strains (28%), respectively. In case of T4 type, it was the most predominant pharyngitidis isolates which contributed to 8 strains (40%). In T serotype distribution of Em-Cm resistant strains, 27 strains (84%) of the thirty two showed T12 serotype. In minimum inhibitory concentration (MIC) values of Em-Cm resistance isolates, thirty two isolates showed resistant to erythromycin 27 strains (84%), had high MIC of >$128\;{\mu}g/ml$. And also to clindamycin, twenty two strains (69%) had high MIC of >$128\;{\mu}g/ml$. When OF detection of Em-Cm resistance of S. pyogenes isolates were analyzed by T serotype distribution, T12 serotype isolates revealed that all of the isolates except one strain were OF negative. In PFGE profile analysis to Em-Cm resistance isolates, of the twenty seven, Em-Cm resistance of T12 serotype isolates, 26 strains showed identical PFGE profile and all of these isolates revealed that OF negative. Eighty four percent of Em-Cm resistance S. pyogenes isolates had identical phenotype and PFGE profile. These results strongly suggested that the Em-Cm resistant S. pyogenes isolates from Seoul area showed close genetic correlation and PFGE could be available tool for molecular epidemiology.

• Proceedings of the KOSOMBE Conference
• /
• v.1997 no.11
• /
• pp.361-364
• /
• 1997

Bayesian MAP-EM approaches have been quite useful or tomographic reconstruction in that they can stabilize the instability of well-known ML-EM approaches, and can incorporate a priori information on the underlying emission object. However, MAP reconstruction algorithms with expressive priors often suffer from the optimization problem when their objective unctions are nonquadratic. In our previous work [1], we showed that the use of deterministic annealing method greatly reduces computational burden or optimization and provides a good solution or nonquadratic objective unctions. Here, we further investigate the convergence of the deterministic annealing algorithm; our experimental results show that, while the solutions obtained by a simple quenching algorithm depend on the initial conditions, the estimates converged via deterministic annealing algorithm are consistent under various initial conditions.

• The Journal of the Acoustical Society of Korea
• /
• v.24 no.3
• /
• pp.141-149
• /
• 2005

Tn general. Gaussian mixture model (GMM) is used to estimate the speaker model from the speech for speaker identification. The parameter estimates of the GMM are obtained by using the Expectation-Maximization (EM) algorithm for the maximum likelihood (ML) estimation. However the EM algorithm has such drawbacks that it depends heavily on the initialization and it needs the number of mixtures to be known. In this paper, to solve the above problems of the EM algorithm. we propose an EM algorithm with the initialization based on incremental ${\cal}k-means$ for GMM. The proposed method dynamically increases the number of mixtures one by one until finding the optimum number of mixtures. Whenever adding one mixture, we calculate the mutual relationship between it and one of other mixtures respectively. Finally. based on these mutual relationships. we can estimate the optimal number of mixtures which are statistically independent. The effectiveness of the proposed method is shown by the experiment for artificial data. Also. we performed the speaker identification by applying the proposed method comparing with other approaches.