Secular trends in the etiology of nosocomial infection at a teaching hospital in Taiwan, 1981-1994

Surveillance system of nosocomial infection was established in 1980 at the National Taiwan University Hospital (NTUH). To identify pathogens and the secular trends in the etiology of nosocomial infection from 1981 to 1994, the prospective, hospital-wide nosocomial surveillance data were analysed. During this period, 22,146 pathogens causing nosocomial infections were isolated. Gram-negative aerobic bacteria remained the major pathogens, but gram-positive cocci and fungi increased rapidly in the past 14 years. When the overall pathogen distribution is examined, Pseudomonas areuginosa was the most frequently isolated pathogen, but Candida albicans and other yeasts have taken the leading position since 1993. Staphylococcus aureus and coagulase-negative staphylococci also increase significantly in recent years. When the pathogens causing infection at the 4 major sites were examined. P. aeruginosa was the pathogen most often associated with respiratory tract and surgical wound infections. In blood stream and urinary tract infections, we observed Escherichia coli was replaced by C. albicans and other yeasts as a most common isolate in these years. In addition, C. albicans and other yeasts and methicillin-resistant S. aureus (MRSA) are emerging as major nosocomial pathogens at NTUH. C. albicans and other yeast increased from 1.8% in 1981 to 14.9% in 1994 in the overall nosocomial infection. The increase was found in the blood stream (2.1% to 16.2%) and urinary tract infections (5.4% to 24.7%). Of 1,742 nosocomial S. aureus isolates, the percentage of MRSA rose from 12.5% in 1981 to 55.2% in 1994. The high percentage of MRSA was observed at 4 major anatomic sites of infection. In summary, significant shifts in the pathogens of nosocomial infection have occurred in the past 14 years at NTUH, and the distribution of nosocomial pathogens was similar to those reported in the United States in recent years.     

In vitro activities of antimicrobial agents, alone and in combination, against Acinetobacter baumannii isolated from blood

In vitro activities of 15 antimicrobial agents against 90 strains of Acinetobacter baumannii isolated from blood cultures from hospitalized patients were determined using the agar dilution method. Imipenem, ofloxacin, and ciprofloxacin had the best antimicrobial activity with minimum inhibitory concentrations (MIC50s) of 0.25 mu g/ml and MIC90s of 0.5-1 mu g/ml. beta-lactam antibiotics other than imipenem had poor activity, with MIC50s ranging from 8 to 64 mu g/ml and MIC90s from 32 to > or = 256 mu g/ml. The checkerboard titration method was used to study the effects of combination of two antimicrobial agents. Combinations of ceftazidime, aztreonam, imipenem, or ciprofloxacin with amikacin showed either synergistic effects or partial synergistic effects for 40.9%-86.4% of 22 tested strains. The best in vitro activity was observed with the combination of imipenem and amikacin. No antagonistic effects were observed with the combination of imipenem and amikacin. Synergistic effects were confirmed by time-kill curve studies. In conclusion, imipenem, ofloxacin, and ciprofloxacin were the three most active agents against human blood isolates of A. baumannii. The combination of a beta-lactam or ciprofloxacin with amikacin was synergistic for some of the isolates.     

Accelerating Hair Rendering by Learning High-Order Scattered Radiance

Efficiently and accurately rendering hair accounting for multiple scattering is a challenging open problem. Path tracing in hair takes long to converge while other techniques are either too approximate while still being computationally expensive or make assumptions about the scene. We present a technique to infer the higher order scattering in hair in constant time within the path tracing framework, while achieving better computational efficiency. Our method makes no assumptions about the scene and provides control over the renderer's bias & speedup. We achieve this by training a small multilayer perceptron (MLP) to learn the higher-order radiance online, while rendering progresses. We describe how to robustly train this network and thoroughly analyze our resulting renderer's characteristics. We evaluate our method on various hairstyles and lighting conditions. We also compare our method against a recent learning based & a traditional real-time hair rendering method and demonstrate better quantitative & qualitative results. Our method achieves a significant improvement in speed with respect to path tracing, achieving a run-time reduction of 40%-70% while only introducing a small amount of bias.