Background

Extraction of clinical information such as medications or problems from clinical text is an important task of clinical natural language processing (NLP). Rule-based methods are often used in clinical NLP systems because they are easy to adapt and customize. Recently, supervised machine learning methods have proven to be effective in clinical NLP as well. However, combining different classifiers to further improve the performance of clinical entity recognition systems has not been investigated extensively. Combining classifiers into an ensemble classifier presents both challenges and opportunities to improve performance in such NLP tasks.

Methods

We investigated ensemble classifiers that used different voting strategies to combine outputs from three individual classifiers: a rule-based system, a support vector machine (SVM) based system, and a conditional random field (CRF) based system. Three voting methods were proposed and evaluated using the annotated data sets from the 2009 i2b2 NLP challenge: simple majority, local SVM-based voting, and local CRF-based voting.

Results

Evaluation on 268 manually annotated discharge summaries from the i2b2 challenge showed that the local CRF-based voting method achieved the best F-score of 90.84% (94.11% Precision, 87.81% Recall) for 10-fold cross-validation. We then compared our systems with the first-ranked system in the challenge by using the same training and test sets. Our system based on majority voting achieved a better F-score of 89.65% (93.91% Precision, 85.76% Recall) than the previously reported F-score of 89.19% (93.78% Precision, 85.03% Recall) by the first-ranked system in the challenge.

Conclusions

Our experimental results using the 2009 i2b2 challenge datasets showed that ensemble classifiers that combine individual classifiers into a voting system could achieve better performance than a single classifier in recognizing medication information from clinical text. It suggests that simple strategies that can be easily implemented such as majority voting could have the potential to significantly improve clinical entity recognition.

Pyrazinoic acid, the active form of the first-line antituberculosis drug pyrazinamide, decreased the proton motive force and respiratory ATP synthesis rates in subcellular mycobacterial membrane assays. Pyrazinoic acid also significantly lowered cellular ATP levels in Mycobacterium bovis BCG. These results indicate that the predominant mechanism of killing by this drug may operate by depletion of cellular ATP reserves.

Purpose

To prospectively demonstrate the feasibility of using indocyanine green, a near-infrared (NIR) fluorophore at the minimum dose needed for noninvasive optical imaging of lymph nodes (LNs) in breast cancer patients undergoing sentinel lymph node mapping (SLNM).

Materials and Methods

Informed consent was obtained from 24 women (age range, 30–85 years) who received intradermal subcutaneous injections of 0.31–100 μg indocyanine green in the breast in this IRB-approved, HIPAA-compliant, dose escalation study to find the minimum microdose for imaging. The breast, axilla, and sternum were illuminated with NIR light and the fluorescence generated in the tissue was collected with an NIR-sensitive intensified charged-coupled device. Lymphoscintigraphy was also performed. Resected LNs were evaluated for the presence of radioactivity, blue dye accumulation, and fluorescence. The associations between the resected LNs that were fluorescent and (a) the time elapsed between NIR fluorophore administration and resection and (b) the dosage of NIR fluorophores were tested with the Spearman rank and Pearson product moment correlation tests, respectively.

Results

Lymph imaging consistently failed with indocyanine green microdosages between 0.31 and 0.77 μg. When indocyanine green dosages were 10 μg or higher, lymph drainage pathways from the injection site to LNs were imaged in eight of nine women; lymph propulsion was observed in seven of those eight. When propulsion in the breast and axilla regions was present, the mean apparent velocities ranged from 0.08 to 0.32 cm/sec, the time elapsed between “packets” of propelled fluid varied from 14 to 92 seconds. In patients who received 10 μg of indocyanine green or more, a weak negative correlation between the fluorescence status of resected LNs and the time between NIR fluorophore administration and LN resection was found. No statistical association was found between the fluorescence status of resected LNs and the dose of NIR fluorophore.

Conclusion

NIR fluorescence imaging of lymph function and LNs is feasible in humans at microdoses that would be needed for future molecular imaging of cancer-positive LNs.

During the past decade a considerable amount of data has been generated concerning polychlorinated dibenzodioxin (PCDD) and polychlorinated dibenzofuran (PCDF) levels in humans from many geographical locations. To organize these data in a useful fashion for environmental purposes and for consideration of human toxicity, selected portions of our data are presented in a somewhat atypical fashion, by percentage contribution of individual congeners to total PCDD/Fs in human tissue, and to the total dioxin equivalents (TEq). This is done to better characterize congener contributions from environmental contamination in various geographical regions at this time and health-related levels. To present the findings in a global perspective, data from widely different locations are presented including the United States, Germany, Vietnam, the former Soviet Union, Thailand, Cambodia, China, South Africa, and Guam.