A quantitative comparison of towed-camera and diver-camera transects for monitoring coral reefs.

Novel tools and methods for monitoring marine environments can improve efficiency but must not compromise long-term data records. Quantitative comparisons between new and existing methods are therefore required to assess their compatibility for monitoring. Monitoring of shallow water coral reefs is typically conducted using diver-based collection of benthic images along transects. Diverless systems for obtaining underwater images (e.g. towed-cameras, remotely operated vehicles, autonomous underwater vehicles) are increasingly used for mapping coral reefs. Of these imaging platforms, towed-cameras offer a practical, low cost and efficient method for surveys but their utility for repeated measures in monitoring studies has not been tested. We quantitatively compare a towed-camera approach to repeated surveys of shallow water coral reef benthic assemblages on fixed transects, relative to benchmark data from diver photo-transects. Differences in the percent cover detected by the two methods was partly explained by differences in the morphology of benthic groups. The reef habitat and physical descriptors of the site-slope, depth and structural complexity-also influenced the comparability of data, with differences between the tow-camera and the diver data increasing with structural complexity and slope. Differences between the methods decreased when a greater number of images were collected per tow-camera transect. We attribute lower image quality (variable perspective, exposure and focal distance) and lower spatial accuracy and precision of the towed-camera transects as the key reasons for differences in the data from the two methods and suggest changes to the sampling design to improve the application of tow-cameras to monitoring.

Biophysical and anthropogenic influences on the status of Tonga's coral reefs and reef fish fishery.

Despite increasing threats to Tonga's coral reefs from stressors that are both local (e.g. overfishing and pollution) and global (e.g. climate change), there is yet to be a systematic assessment of the status of the country's coral reef ecosystem and reef fish fishery stocks. Here, we provide a national ecological assessment of Tonga's coral reefs and reef fish fishery using ecological survey data from 375 sites throughout Tonga's three main island groups (Ha'apai, Tongatapu and Vava'u), represented by seven key metrics of reef health and fish resource status. Boosted regression tree analysis was used to assess and describe the relative importance of 11 socio-environmental variables associated with these key metrics of reef condition. Mean live coral cover across Tonga was 18%, and showed a strong increase from north to south correlated with declining sea surface temperature, as well as with increasing distance from each provincial capital. Tongatapu, the southernmost island group, had 2.5 times greater coral cover than the northernmost group, Vava'u (24.9% and 10.4% respectively). Reef fish species richness and density were comparable throughout Tongatapu and the middle island group, Ha'apai (~35 species/transect and ~2500 fish/km2), but were significantly lower in Vava'u (~24 species/transect and ~1700 fish/km2). Spatial patterns in the reef fish assemblage were primarily influenced by habitat-associated variables (slope, structural complexity, and hard coral cover). The biomass of target reef fish was greatest in Ha'apai (~820 kg/ha) and lowest in Vava'u (~340 kg/ha), and was negatively associated with higher human influence and fishing activity. Overall mean reef fish biomass values suggest that Tonga's reef fish fishery can be classified as moderately to heavily exploited, with 64% of sites having less than 500 kg/ha. This study provides critical baseline ecological information for Tonga's coral reefs that will: (1) facilitate ongoing management and research; and (2) enable accurate reporting on conservation targets locally and internationally.

The UAB Informatics Institute and 2016 CEGS N-GRID de-identification shared task challenge.

Clinical narratives (the text notes found in patients' medical records) are important information sources for secondary use in research. However, in order to protect patient privacy, they must be de-identified prior to use. Manual de-identification is considered to be the gold standard approach but is tedious, expensive, slow, and impractical for use with large-scale clinical data. Automated or semi-automated de-identification using computer algorithms is a potentially promising alternative. The Informatics Institute of the University of Alabama at Birmingham is applying de-identification to clinical data drawn from the UAB hospital's electronic medical records system before releasing them for research. We participated in a shared task challenge by the Centers of Excellence in Genomic Science (CEGS) Neuropsychiatric Genome-Scale and RDoC Individualized Domains (N-GRID) at the de-identification regular track to gain experience developing our own automatic de-identification tool. We focused on the popular and successful methods from previous challenges: rule-based, dictionary-matching, and machine-learning approaches. We also explored new techniques such as disambiguation rules, term ambiguity measurement, and used multi-pass sieve framework at a micro level. For the challenge's primary measure (strict entity), our submissions achieved competitive results (f-measures: 87.3%, 87.1%, and 86.7%). For our preferred measure (binary token HIPAA), our submissions achieved superior results (f-measures: 93.7%, 93.6%, and 93%). With those encouraging results, we gain the confidence to improve and use the tool for the real de-identification task at the UAB Informatics Institute.

Classification of Clinical Research Study Eligibility Criteria to Support Multi-Stage Cohort Identification Using Clinical Data Repositories.

One of the challenges to using electronic health record (EHR) repositories for research is the difficulty mapping study subject eligibility criteria to the query capabilities of the repository. We sought to characterize criteria as "easy" (searchable in a typical repository), "hard" (requiring manual review of the record data), and "impossible" (not typically available in EHR repositories). We obtained 292 criteria from 20 studies available from Clinical Trials.gov and rated them according to our three types, plus a fourth "mixed" type. We had good agreement among three independent reviewers and chose 274 criteria that were characterized by single types for further analysis. The resulting analysis showed typical features of criteria that do and don't map to repositories. We propose that these features be used to guide researchers in specifying eligibility criteria to improve development of enrollment workflow, including the definition of EHR repository queries for self-service or analyst-mediated retrievals.