Modelling expertise at different levels of granularity using semantic similarity measures in the context of collaborative knowledge-curation platforms.

Collaboration platforms provide a dynamic environment where the content is subject to ongoing evolution through expert contributions. The knowledge embedded in such platforms is not static as it evolves through incremental refinements - or micro-contributions. Such refinements provide vast resources of tacit knowledge and experience. In our previous work, we proposed and evaluated a Semantic and Time-dependent Expertise Profiling (STEP) approach for capturing expertise from micro-contributions. In this paper we extend our investigation to structured micro-contributions that emerge from an ontology engineering environment, such as the one built for developing the International Classification of Diseases (ICD) revision 11. We take advantage of the semantically related nature of these structured micro-contributions to showcase two major aspects: (i) a novel semantic similarity metric, in addition to an approach for creating bottom-up baseline expertise profiles using expertise centroids; and (ii) the application of STEP in this new environment combined with the use of the same semantic similarity measure to both compare STEP against baseline profiles, as well as to investigate the coverage of these baseline profiles by STEP.

A supervised approach to quantifying sentence similarity: with application to evidence based medicine.

Following the Evidence Based Medicine (EBM) practice, practitioners make use of the existing evidence to make therapeutic decisions. This evidence, in the form of scientific statements, is usually found in scholarly publications such as randomised control trials and systematic reviews. However, finding such information in the overwhelming amount of published material is particularly challenging. Approaches have been proposed to automatically extract scientific artefacts in EBM using standardised schemas. Our work takes this stream a step forward and looks into consolidating extracted artefacts-i.e., quantifying their degree of similarity based on the assumption that they carry the same rhetorical role. By semantically connecting key statements in the literature of EBM, practitioners are not only able to find available evidence more easily, but also can track the effects of different treatments/outcomes in a number of related studies. We devise a regression model based on a varied set of features and evaluate it both on a general English corpus (the SICK corpus), as well as on an EBM corpus (the NICTA-PIBOSO corpus). Experimental results show that our approach performs on par with the state of the art on the general English and achieves encouraging results on the biomedical text when compared against human judgement.

Semantic interestingness measures for discovering association rules in the skeletal dysplasia domain.

BACKGROUND: Lately, ontologies have become a fundamental building block in the process of formalising and storing complex biomedical information. With the currently existing wealth of formalised knowledge, the ability to discover implicit relationships between different ontological concepts becomes particularly important. One of the most widely used methods to achieve this is association rule mining. However, while previous research exists on applying traditional association rule mining on ontologies, no approach has, to date, exploited the advantages brought by using the structure of these ontologies in computing rule interestingness measures. RESULTS: We introduce a method that combines concept similarity metrics, formulated using the intrinsic structure of a given ontology, with traditional interestingness measures to compute semantic interestingness measures in the process of association rule mining. We apply the method in our domain of interest - bone dysplasias - using the core ontologies characterising it and an annotated dataset of patient clinical summaries, with the goal of discovering implicit relationships between clinical features and disorders. Experimental results show that, using the above mentioned dataset and a voting strategy classification evaluation, the best scoring traditional interestingness measure achieves an accuracy of 57.33%, while the best scoring semantic interestingness measure achieves an accuracy of 64.38%, both at the recall cut-off point 5. CONCLUSIONS: Semantic interestingness measures outperform the traditional ones, and hence show that they are able to exploit the semantic similarities inherently present between ontological concepts. Nevertheless, this is dependent on the domain, and implicitly, on the semantic similarity metric chosen to model it.

Identifying scientific artefacts in biomedical literature: the Evidence Based Medicine use case.

Evidence Based Medicine (EBM) provides a framework that makes use of the current best evidence in the domain to support clinicians in the decision making process. In most cases, the underlying foundational knowledge is captured in scientific publications that detail specific clinical studies or randomised controlled trials. Over the course of the last two decades, research has been performed on modelling key aspects described within publications (e.g., aims, methods, results), to enable the successful realisation of the goals of EBM. A significant outcome of this research has been the PICO (Population/Problem-Intervention-Comparison-Outcome) structure, and its refined version PIBOSO (Population-Intervention-Background-Outcome-Study Design-Other), both of which provide a formalisation of these scientific artefacts. Subsequently, using these schemes, diverse automatic extraction techniques have been proposed to streamline the knowledge discovery and exploration process in EBM. In this paper, we present a Machine Learning approach that aims to classify sentences according to the PIBOSO scheme. We use a discriminative set of features that do not rely on any external resources to achieve results comparable to the state of the art. A corpus of 1000 structured and unstructured abstracts - i.e., the NICTA-PIBOSO corpus - is used for training and testing. Our best CRF classifier achieves a micro-average F-score of 90.74% and 87.21%, respectively, over structured and unstructured abstracts, which represents an increase of 25.48 percentage points and 26.6 percentage points in F-score when compared to the best existing approaches.

Inferring characteristic phenotypes via class association rule mining in the bone dysplasia domain.

Finding, capturing and describing characteristic features represents a key aspect in disorder definition, diagnosis and management. This process is particularly challenging in the case of rare disorders, due to the sparse nature of data and expertise. From a computational perspective, finding characteristic features is associated with some additional major challenges, such as formulating a computationally tractable definition, devising appropriate inference algorithms or defining sound validation mechanisms. In this paper we aim to deal with each of these problems in the context provided by the skeletal dysplasia domain. We propose a clear definition for characteristic phenotypes, we experiment with a novel, class association rule mining algorithm and we discuss our lessons learned from both an automatic and human-based validation of our approach.

Creating a behavioural classification module for acceleration data: using a captive surrogate for difficult to observe species.

Distinguishing specific behavioural modes from data collected by animal-borne tri-axial accelerometers can be a time-consuming and subjective process. Data synthesis can be further inhibited when the tri-axial acceleration data cannot be paired with the corresponding behavioural mode through direct observation. Here, we explored the use of a tame surrogate (domestic dog) to build a behavioural classification module, and then used that module to accurately identify and quantify behavioural modes within acceleration collected from other individuals/species. Tri-axial acceleration data were recorded from a domestic dog whilst it was commanded to walk, run, sit, stand and lie-down. Through video synchronisation, each tri-axial acceleration sample was annotated with its associated behavioural mode; the feature vectors were extracted and used to build the classification module through the application of support vector machines (SVMs). This behavioural classification module was then used to identify and quantify the same behavioural modes in acceleration collected from a range of other species (alligator, badger, cheetah, dingo, echidna, kangaroo and wombat). Evaluation of the module performance, using a binary classification system, showed there was a high capacity (>90%) for behaviour recognition between individuals of the same species. Furthermore, a positive correlation existed between SVM capacity and the similarity of the individual's spinal length-to-height above the ground ratio (SL:SH) to that of the surrogate. The study describes how to build a behavioural classification module and highlights the value of using a surrogate for studying cryptic, rare or endangered species.

Recognizing scientific artifacts in biomedical literature.

Today's search engines and digital libraries offer little or no support for discovering those scientific artifacts (hypotheses, supporting/contradicting statements, or findings) that form the core of scientific written communication. Consequently, we currently have no means of identifying central themes within a domain or to detect gaps between accepted knowledge and newly emerging knowledge as a means for tracking the evolution of hypotheses from incipient phases to maturity or decline. We present a hybrid Machine Learning approach using an ensemble of four classifiers, for recognizing scientific artifacts (ie, hypotheses, background, motivation, objectives, and findings) within biomedical research publications, as a precursory step to the general goal of automatically creating argumentative discourse networks that span across multiple publications. The performance achieved by the classifiers ranges from 15.30% to 78.39%, subject to the target class. The set of features used for classification has led to promising results. Furthermore, their use strictly in a local, publication scope, ie, without aggregating corpus-wide statistics, increases the versatility of the ensemble of classifiers and enables its direct applicability without the necessity of re-training.

Decomposing phenotype descriptions for the human skeletal phenome.

Over the course of the last few years there has been a significant amount of research performed on ontology-based formalization of phenotype descriptions. The intrinsic value and knowledge captured within such descriptions can only be expressed by taking advantage of their inner structure that implicitly combines qualities and anatomical entities. We present a meta-model (the Phenotype Fragment Ontology) and a processing pipeline that enable together the automatic decomposition and conceptualization of phenotype descriptions for the human skeletal phenome. We use this approach to showcase the usefulness of the generic concept of phenotype decomposition by performing an experimental study on all skeletal phenotype concepts defined in the Human Phenotype Ontology.

Mining skeletal phenotype descriptions from scientific literature.

Phenotype descriptions are important for our understanding of genetics, as they enable the computation and analysis of a varied range of issues related to the genetic and developmental bases of correlated characters. The literature contains a wealth of such phenotype descriptions, usually reported as free-text entries, similar to typical clinical summaries. In this paper, we focus on creating and making available an annotated corpus of skeletal phenotype descriptions. In addition, we present and evaluate a hybrid Machine Learning approach for mining phenotype descriptions from free text. Our hybrid approach uses an ensemble of four classifiers and experiments with several aggregation techniques. The best scoring technique achieves an F-1 score of 71.52%, which is close to the state-of-the-art in other domains, where training data exists in abundance. Finally, we discuss the influence of the features chosen for the model on the overall performance of the method.

Decision support methods for finding phenotype--disorder associations in the bone dysplasia domain.

A lack of mature domain knowledge and well established guidelines makes the medical diagnosis of skeletal dysplasias (a group of rare genetic disorders) a very complex process. Machine learning techniques can facilitate objective interpretation of medical observations for the purposes of decision support. However, building decision support models using such techniques is highly problematic in the context of rare genetic disorders, because it depends on access to mature domain knowledge. This paper describes an approach for developing a decision support model in medical domains that are underpinned by relatively sparse knowledge bases. We propose a solution that combines association rule mining with the Dempster-Shafer theory (DST) to compute probabilistic associations between sets of clinical features and disorders, which can then serve as support for medical decision making (e.g., diagnosis). We show, via experimental results, that our approach is able to provide meaningful outcomes even on small datasets with sparse distributions, in addition to outperforming other Machine Learning techniques and behaving slightly better than an initial diagnosis by a clinician.

Supervised segmentation of phenotype descriptions for the human skeletal phenome using hybrid methods.

BACKGROUND: Over the course of the last few years there has been a significant amount of research performed on ontology-based formalization of phenotype descriptions. In order to fully capture the intrinsic value and knowledge expressed within them, we need to take advantage of their inner structure, which implicitly combines qualities and anatomical entities. The first step in this process is the segmentation of the phenotype descriptions into their atomic elements. RESULTS: We present a two-phase hybrid segmentation method that combines a series individual classifiers using different aggregation schemes (set operations and simple majority voting). The approach is tested on a corpus comprised of skeletal phenotype descriptions emerged from the Human Phenotype Ontology. Experimental results show that the best hybrid method achieves an F-Score of 97.05% in the first phase and F-Scores of 97.16% / 94.50% in the second phase. CONCLUSIONS: The performance of the initial segmentation of anatomical entities and qualities (phase I) is not affected by the presence / absence of external resources, such as domain dictionaries. From a generic perspective, hybrid methods may not always improve the segmentation accuracy as they are heavily dependent on the goal and data characteristics.

The Bone Dysplasia Ontology: integrating genotype and phenotype information in the skeletal dysplasia domain.

BACKGROUND: Skeletal dysplasias are a rare and heterogeneous group of genetic disorders affecting skeletal development. Patients with skeletal dysplasias suffer from many complex medical issues including degenerative joint disease and neurological complications. Because the data and expertise associated with this field is both sparse and disparate, significant benefits will potentially accrue from the availability of an ontology that provides a shared conceptualisation of the domain knowledge and enables data integration, cross-referencing and advanced reasoning across the relevant but distributed data sources. RESULTS: We introduce the design considerations and implementation details of the Bone Dysplasia Ontology. We also describe the different components of the ontology, including a comprehensive and formal representation of the skeletal dysplasia domain as well as the related genotypes and phenotypes. We then briefly describe SKELETOME, a community-driven knowledge curation platform that is underpinned by the Bone Dysplasia Ontology. SKELETOME enables domain experts to use, refine and extend and apply the ontology without any prior ontology engineering experience--to advance the body of knowledge in the skeletal dysplasia field. CONCLUSIONS: The Bone Dysplasia Ontology represents the most comprehensive structured knowledge source for the skeletal dysplasias domain. It provides the means for integrating and annotating clinical and research data, not only at the generic domain knowledge level, but also at the level of individual patient case studies. It enables links between individual cases and publicly available genotype and phenotype resources based on a community-driven curation process that ensures a shared conceptualisation of the domain knowledge and its continuous incremental evolution.

Federated repositories of X-ray diffraction images.

There is a pressing need for the archiving and curation of raw X-ray diffraction data. This information is critical for validation, methods development and improvement of archived structures. However, the relatively large size of these data sets has presented challenges for storage in a single worldwide repository such as the Protein Data Bank archive. This problem can be avoided by using a federated approach, where each institution utilizes its institutional repository for storage, with a discovery service overlaid. Institutional repositories are relatively stable and adequately funded, ensuring persistence. Here, a simple repository solution is described, utilizing Fedora open-source database software and data-annotation and deposition tools that can be deployed at any site cheaply and easily. Data sets and associated metadata from federated repositories are given a unique and persistent handle, providing a simple mechanism for search and retrieval via web interfaces. In addition to ensuring that valuable data is not lost, the provision of raw data has several uses for the crystallographic community. Most importantly, structure determination can only be truly repeated or verified when the raw data are available. Moreover, the availability of raw data is extremely useful for the development of improved methods of image analysis and data processing.

Radiation-induced trismus in head and neck cancer patients.

PURPOSE: To determine the incidence of trismus in patients who had previously received curative doses of radiation therapy (RT) for head and neck cancer. In addition, we assessed if trismus was associated with quality of life deficits and radiation toxicity. METHODS AND MATERIALS: Between February, 2005 and December, 2006, 40 patients with histologically confirmed head and neck cancer who had received curative doses of RT to the area(s) of the masticatory muscles and/or the ligaments of the temporomandibular joint (TMJ) were enrolled in this study. Differences in trismus incidence were compared between cancer treatment modalities [i.e., RT vs RT/chemotherapy (CT) and conventional RT vs intensity modulated RT]. Quality of life (QOL) was measured by using four questions from the EORTC QLQ-C30 that address pain and difficulty opening the jaw. Scores regarding impaired eating as a result of decreased range of motion of the mouth were derived from the Modified Common Toxicity Criteria (CTCAE Version 3.0). RESULTS: Trismus was identified in 45% of subjects who had received curative doses of RT. No differences were noted in the incidence of trismus between RT and RT/CT or between conventional RT and intensity modulated RT (IMRT). Those with trismus demonstrated more QOL deficits than the non-trismus group. CONCLUSIONS: Curative doses of RT for head and neck cancer result in trismus in a high percentage of patients, independent of other treatment modalities. Trismus has a negative impact on quality of life in this population.

Prescribing incentive schemes : a useful approach?

INTRODUCTION: From 2000 to 2004, primary care organisations (PCOs) in England were legally required to operate a prescribing incentive scheme for their general practices. A statutory framework specified the types of target, maximum rewards and use of 'good cause for failure' provisions that schemes should include. Our objective was to explore whether schemes might be a useful approach to encourage 'good quality' prescribing. METHODS: We requested copies of the original schemes from all PCOs in England in 2001 and 2002. Data were extracted on the rewards offered, types of budgetary targets set and additional conditions specified. RESULTS: Many schemes had not been finalised, some PCOs had no scheme, and one scheme operated without rewards. Although schemes covered similar therapeutic areas, they varied considerably in their length, complexity, reward levels and reward structure. Over half the schemes contained no 'good cause for failure' provision. DISCUSSION/CONCLUSION: PCOs are offering diverse incentives to general practices and some have interpreted the statutory framework imaginatively. Better use of the 'good cause for failure' provision may help to overcome inflationary pressures on prescribing, but further research is needed to clarify the role of financial incentives in influencing prescribing.