Clinical coverage of an archetype repository over SNOMED-CT.

Clinical archetypes provide a means for health professionals to design what should be communicated as part of an Electronic Health Record (EHR). An ever-growing number of archetype definitions follow this health information modelling approach, and this international archetype resource will eventually cover a large number of clinical concepts. On the other hand, clinical terminology systems that can be referenced by archetypes also have a wide coverage over many types of health-care information. No existing work measures the clinical content coverage of archetypes using terminology systems as a metric. Archetype authors require guidance to identify under-covered clinical areas that may need to be the focus of further modelling effort according to this paradigm. This paper develops a first map of SNOMED-CT concepts covered by archetypes in a repository by creating a so-called terminological Shadow. This is achieved by mapping appropriate SNOMED-CT concepts from all nodes that contain archetype terms, finding the top two category levels of the mapped concepts in the SNOMED-CT hierarchy, and calculating the coverage of each category. A quantitative study of the results compares the coverage of different categories to identify relatively under-covered as well as well-covered areas. The results show that the coverage of the well-known National Health Service (NHS) Connecting for Health (CfH) archetype repository on all categories of SNOMED-CT is not equally balanced. Categories worth investigating emerged at different points on the coverage spectrum, including well-covered categories such as Attributes, Qualifier value, under-covered categories such as Microorganism, Kingdom animalia, and categories that are not covered at all such as Cardiovascular drug (product).

Meaningful Activity Replacement Recommendations in Dementia.

Exercise of meaningful activities is important for people living with dementia, both for quality of life and to maintain the necessary basic activities of daily living. A method is proposed for recommendation of replacements for lost meaningful activities that accounts for the need to maintain activities of daily living.

Towards an ontology for soft robots: what is soft?

The advent of soft robotics represents a profound change in the forms robots will take in the future. However, this revolutionary change has already yielded such a diverse collection of robots that attempts at defining this group do not reflect many existing 'soft' robots. This paper aims to address this issue by scrutinising a number of descriptions of soft robots arising from a literature review with the intention of determining a coherent meaning for soft. We also present a classification of existing soft robots to initiate the development of a soft robotic ontology. Finally, discrepancies in prescribed ranges of Young's modulus, a frequently used criterion for the selection of soft materials, are explained and discussed. A detailed visual comparison of these ranges and supporting data is also presented.

Using a generalised identity reference model with archetypes to support interoperability of demographics information in electronic health record systems.

Computerised identity management is in general encountered as a low-level mechanism that enables users in a particular system or region to securely access resources. In the Electronic Health Record (EHR), the identifying information of both the healthcare professionals who access the EHR and the patients whose EHR is accessed, are subject to change. Demographics services have been developed to manage federated patient and healthcare professional identities and to support challenging healthcare-specific use cases in the presence of diverse and sometimes conflicting demographic identities. Demographics services are not the only use for identities in healthcare. Nevertheless, contemporary EHR specifications limit the types of entities that can be the actor or subject of a record to health professionals and patients, thus limiting the use of two level models in other healthcare information systems. Demographics are ubiquitous in healthcare, so for a general identity model to be usable, it should be capable of managing demographic information. In this paper, we introduce a generalised identity reference model (GIRM) based on key characteristics of five surveyed demographic models. We evaluate the GIRM by using it to express the EN13606 demographics model in an extensible way at the metadata level and show how two-level modelling can support the exchange of instances of demographic identities. This use of the GIRM to express demographics information shows its application for standards-compliant two-level modelling alongside heterogeneous demographics models. We advocate this approach to facilitate the interoperability of identities between two-level model-based EHR systems and show the validity and the extensibility of using GIRM for the expression of other health-related identities.

Confident predictability: identifying reliable gene expression patterns for individualized tumor classification using a local minimax kernel algorithm.

BACKGROUND: Molecular classification of tumors can be achieved by global gene expression profiling. Most machine learning classification algorithms furnish global error rates for the entire population. A few algorithms provide an estimate of probability of malignancy for each queried patient but the degree of accuracy of these estimates is unknown. On the other hand local minimax learning provides such probability estimates with best finite sample bounds on expected mean squared error on an individual basis for each queried patient. This allows a significant percentage of the patients to be identified as confidently predictable, a condition that ensures that the machine learning algorithm possesses an error rate below the tolerable level when applied to the confidently predictable patients. RESULTS: We devise a new learning method that implements: (i) feature selection using the k-TSP algorithm and (ii) classifier construction by local minimax kernel learning. We test our method on three publicly available gene expression datasets and achieve significantly lower error rate for a substantial identifiable subset of patients. Our final classifiers are simple to interpret and they can make prediction on an individual basis with an individualized confidence level. CONCLUSIONS: Patients that were predicted confidently by the classifiers as cancer can receive immediate and appropriate treatment whilst patients that were predicted confidently as healthy will be spared from unnecessary treatment. We believe that our method can be a useful tool to translate the gene expression signatures into clinical practice for personalized medicine.