Topic: Children in care

The MIDAS pilot for “Looked After Children” is based on the existing databases, such as  admissions, social service interventions, etc. The big data analysis on the MIDAS platform should enable the pilot site users to carry out a longitudinal analysis and track a cohort of Looked After Children as they move in and out of care, use a variety of health services, and look at patterns of behaviours and changes over time. The MIDAS platform should enable analysis of the available datasets to identify effective preventative measures and to get new insights about alarming pointers.


The portal SearchPoint exhibits the clustered keywords of a query on the MEDLINE/PubMed open dataset, after searching for a keyword. This interactive visual tool helps to surface information we are looking for, avoiding the standard answer that is biased by definition. SearchPoint can be used in any language with any document set (public or private) that can be indexed, analysed and visualised with this approach.
  • designed to improve the search engine experience; the user provides further information to the search by interacting with the system by dragging a pointer over word clouds
  • these word clouds are produced by cosine similarity to an "average" centred on the topics in each abstract of the set of selected papers, clustered using the k-means algorithm
  • it can be used with any document set (public or private) that can be indexed, analysed and visualised with this approach. 

MIDAS SearchPoint Demo

For example, when we enter a search term ‘Child Care’, the system performs an elasticSearch search over the MEDLINE dataset, extracts groups of keywords that best describe different subgroups of results (these are most relevant, and not most frequent terms). We can also specify that are looking for articles annotated with the MeSH Heading 'Child Care' by inputing MeshHeadingList.desc:"Child Care". It gives us an overview of the content of the retrieved documents (eg. we see groups of results about services, treatments, etc). By moving the cursor over word-groups, we provide the relevance criteria to the search result, thus bringing at the top results the articles we are interested in. For example, the article on the Swedish childhood diabetes studies that occupied the position 153 is now in position 4. The user can read its title and first lines of abstract, and when clicking on it, the system opens the article in the browser at its PubMed url location. 



The news monitoring system in the backend of the MIDAS News Monitoring Dashboard, collects and annotates in real-time news articles published by over 100,000 news publishers worldwide. It provides the user with public health news articles in 10+ languages (including Finish, Basque and Irish) as well as world events mentioned in these articles, permitting to explore what is currently being reported about in the media worldwide.

In the first example, the event of the Zika outbreak is identified immediately after the news articles that report about it are collected. One can explore the evolution of the news publishers awareness of the epidemics in a timeline by looking at the related news articles represented in a world map, as they were identified or updated during a selected period of time. MIDAS News Monitoring tool can find articles and events related to a particular entity, topic, etc.

For each event, the MIDAS News Monitoring Dashboard is able to provide extensive information. Its’ Event clustering permits us to distinguish between subtopics and perspectives in the stories relating to a certain news. It provides us with a list of cross-lingual articles that describe the several aspects of the event, as well as its date, location, and their impact on social media (Twitter). It permits the user to see a real-time dynamic world map of events, or to explore a static set o of news over a determined time range.

A version of this news explorer with MeSH Headings integration (and thus supporting only english language, due to the limitation of the MeSH Headings itself) is available with a two year coverage (2017-2018) at: http://qmidas-news.quintelligence.com/.

This system is based in the Event Registry (eventregistry.org) technology, licensed to the MIDAS project.


The Kibana dashboard permits the user to profit of data visualization modules that feed on his/her datasets built in ElasticSearch. This tool enables one to query the dataset and produce different types of data visualization modules that can later integrate a customized dashboard. Kibana can be used in any language with any document set (public or private) that can be indexed, analyzed and visualized with this approach.
  • designed to improve the user experience in exploring the MEDLINE dataset through visualization modules composing topic dedicated live monitoring dashboards 
  • the visualization modules based on Kibana do not require technical skills and enable the data exploration by a diversity of professionals
  • the system includes a powerful querying engine based on the open-source information retrieval software library Lucene    

MIDAS ElasticSearch Demo

To extract meaningful information from MEDLINE, QUINT is using the underlying MeSH (Medical Subject Headings) ontology-like structure. Most of the articles in MEDLINE are annotated by humans with MeSH Heading descriptors. These permit the user to explore a certain biomedical related topic relying in curated information made available by the NIH. MEDLINE data, together with the MeSH annotation, is indexed with ElasticSearch and made available to analytics and visualisation tools. The MEDLINE dashboard permits the user to profit of data visualisation modules that feed on an instance of the MEDLINE open data sets built in ElasticSearch. In that, Kibana is used for prototyping this tool. It also enables one to query the dataset and produce different types of data visualisation modules that can later integrate a customised dashboard, designed in agreement with the workflow of the end-user.

This system was developed by the AI Lab at the IJS and refocused by Quintelligence within the MIDAS project to visually analyze the MEDLINE dataset. It can be implemented in premises to work with proprietary data. It is currently available as Open Source under the BSD license.


The MeSH Classifier is a tool developed by Quintelligence to classify free text with the latest MeSH Headings provided by NHS. It is based on the DMOZ classifier, learning over 80+ years of MEDLINE data, and over the MeSH tree with 16 major categories and a max of 13 levels of deepness. It provides all the classifying categories with position number and (cosine) similarity weight, with a slider and a number of max categories visible. It available through a web app and an API. 

  • designed to classify free text of any nature with the classes of MeSH Headings where MEDLINE is based on, to which health professionals are familiar with 
  • it can classify articles that haven't yet been annotated by the NHS, or official WHO documents of interest
  • it can also classify news articles and be used to monitor worldwide news based on the classification provided at the MeSH Headings  

    MIDAS MeSH Classifier Demo

The automatic annotation tool of free text provided by Quintelligence permits the user to annotate any piece of text with the MeSH Heading descriptors as made available by the NIH. The system learned over the dataset of years of MEDLINE records MeSH annotated by humans, to enable this automatic annotation based on text similarity. In detail it considers the average bag of words matching to centroids. It provides an ordered list of MeSH terms and their relevance rank in this classification. It permits the annotation of abstracts in MEDLINE that were not yet annotated. Also it potentiates the annotation of free text over medical records or even health related news. It is accessible over an API through a POST call where the user specifies the body of text. Having those classes available for any free text permits the integration with systems where one can query the data based on MeSH Headings.

The core system was developed by the AI Lab at the IJS and refocused by Quintelligence within the MIDAS project to use the MeSH Headings to classify free text. It can be implemented in premises to work with proprietary data. It is currently available as Open Source under the BSD license.