convert into RDF MIMIC dataset

[gatemezing]

Hi, would it be possible to convert the MIMIC dataset in RDF (graph model) to be used by those interested in easy mapping and SPARQL? I can help if I can have access to the raw data.

[tompollard]

Hi @gatemezing, yes this should be possible. What do you mean by the "raw" data? The full MIMIC dataset is available via the process at: https://mimic.physionet.org/gettingstarted/access/. We are not able to share the data used to generate MIMIC because it contains sensitive information.

[gatemezing]

Hi @tompollard Thanks for the link! I meant having access to the anonymized data so that I can generate an RDF version of the data for further research.

[tompollard]

Thanks @gatemezing. You could begin by developing an approach using the demo dataset at: https://mimic.physionet.org/gettingstarted/demo/. To access the full MIMIC-III dataset, you will need to complete a formal data access request (see website).

To download MIMIC (and the MIMIC demo) you agree not to share the data with anyone else, but if you were able to generate an RDF version then we could arrange to publish it alongside MIMIC with the same data access restrictions.

[gatemezing]

Sure! Let me try to play with the demo dataset and with that version in RDF, I will request the full version of MIMIC to apply the same conversion. Thanks!

[gatemezing]

I've managed to get a first version of RDF dataset. Now I will modify and/do some alignments of the properties and classes mostly used in Linked Data Space as per Linked Open Data Vocabulary http://lov.okfn.org/dataset/lov/

[alistairewj]

Great! I look forward to taking a look. One of my colleagues also suggested looking at neo 4j if you haven't.

[gatemezing]

Sorry for the delay, busy and forgot to send you the first step of the work :)
So, I upload a local copy of the DB and use this nice tool Dr2Q

jdbc:postgresql://localhost:5432/mimiciii # connection to DB
./generate-mapping.sh -u -p -o mapping.ttl jdbc:postgresql://localhost:5432/mimiciii ./dump-rdf.sh -f TURTLE -b http://data.mimic.org/id/ mapping.ttl > mimic_demo.ttl

IMHO, the next step could be to build a consensus vocabulary for concepts and properties, and then define a policy for the URIs.
HTH