maastrichtu-cds / datafairifier Goto Github PK

In the Jupyternotebook:
When the user is writing the SQL query, including mapping the column headers to required variable names,
it would be convenient if he/she could see a list with these column headers displayed in the notebook. Otherwise, he/she has to open and inspect the database elsewhere.

When there is NaN information in the age column, stored as string in postgres. The R2RML conversiont tool breaks on these values, and discards all subsequent rows. This results in a list of missing patients.
The same happens with NULL values in the database.

We need to run SeDI as a Docker

Although we have defined R2RML mappings, we need to test the performance of Ontop in comparison to constructing all triples. Is there a performance gain/reduction?

The RAM memory threshold (1 GB) seems to be limiting loading RDF triples in GraphDB.
-> increase threshold

• Nice to have: selectboxes for X and Y vars

Create dcm4che docker image to use in the image pathway

This is possible in Protege, but is an ontop-proprietary solution.

The question is how we can make an R2RML mapping using a graphical tool which helps in selecting the right tables/columns/values.

• SQL syntax highlighting
• index for pandas dataframe

Currently, the user has to manually create two new repositories in GraphDB when starting the system:

• One for database
• One for the R2RML file
  This should be automated and part of starting GraphDB

R2RML conversion tool can't handle different SQL 'dialects', i.e., server database access versus direct access to postgres database within our system

A Docker containing all MIA micro-services except the UniversalWorker

We need a standard CAT CTP Docker image to anonymize the DICOM data

The paragraph on configuration of the infrastructure is not really clear to me @jvsoest : Is this targeted at people who attend a workshop or hackathon at Maastro?
Maybe we could move this paragraph to the documentation files, and have only generic high-level install&run&use instructions in the ReadMe.

Visualization of database output is currently based on plotting the values, but the database columns often contain non-numeric values.
Therefore we would rather visualize the count of variables

We start the jupyter notebook assuming that the data is already loaded in Virtuoso. For automatic import we assume there is a csv file in the same directory as docker-compose.

To deploy the imaging pathway we need to deploy all.

security and audit log for the imaging pathway

I created a link between an object and two different subjects through two separate predicates. However, when visualizing, the function allocates them to the same Literal in this case. See figure below:

This should be visualized as one object with two separate LITERAL strings, connected by one arrow each.

We need to reuse/test an available blazegraph Docker or create a new one ourselves.

For the imaging and database pathway we need a Docker image with a postgreSQL server with the CAT key database

To build the Universal Worker for Docker we need to make a linux compatible matlab jar.
This should be used to create a Linux UW and stored into a docker container.
We need to add a linux build agent to our Bamboo

Placeholder for enhancement.

Currently the jupyter notebook reads the ontology to feed the terminology mapping. However, we can also load the ontology in a separate database, when importing the R2RML.

Advantages:

• we don't need to reload every time the ontology
• we can use the ontology during querying (using SERVICE <….> options in sparql)
• we can use it in the mapping tool, including inferencing on subClassOf* reasoning. Some terminology options are sometimes a logical statement between the concepts "delineation" AND "specific organ/tissue".