Graham Else - Researching the use of Neo4j for SNOMED patient data analysis

[ge1]

Hi Everyone,

I've been working on a Cardiac research data repository, using SQL to store anonymised patient diagnostics and procedural notes and am intrigued on how a much better analytical tool can be built using a Graph database. Having seen a number of examples on You Tube etc. it looks a truly exciting prospect and I also want to integrate links to Imaging studies from the results.

I've downloaded and built the SNOMED schema into a Neo4j instance locally but am now stumped. I cannot get to grips with how patient data is added. Part of this is , I think, the understanding of SNOMED and combining this with the normal analysis based on things like ethnicity, age, date of event, but, even without these considerations, don't really know where to start.

Having started reading the documentation, any real world advice would be greatly appreciated to make the intellectual leap to understand how relations are built between what is really a coding umbrella and the data that has employed it. All the examples I have seen only talk about loading the SNOMED coding system (and a lot about the verification) and then a one liner about loading the patient data.

Any tips, links, and advice will be greatly appreciated!

Kind regards

Graham

[clem]

Presumably, your data is in a CSV file?

If so, look at this:

Neo4j Graph Database Platform

Importing CSV Data into Neo4j - Developer Guides

This article demonstrates different approaches to importing CSV data into Neo4j and solutions to potential issues that might arise during the process.

I also have this hint using CSV files:

Help creating graph from json file Neo4j Graph Platform

What I have found helpful (at least with CSV files), is to take it step by step. First, see if you can create a Node. And as a first step to that, see if you can create a table of the properties of the nodes (not yet created). CALL apoc.some_call("file:///somefile") YIELD value RETURN value.field1, value.field2, ... LIMIT 10 Then you should get a table. It's going to be a bit more complex than what I've shown because it's JSON, and you have to chase all the JSON elements. If that works, t…

[ge1]

Thanks Clem, I've worked through the basics now and can see how this works! I just need to understand whether Cypher can do the same sorts of queries as the SNOMED CT expression constraint language...
Thanks again for the tips.
Kind regards
Graham

[clem]

I'm sure you will do fine then. The Cypher language is pretty easy compared to SQL.

There are a few quirks but usually you can do a Google search and find the answer. The tutorial and documentation is generally pretty good too.

Look to the APOC libraries if there's something that seems a bit too sophisticated for Cypher to do. (It's easier to add to the APOC library vs. tweaking the Cypher Language.)

The hardest problem that I see newbies have, is that they are used to computer languages are almost always procedural, but most query languages are declarative. Hence creating a complex Cypher query right off the bat can be hard.

As a beginner, it's best to start with as a simple version of your query as possible and add to it gradually.

Have fun!