book recommendation - basic formal ontology (BFO) for biomedicine

Clinical (archive)
1

A new book, Building Ontologies with Basic Formal Ontology (BFO2) has been published by Robert Arp, Barry Smith and Andrew Spear.

Amazon.com
Amazon.co.uk

I’ve read a pre-print, and this is an excellent book. Many of us here have studied or used some of the key upper level ontologies (BFO, BioTopLite etc), and I think BFO2 will probably end up being the one of choice for general biomedicine. It contains all the concepts from Barry Smith’s earlier work (SNAP and SPAN - spatial and temporal regions, partonomy concepts etc) and will ultimately (I have been told) contain/be merged with a new version of the Information Artefact Ontology (IAO) which will probably become the upper level ontology for describing types of information that stand in the IS-ABOUT and similar relationships with real world referents - in other words, EHR information items.

My personal feeling is that in the next 2-5 years, we will finally see the joining up of these key ontologies with information models and archetypes in the clinical information space. It will be up to us in the openEHR community and other related communities (13606, CIMI, HL7 etc) to engage with this material and consider how this integration will be achieved. A few very early ideas are mentioned in the Archetype Technology Overview document, but of course this is just one narrow area of application.

We need to all get on the same page in terms of understanding and conceptual nomenclature in this space, and BFO2 I believe is an excellent foundation for that. We will be using it increasingly in the openEHR specification space, and I think others will find it useful as well.

  • thomas
2

Thank you, Thomas, this looks very promising. I have placed an order so now I am waiting :slight_smile: Very considerate of you, many thanks. Kind regards, Rikard

3

is there a kindle version

4

Yes, - $32!

Sincerely

Einar Fosse

5

Me too - nice bed-time reading.

I will reserve judgement for the “2-5 years and we will be using this”. I agree this is the future but it still feels a lot like nuclear fusion to me - nice to have but a b****r to use (at least in our messy world of clinical documentation).

Ian

6

Kindle - yes

7

Dear All,

agree partly with Ian’s assessment, i.e. about the messiness. While I much appreciate what I have read, and I’ve had much help from earlier texts from the authors (as I’m sure I will from this book), there is in the medical informatics community a widespread belief that the position held by (some part of) the BFO community is undisputed and sort-of final. There are still issues which requires careful consideration, especially regarding information artefacts and the is-about relationship [1, 2], but also about e.g. dispositions [3], and functions [4].

Additionally, while ontologies deal with what is universally true, it is my belief that universal truth takes, and should take, the back seat compared to user needs and practicality in information modelling. First-world (using Popper’s ontology [5]) ontologies are outcomes of our understanding of the physical world and evolve as science evolves (at least good ones). Information models and other second-third-world ontologies are always constructs and, like with fictional characters, nothing can be discovered by examining those models in addition to what has been explicitly stated. For this reason, ontology as a method isn’t as helpful for information modellers as it is for others.

  1. https://www.researchgate.net/publication/266021648_An_Ontological_Analysis_of_Reference_in_Health_Record_Statements

  2. http://www.amazon.com/Aboutness-Carl-G-Hempel-Lecture/dp/0691144958

  3. http://www.amazon.com/Dispositions-Stephen-Mumford/dp/0199259828

  4. http://www.amazon.com/Functions-Biological-Artificial-Worlds-Philosophical/dp/026211321X

  5. https://en.wikipedia.org/wiki/Popper%27s_three_worlds

/Daniel

8

Thanks Thomas! Will study all
Of it and get back!
Beatriz

9

Daniel,

nice observations and thanks for the references - I had intended to circulate the first one as well, at some point. For my part, I think ontology will become useful in information modelling precisely for the reason that it offers ways of representing distinctions between real world referents and the things that refer to them (information entities). In other words, to help the IT sector understand 'how to model'. Historically it has been completely confused (I would go so far as to say not even conscious) of the difference between real world entities and events and the information items that document them.

With no understanding of the in-principle divide between ontological and epistemological points of view (or equivalently of Popper's 3 worlds), information modelling can't possibly achieve much clarity or computability.

So in agreeing with you, I would add that ontology-thinking is not a recipe for how to do information modelling, but it is useful for understanding what not to express in information models - mind-independent truths.

- thomas

10

I think within the timeframe stated by Tom we will definitely see these two worlds coming closer and producing some tangible benefits. Did you all notice an increase in research in this area these days? Just my quick 2 cents…still on travel

Cheers,

-koray

11

Dear Colleagues,

Some personal thoughts and opinion.

1- Scopes
EHR Scope: Document, archive and exchange Statements by one author about one patient using concepts as terms
Ontology scope: Creating a definition about one concept using other concepts expressed as rules.

2- Universal and Particular truths
Authored Statements about one patient by one author are always Particular in nature.
Definitions about a concept are Universal in nature.

3- Open and Closed world Assumption
Statements in the database of the EHR are authored by one person and about one person.
What is not entered in the database does not exist. It is not authored.
The data in an EHR database are following the Closed World Assumption.
The consequence is that it is much better to ascertain patient safety.

In Ontologies that are expressed using rules that define the relationships, new rules can be inferred.
The Concept definitions in an Ontology are following the Open World Assumption.
The consequence is that it is impossible to ascertain patient safety.

4- Deployment of systems that combine both worlds need to think about the interfaces where these Models intersect.
In the EHR several models are used:

  • Model for documentation, archiving and exchange (RM of part 1)
  • Model for Statement expression (AOM part 2)
  • Model for Statement content: Reference Archetypes (e.g. SIAMM, DCM) that use Concepts
  • Model for the definition of Concepts that are used in Statements
    The first 3 are Structured, Particular, Closed World Assumption models.
    The last one is an Universal, Open World Assumption model.

Deploying all four models using ontological techniques will (possibly) create problems.
The language used, at present, is OWL-DL. OWL-DL is instantiating first order logic, partially.
All four models constitute all together multiple order logic.
There is not enough evidence that we (in eHealth) are capable to handle multiple order logic using ontological methods.

5- Taken all in consideration.
For reasons given above I agree with Ian and Daniel, that we had better create clear, simple, well understood, delineations between all model intersections (e.i. their interfaces).

6- I propose to use the Two Level Modeling Paradigm creating structures used for the documentation of Statements.
And have Rules Engines do the inferencing in a controlled way.

Only use Ontologies for the definition of ‘simple’, non-complex, terms/concepts and use these defined terms in a Terminology in Statements.
Use them in Statements as lemma’s from a dictionary.
This is using ontological defined terms in their natural well understood role,

Gerard Freriks
+31 620347088
gfrer@luna.nl