----- Original Message ----- From: "Jeremy Rogers" <[EMAIL PROTECTED]> To: <[EMAIL PROTECTED]> Sent: Monday, October 01, 2001 5:23 PM Subject: Re: interface language important ?
> -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > > elaborate an (open component) ontology you can use for : > > - interface > > - data storage > > - document publishing > > What about data analysis, data merging and data abstraction ? Next step > I would agree that it is very important to be able to build a useable > interface - and also that this is one area where the more popular > offerings (ICD etc) have been especially poor. But, equally, I > believe that a solution which supports a good interface and adequate > storage is still suboptimal if, in fact, it can not support data > analysis for the purposes of decision support - and note here that my > requirements for decision support are very different from those of > statistical aggregation. We will start the elaboration of a blackboard next month, precisely to handle decision support. > I would not, for example, by impressed by evidence that data entered > could be converted into (for example) ICD codes. From the point of > view of those of us trying to implement decision support, that is not > progress. I perfectly agree. However finding the proper ICD codes is a decision support process (not the most interesting one, I confess). > Can the data that Odyssee captures function as the input to a machine > reasoning or decision support system, where data abstraction and > classification can take place ? What do you mean by "data abstraction and classification can take place". We currently are able to store, in a fully structured way, a whole echocardiographic report (I give this example because it is one of the most complicated kind of medical report - the heart is a pump). > I could well believe that translation > between human languages can be supported, precisely because human > language is so imprecise and ambiguous. I never wrote anywhere we are doing langage translation : what we do is langage synthesis from structured datas. Furthermore, I would say that I don't believe in medical langage translation precisely because langage is imprecise and ambigous. The major part of our job in Odyssee is building domain models, and interfaces that are able to softly guide the user towards these models (the Fils guides). > But I remain concerned that > what Odyssee captures does not appear to me to be suitable as the > input to the kinds of machine reasoning system that I have > encountered. I can only say that the problem has never been to capture "perfectly suitable datas for machine reasoning" ; the problem is to be able to have doctors and health networks use in daily practice "reasonably structured" systems. Because if your system is not everywhere in the care network, it is not worth reasoning at all. The "Mod�le Unifi�" already has the proper techniques for capture, storage and publishing of fine grained medical datas. What is the difference between a "daily used" software and a good prototype ? : for echographic systems it can mean 10 years of continuous improvement before you can have as well the "ordinary cardiologist" and the "university hospital expert" use it. > There is much similarity of approach between Odyssee and > the Rotterdam ORCA system, and I am familiar with the analytical > difficulties experienced with the data captured by ORCA. Jeremy, I don't know ORCA, but what I know is that - if I have been able to store in a fully structured way a whole echographic or endoscopic report - if from these data I can elaborate a natural langage report with a sufficient quality for the doctor to sign it (in more than 80 hospitals now) I would be very stupid if I can't find what I want in these datas for decision support purposes. <positive behaviour> Could you give me/us informations about what you do in Galen in the decision support domain - and what kind of problems were faced by ORCA ? I am currently looking for an open source blackboard (C++ if possible) - does the blackboard approach seem good ? do you know some piece of code that could allow me not to start from scratch (and contribute to a project) ? Regards, Philippe PS : I would really like to see Gallen as a material I can use to go further in my projects - maybe you can help.
