pgabriel76
commented
3 years ago Here is what I would argue:
"Modality" as the term is commonly used does not refer to a single universal. i.e. it is not really a definable class and doesn't belong anywhere in the ontology. In the "minimum data elements" proposed so far by the ASTRO white paper and by Chuck's recent BDSC work, I think it refers to a composite of two distinct (although related) entities:
- The source of ionizing radiation (e.g. proton, photon, electron, isotope)
- The delivery mechanism, described at a broad categorical level (e.g. external beam radiotherapy, high dose rate brachytherapy, low dose rate brachytherapy, systemically ingested radiopharmaceuticals, etc.)
The composite nature of the terms included as permissible values for the "modality" data element in both the ASTRO and BDSC specs makes this clear -- for example, "EBRT photon."
I think "technique" is used to specify further details, and also seems to involve a combination of at least two more atomic concepts:
- The method used to develop the plan for how the radiation is to be delivered (e.g. inverse planning)
- The delivery mechanism, described at a more detailed level than in #2 above (e.g. intensity modulation, volume modulation, movement of the gantry in a continuous arc, placement of a radioactive source in an interstitial or intracavitary space, etc.).
And BDSC has also proposed a third data element -- "delivery device" -- that provides yet another piece of information about how the radiation is actually delivered. i.e. it describes the physical technology that is employed to realize the delivery method.
So, my proposal: I think we should put our stake in the ground and say that modality and technique are not classes that belong in the ontology. They can be data elements in a data dictionary -- there is no problem with that, per se. It is just that the permissible values for the data elements have to be "post-coordinated" terms that combine more than one conceptual entity into a single shorthand expression -- e.g. "IMRT," which is handy shorthand for: 1) photon radiation source, 2) generated external to the body and delivered as a beam of energy, 3) the plan for delivery is developed using optimization algorithms constrained by prespecified dose objectives, and 4) the delivery involves modulating both the intensity and shape of the radiation beam.
I think we should work to better define all the truly atomic entities outlined above. Radiation source is fairly straightforward and we basically have it modeled already -- it is an independent continuant.
I'm not sure we've really captured the key aspects of planning and delivery yet. We have defined a radiation therapy procedure, which is a type of planned process that involves many different sub-processes. I think we could go deeper and define all of those different sub-processes, down to the level of things like movement of collimator leaves, but I'm not sure that would really give us what we want and we can all feel is still missing.
Is it possible that we are missing a different type of entity here? Not a planned process, not a material entity, and not an information artifact. I am thinking that perhaps we are dealing with some "specifically-dependent continuants" (in BFO-speak) that we have yet to define. From the Arp book, a specifically-dependent continuant depends on another, more concrete entity for its existence. Examples include qualities (e.g. the color of a tomato, the mass of an object), dispositions (which is how diseases are modeled), roles (e.g. the role of a doctor), and functions (e.g. the function of the heart to pump blood). I could see modeling something like the concept of "brachytherapy" in this way. On the one hand, brachytherapy could be a type of planned process, but I think that might miss something important. I think brachytherapy can be seen as a sort of quality of another entity that is more concrete -- a radiation therapy procedure. It seems like it would allow us to say "this radiation therapy procedure has the quality of being a brachytherapy type of procedure." Which I think is ultimately one of the main things we want to be able to do.
But my head is now spinning and it's getting late on the East Coast, so I'm going to stop here and hope that Jonathan can pick this up and set me straight or guide us somewhere more helpful!
Bayesianworld
It doesn't appear possible to comment on a wiki page, so I'm adding comments here to Mark's post here: https://github.com/aapm-bdsc-ontology-tg/radont/wiki/modality-and-technique
Mark, I see where you are going with this, but I don't think it works to model 'modality' and 'technique' as information artifacts and then say with axioms that they have subatomic particles and devices and such as their parts. An information artifact is a piece of information -- e.g. a document that describes something, or a value recorded in a database field. I don't think 'modality' -- as in the thing itself, whatever it is -- is an information artifact. We could use information artifact to model what someone wrote down that the modality was in their end of treatment summary document, or something like that. And maybe that is all we can do, but I think that 1) that would be of little value, and 2) then we definitely couldn't say that concept has parts that include subatomic particles and machines and such.
Logging this thought and then will continue with some alternative ideas...