| Table of Contents |
|---|
General discussions - anatomy
...
| Title | Content | Created By | Date Created | Forum ID | ID | Project ID | Reply To ID | Status | Topic ID |
|---|---|---|---|---|---|---|---|---|---|
| rPartOf bones; bone marrow | I'm reviewing with the use case of fractures in mind. In our paper on the fracture ontology, we said that fractures occurred in bone regions, which were regional parts of bone; bone fractures can not occur in the periosteum, bone marrow, bone vasculature, etc, as Rector has pointed out. Bone marrow is problematic as it is currently a subclass of bone zone, which makes it a regional part of bone. It is also that way in FMA. I think the bone marrow should be at most a constitutional part of bone organ and not a zone, else I would anticipate unintended inferences elsewhere. Cavity of bone marrow is currently not included in the hierarchy, but maybe should be treated like lumen of the GI tract. Lumens are cavities that are PartOf organs; the lumen of the GI tract is a cPart of the tract of the stomach and intestines. | hgoldberg | Wed Oct 16 15:26:00 Z 2013 | forum1410 | post9093 | proj1113 | A | topc6197 | |
| Re: rPartOf bones; bone marrow | The difficulty with fractures being located in (any part of) a bone region is that we would have to say that periosteum and articular cartilage aren't part of the region. But then there has to be some other way to associate the periosteum and articular cartilage with the bone. (if not 'part of', then what?) Another way to do it is to allow a fracture to be defined according to any bone region or regional part, but not a cPartOf the bone region. If periosteum and articular cartilage and bone marrow would be cPartOf the region, the specialization of the fracture site(s) could not go down to the level of those parts. Finally you could add a "necessarily-also" statement in fracture disorders that says the fracture is necessarily also part of the bone tissue of the region. Anyway that is one way to approach it. | kspackman | Wed Oct 16 17:01:44 Z 2013 | forum1410 | post9095 | proj1113 | post9093 | A | topc6197 |
| Re: rPartOf bones; bone marrow | One more note just to clarify that I am agreeing that bone marrow should not be a zone of bone organ, but rather a constitutional part. Also agree that we could add a bone marrow cavity as a space - i.e. an immaterial anatomical entity. We already have 'subperiosteal space' as an example. I think the bone marrow space could be regarded as an anatomical compartment space (a space demarcated by bona fide boundaries), although the internal surface of the bone is quite irregular, and there are bony trabeculae that criss-cross the space. Still .. I think it works. | kspackman | Wed Oct 16 17:09:16 Z 2013 | forum1410 | post9096 | proj1113 | post9093 | A | topc6197 |
| Re: rPartOf bones; bone marrow | > Another way to do it is to allow a fracture to be defined according to any > bone region or regional part, but not a cPartOf the bone region. If periosteum > and articular cartilage and bone marrow would be cPartOf the region, the > specialization of the fracture site(s) could not go down to the level of those > parts. > Yes, this is what I am saying--the finding site for fractures is restricted to bone regions. In the fracture paper, bone regions are rPartOf bone organ and rPartOf is reflexive, so that bone regions are the bone itself or subregions. Bone regions can certainly have constitutional parts, but these won't be admitted as fracture sites. An issue with PartOf and AllorPartOf is that they will admit cPartOf, rPartOf, and sPartOf, which may have unintended inferences. I've attached the fracture paper for reference, it also includes examples of articular bone regions and articular fractures. This pattern should work for entities such as cellulitis and vascular diseases. Need to look for any examples where this pattern will break expected inferences. | hgoldberg | Thu Oct 17 01:31:19 Z 2013 | forum1410 | post9097 | proj1113 | post9095 | A | topc6197 |
...
| Title | Content | Created By | Date Created | Forum ID | ID | Project ID | Reply To ID | Status | Topic ID |
|---|---|---|---|---|---|---|---|---|---|
| containment | I've added a "contained-in" object property to the alpha draft. This allows queries such as (PartOf some (containedIn some X)). Example of the orbit and contents of the orbit - screen shot attached. | kspackman | Fri Apr 08 19:28:27 Z 2011 | forum1410 | post4073 | proj1113 | A | topc2959 | |
| Re: containment | I'd agree that containment is a useful attribute to have in an anatomical model, but I'm worried that it can be a slippery slope towards still greater complexity. I'm not sure that the point on that slope represented in this draft model is the best place to stop; I suspect we need either more, or nothing. To begin with, there's the issue that some containedIn implies partOf, but not all. So, for example, the stuff contained within the space defined by a cyst is also part of that cyst (at least that's what clinicians would probably expect) whereas the stuff contained within the space defined by a blood vessel isn't part of the vessel. A way to achieve this is to have two subflavours of containedIn: one that is also a subtype of PartOf, and one that isn't (though note this would also be moving the SNOMED attribute hierarchy from a monohierarchy to a polyhierarchy). But then you have to know which flavour to use when and why (and, more problematically, so does anybody writing postcoordinated EPR or query expressions). Then there's the case for a property chain axiom to achieve subsumptions of the general pattern: A and containedIn some (B and PartOf some C) -> A and containedIn some C (ie if I'm contained in something which is part of a larger structure, then I'm presumably contained in the larger structure also). A (somewhat dense) 2005 paper by Bittner and Donnelly expands on the issue: Bittner, T. and Donnelly, M. (2005). Computational ontologies of parthood, componenthood, and containment, In L. Kaelbling (ed.): Proceedings of the Nineteenth International Joint Conference on Artificial Intelligence, 382-387 (available by FTP from here: http://www.acsu.buffalo.edu/~bittner3/ParthoodCompositionContainment.pdf) | jrogers | Thu Apr 14 08:30:18 Z 2011 | forum1410 | post4094 | proj1113 | post4073 | A | topc2959 |
| Re: containment | The Bittner/Donnelly paper is dense, I agree. I think we can get value from a containment relation that has a domain of anatomical space, and has a range of anatomical structure, and where the space is proper-part-of the structure that surrounds/defines the space, but the contents of the space are not proper-part-of it. Think orbit-proper (the orbital bone complex and walls of the orbit) defining the orbital space, which contains the globe, extraocular muscles, etc. Or think the abdominal cavity. FMA uses "content of abdomen" as a primitive grouper, so perhaps we could go with the suggestion to not have a contained-in relation but just use primitive supertypes to express the idea that something is contents of a space. But I tend to like moving incrementally in the direction of using attributes instead of primitive supertypes for these purposes. Caution about the multiple possible interpretations of "contained in" duly noted though. | kspackman | Thu Apr 21 19:23:27 Z 2011 | forum1410 | post4120 | proj1113 | post4094 | A | topc2959 |
...
| Title | Content | Created By | Date Created | Forum ID | ID | Project ID | Reply To ID | Status | Topic ID |
|---|---|---|---|---|---|---|---|---|---|
| Representation of parts of transplanted organs | Not really a new question, just relevant given some recent UK requests... What is the likely SNOMED CT anatomy chapter mechanism for support arbitrarily precise representation of parts of transplanted organs? In the UK we've been trying to regularize the terms used in a number of 'disorder of vein of transplanted structure' - many of which routinely carry domain-specific terms without reference to the 'organ itself' (e.g. obstruction of transplant vein (renal is implied)). This analysis has thrown a bit of light on the existing 'disorder of transplanted artery/vein' content, which is generally modelled with a finding site values of 68273001 | Structure of transplanted artery & 49622001 | Structure of transplanted vein. Superficially this is OK, but treats vessel transplants as independent of their associated organ. We therefore have 15842009 | Thrombosis of renal vein (disorder) is a kind of renal vascular disorder but 234046009 | Transplant renal vein thrombosis (disorder) is just a 234045008 | Thrombosis of transplanted vein (disorder) - no mention of being a kidney disorder [ultimately disconnected from transplants in general due to the inclusion of Associated with = Transplantation (procedure) in the definition of 429490004 | Disorder affecting transplanted structure (disorder)]. Whatever, I am wondering whether there is value in agreeing a convention which would allow the formal defining of concepts know to refer to the vessels of specific transplanted organs in a suitably explicit way? It may well need to be a fractal introduction of new primitive 'vein of transplanted kidney' concepts (like the Matryoshka representation of fetal substructures), but I'm happy to hear other suggestions. I note the principles in the editorial guide (6.5.1.1.8) and the available partonomy in the owl data (including transplanted structures carrying an "allOrPartOf some transplant"), but I can't think of an elegant compositional mechanism for representing 'vein of transplanted kidney' by exploiting this partonomy. Kind regards Ed | edcheetham | Mon Jul 02 11:38:06 Z 2012 | forum1410 | post6947 | proj1113 | A | topc4772 | |
| RE: Representation of parts of transplanted organs | Since the usual anatomy concepts are considered canonical in their parts and locations, we can't properly create subtypes (subclasses). However, we could have a separate hierarchy that references the canonical concepts via attributes "derived-from" and "located-in". So, for example, a saphenous vein graft to the left coronary artery could be modeled as: Isa graft Derived-from saphenous-vein Located-in left-coronary-artery We aren't currently using "derived-from" or "located-in", but they are attributes that have been defined and used in Biotop and other formal ontologies. Some further design and testing is obviously necessary, and we might want to consider other attributes as well. This is just the first stab at a general direction of travel for solving this issue. | kspackman | Mon Jul 02 20:13:39 Z 2012 | forum1410 | post6952 | proj1113 | post6947 | A | topc4772 |
| Re: RE: Representation of parts of transplanted organs | Thanks Kent. Not quite the answer I was expecting... I can see how the "derived-from" or "located-in" attributes would result in a more 'honest' representation for the transplant/graft structures which are 'playing a role' of a canonical structure. The example you cite is a good one in this sense - A grafted saphenous vein doesn't magically 'become' a coronary artery after grafting and should not imply all its properties. However my question related to the parts of a transplanted structure. The veins and arteries of a transplanted kidney or liver are just that - veins and arteries of a transplanted kidney or liver. What I hoping we could have is common retrieval behaviour for part-whole relationships regardless of whether the 'whole' is transplanted or not. In the example I gave above we currently have different answer to the question 'is this a kind of renal vascular disorder?' if we ask it of a 'canonical kidney' or a 'transplanted kidney'. This is not the case for 'kidneys as a whole' [admittedly there are some exceptions because of the 'Associated with = Transplantation (procedure)' issue in the current data] where 'disorder of a transplanted kidney' implies 'disorder of kidney' (because 'transplanted kidney' implies 'kidney'). Kind regards Ed | edcheetham | Sat Jul 07 06:54:32 Z 2012 | forum1410 | post6977 | proj1113 | post6952 | A | topc4772 |
...
| Title | Content | Created By | Date Created | Forum ID | ID | Project ID | Reply To ID | Status | Topic ID |
|---|---|---|---|---|---|---|---|---|---|
| zones | I've seen zones in FMA, but what is the use case for a zone versus a region? Is there a reason that we don't want to specify all 'zone of X' as fully-defined versus primitive? Howard | hgoldberg | Thu Oct 10 17:57:26 Z 2013 | forum1410 | post9074 | proj1113 | A | topc6186 | |
| Re: zones | FMA defines an organ zone (or zone of organ) as: Organ region with one or more floating fiat boundaries. Examples: apical zone of lung, superior pole of kidney, gingiva, apex of prostate, ascending colon proper. The use case is to separate regions of bone that "float", i.e. that depend on the bone's relationship to other structures, from those that are fixed based on the structure of the bone itself. Actually the practical use case of identifying organ zones, distinct from organ regions, is quite limited. The use would have to be something related to the exact boundary of a region or zone, and the fact that zone boundaries can't be determined in a fixed way. Thinking about it, I can't really come up with a compelling clinical use. So maybe we can collapse zone of bone into region of bone. But it isn't possible to fully define, say, "head of fibula", without having a structure that delineates the head from the rest of the bone, and a relationship that relates it to the whole bone. So zone of bone organ and its subtypes will tend to be primitive. | kspackman | Fri Oct 11 03:06:49 Z 2013 | forum1410 | post9075 | proj1113 | post9074 | A | topc6186 |
| Re: zones | > But it isn't possible to fully define, say, "head of fibula", without having a > structure that delineates the head from the rest of the bone, and a > relationship that relates it to the whole bone. So zone of bone organ and its > subtypes will tend to be primitive. Yes, here I meant the 'zone of *' direct subclasses of 'zone of organ', e.g., 'zone of prostate', 'zone of lung', etc. These could be fully-defined; their children would obviously be primitive. | hgoldberg | Tue Oct 15 18:34:33 Z 2013 | forum1410 | post9083 | proj1113 | post9075 | A | topc6186 |
...
| Title | Content | Created By | Date Created | Forum ID | ID | Project ID | Reply To ID | Status | Topic ID |
|---|---|---|---|---|---|---|---|---|---|
| Tooling Issues : which Protege and reasoner? | Hi Kent, Which versions of Protege and - probably more importantly - which reasoner do you recommend using for this? I've loaded the v2 release into Protege 4.1.0 build 213 under an 8Gb 64-bit platform, and then tried classifying under Pellet, Fact and Hermit. But classification progress using any of the three doesn't look like its a five-minute process. Given that there can be very different classification times over the same model depending on which reasoner you select, it would be helpful to know the optimum reasoner(s) to use with this draft model, and some idea of how long it should take to classify. | jrogers | Tue Apr 12 10:31:19 Z 2011 | forum1410 | post4088 | proj1113 | A | topc2964 | |
| Re: Tooling Issues : which Protege and reasoner? | I use FaCT and it takes about 30 minutes on a Macbook Pro with 8GB of memory and an SSD. There are, after all, over 36,000 sufficiently defined concepts. So I'm not surprised it takes a long time. I haven't experimented with the others, except to say they also seem to require some significant time - but if you find a fast one let me know! | kspackman | Tue Apr 12 13:42:57 Z 2011 | forum1410 | post4089 | proj1113 | post4088 | A | topc2964 |
| Re: Tooling Issues : which Protege and reasoner? | I also get classification using FACT++ in 28.6 minutes, using a 3.33 GHz Core i5 PC with 8Gb memory and 64-bit Java on Windows 7 But HermiT probably never terminates or, at least, appeared to have made zero progress after an hour. Pellet runs out of memory (within a 5.5Gb Java image) after about 10 minutes. The Snorocket 1.3.4 plugin crashes Protege 4.1 before you've even loaded the ontology (it may work on Protege 4.0). So the choice is simple! | jrogers | Wed Apr 13 07:54:29 Z 2011 | forum1410 | post4091 | proj1113 | post4089 | A | topc2964 |
...
| Title | Content | Created By | Date Created | Forum ID | ID | Project ID | Reply To ID | Status | Topic ID |
|---|---|---|---|---|---|---|---|---|---|
| "Body cavity" content | In the UK we are undertaking a review of the ancestry of content in our draft 'renal finding' subset (the vast majority of which are renal disorders). One ancestor we keep encountering is "399902003 | Disorder of body cavity (disorder)". This appears wrong when considering disorders of a solid organ such as the kidney, and would seem to stem from mixing up subclass and containment relationships in the current data. FMA is much stricter about its "anatomical/body cavity" content (e.g. FMAID 85006), requiring all subclasses to be massless, and I was rather hoping that this stringency would be reflected in the alpha data. However this doesn't seem to be the case. Admittedly kidney structures are no longer classified as kinds of body cavity structure, but I do wonder if this is more because of the separate existence of 'retroperitoneal space' and 'retroperitoneum' in the data. In addition many solid structures (notably thoracic structures such as 'heart' and pelvic structures such as 'ovary') are still classified as body cavity structures - but I am ashamed to say that despite scrutinising Elk's inference explanations I'm at a loss to know what modelling pattern is causing this behaviour. First off, therefore, it would be good to know whether our concern (regarding the renal content is correct, and secondly it would be good to explore how this should be addressed in the alpha data. Kind regards Ed | edcheetham | Wed Oct 30 12:57:25 Z 2013 | forum1410 | post9158 | proj1113 | A | topc6225 | |
| Re: "Body cavity" content | Ed, Interesting question. I'm looking at the previous alpha release of anatomy. There are 9 explanations from protégé for why Heart is a descendant class of (S) body cavity, here is 1: heart SubClassOf PartOf some 'heart and pericardium' PartOf SubPropertyOf allOrPartOf 'heart and pericardium' SubClassOf sPartOf some 'intrathoracic cardiovascular system' sPartOf SubPropertyOf PartOf 'intrathoracic cardiovascular system' SubClassOf PartOf some 'thoracic cavity' 'thoracic cavity' SubClassOf rPartOf some 'body cavity' rPartOf SubPropertyOf PartOf allOrPartOf some 'body cavity' SubClassOf '(S) body cavity' allOrPartOf o PartOf SubPropertyOf PartOf There appear to be 2 offending axioms. From the above explanation, the offending axiom is 'intrathoracic cardiovascular system' SubClassOf PartOf some 'thoracic cavity'. Heart is also modeled to be 'PartOf some middle mediastinum'. When I update both these axioms from part of to contained in, heart is no longer a subclass of (S) body cavity. There are a significant number of anatomic structures that are getting classified under (S) body cavity, so it looks like there is a bit of cleanup to remodel 'Things partof body cavity' to 'Things containedin body cavity'. We should think more whether we can use some well-placed disjointness axioms in a debugging mode to get the classifier identify these issues; a body organ should be a priori disjoint from a body cavity. Howard Howard S. Goldberg, MD Sr Corp Mgr, Enterprise Informatics Infrastructure Services Partners Healthcare System 93 Worcester Rd. Wellesley, MA 02481 On 10/30/13 8:57 AM, "Ed Cheetham (ihtsdo UK)" <ed.cheetham@hscic.gov.uk> wrote: >In the UK we are undertaking a review of the ancestry of content in our >draft 'renal finding' subset (the vast majority of which are renal >disorders). > >One ancestor we keep encountering is "399902003 | Disorder of body cavity >(disorder)". This appears wrong when considering disorders of a solid >organ such as the kidney, and would seem to stem from mixing up subclass >and containment relationships in the current data. > >FMA is much stricter about its "anatomical/body cavity" content (e.g. >FMAID 85006), requiring all subclasses to be massless, and I was rather >hoping that this stringency would be reflected in the alpha data. However >this doesn't seem to be the case. Admittedly kidney structures are no >longer classified as kinds of body cavity structure, but I do wonder if >this is more because of the separate existence of 'retroperitoneal space' >and 'retroperitoneum' in the data. > >In addition many solid structures (notably thoracic structures such as >'heart' and pelvic structures such as 'ovary') are still classified as >body cavity structures - but I am ashamed to say that despite >scrutinising Elk's inference explanations I'm at a loss to know what >modelling pattern is causing this behaviour. > >First off, therefore, it would be good to know whether our concern >(regarding the renal content is correct, and secondly it would be good to >explore how this should be addressed in the alpha data. > >Kind regards > >Ed > > > >_______________________________________________ >IHTSDO anatomy redesign project group discussion list > >General discussions - anatomy >https://csfe.aceworkspace.net/sf/go/post9158 >To cancel your subscription to this discussion, please e-mail >discuss-anatomy_model_project-unsubscribe@csfe.aceworkspace.net The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail. | hgoldberg | Wed Oct 30 15:15:16 Z 2013 | forum1410 | post9159 | proj1113 | post9158 | A | topc6225 |
| Re: "Body cavity" content | Thanks Howard Interestingly I had tried something similar myself, but I think I only modified the first axiom you mention. Consequently heart was still returned as a kind of '(S) body cavity' and I gave up! I've just made both changes you suggest - with the same result, however I note that testing '(S) body cavity' and '(S) heart' in the DL Query pane suggests that there are still some "heart structures" such as coronary arteries returned as subtypes of body cavity. The disjointness suggestion looks very sensible - interested to hear suggestions about how to do this (both in terms of what to apply, where to apply it and how to make sense of the result!). Kind regards Ed | edcheetham | Wed Oct 30 15:57:33 Z 2013 | forum1410 | post9160 | proj1113 | post9159 | A | topc6225 |
| Re: "Body cavity" content | Yes, there needs to be some clean-up so that immaterial entities can still be part-of material entities (a lumen is considered a constitutional part of a vessel, for example), but not the converse, i.e. no material entity can be a part of a material entity. This is currently the case in the inner ear, and, as Ed has pointed out, with some structures like the body cavity. Immaterial and material anatomical entities are disjoint already (with some effort required to get them to that point, I might add). But their sub-parts, as pointed out above, are not necessarily all disjoint. '(S) material anatomical entity' can have immaterial subparts. '(S) immaterial anatomical entity' should _not_ have material subparts. Coming soon to an alpha revision near you.... --Kent On Wed, Oct 30, 2013 at 8:14 AM, Goldberg, Howard Seth,M.D. <HGOLDBERG@partners.org> wrote: > > Ed, > Interesting question. I'm looking at the previous alpha release of > anatomy. There are 9 explanations from protégé for why Heart is a > descendant class of (S) body cavity, here is 1: > > heart SubClassOf PartOf some 'heart and pericardium' > PartOf SubPropertyOf allOrPartOf > 'heart and pericardium' SubClassOf sPartOf some 'intrathoracic > cardiovascular system' > sPartOf SubPropertyOf PartOf > 'intrathoracic cardiovascular system' SubClassOf PartOf some 'thoracic > cavity' > 'thoracic cavity' SubClassOf rPartOf some 'body cavity' > rPartOf SubPropertyOf PartOf > allOrPartOf some 'body cavity' SubClassOf '(S) body cavity' > allOrPartOf o PartOf SubPropertyOf PartOf > > There appear to be 2 offending axioms. From the above explanation, the > offending axiom is 'intrathoracic cardiovascular system' SubClassOf PartOf > some 'thoracic cavity'. Heart is also modeled to be 'PartOf some middle > mediastinum'. When I update both these axioms from part of to contained > in, heart is no longer a subclass of (S) body cavity. > > There are a significant number of anatomic structures that are getting > classified under (S) body cavity, so it looks like there is a bit of > cleanup to remodel 'Things partof body cavity' to 'Things containedin body > cavity'. We should think more whether we can use some well-placed > disjointness axioms in a debugging mode to get the classifier identify > these issues; a body organ should be a priori disjoint from a body cavity. > > > Howard > > > Howard S. Goldberg, MD > Sr Corp Mgr, Enterprise Informatics Infrastructure Services > Partners Healthcare System > 93 Worcester Rd. > Wellesley, MA 02481 > > > > > > > On 10/30/13 8:57 AM, "Ed Cheetham (ihtsdo UK)" <ed.cheetham@hscic.gov.uk> > wrote: > >>In the UK we are undertaking a review of the ancestry of content in our >>draft 'renal finding' subset (the vast majority of which are renal >>disorders). >> >>One ancestor we keep encountering is "399902003 | Disorder of body cavity >>(disorder)". This appears wrong when considering disorders of a solid >>organ such as the kidney, and would seem to stem from mixing up subclass >>and containment relationships in the current data. >> >>FMA is much stricter about its "anatomical/body cavity" content (e.g. >>FMAID 85006), requiring all subclasses to be massless, and I was rather >>hoping that this stringency would be reflected in the alpha data. However >>this doesn't seem to be the case. Admittedly kidney structures are no >>longer classified as kinds of body cavity structure, but I do wonder if >>this is more because of the separate existence of 'retroperitoneal space' >>and 'retroperitoneum' in the data. >> >>In addition many solid structures (notably thoracic structures such as >>'heart' and pelvic structures such as 'ovary') are still classified as >>body cavity structures - but I am ashamed to say that despite >>scrutinising Elk's inference explanations I'm at a loss to know what >>modelling pattern is causing this behaviour. >> >>First off, therefore, it would be good to know whether our concern >>(regarding the renal content is correct, and secondly it would be good to >>explore how this should be addressed in the alpha data. >> >>Kind regards >> >>Ed >> >> >> >>_______________________________________________ >>IHTSDO anatomy redesign project group discussion list >> >>General discussions - anatomy >>https://csfe.aceworkspace.net/sf/go/post9158 >>To cancel your subscription to this discussion, please e-mail >>discuss-anatomy_model_project-unsubscribe@csfe.aceworkspace.net > > > > The information in this e-mail is intended only for the person to whom it is > addressed. If you believe this e-mail was sent to you in error and the e-mail > contains patient information, please contact the Partners Compliance HelpLine at > http://www.partners.org/complianceline . If the e-mail was sent to you in error > but does not contain patient information, please contact the sender and properly > dispose of the e-mail. > > > _______________________________________________ > General discussions - anatomy > https://csfe.aceworkspace.net/sf/go/post9159 | kspackman | Wed Oct 30 16:44:57 Z 2013 | forum1410 | post9161 | proj1113 | post9159 | A | topc6225 |
| Re: "Body cavity" content | I tried the experiment to make 'material anatomic entity' disjoint from 'immaterial anatomic entity'. Of course, that changes the DL expressivity from ALE to S, out of the domain of ELK. Both FACT++ and Hermit fail and complain about a cyclic dependency in the partof object property hierarchy. Howard S. Goldberg, MD Sr Corp Mgr, Enterprise Informatics Infrastructure Services Partners Healthcare System 93 Worcester Rd. Wellesley, MA 02481 On 10/30/13 11:57 AM, "Ed Cheetham (ihtsdo UK)" <ed.cheetham@hscic.gov.uk> wrote: >Thanks Howard > >Interestingly I had tried something similar myself, but I think I only >modified the first axiom you mention. Consequently heart was still >returned as a kind of '(S) body cavity' and I gave up! I've just made >both changes you suggest - with the same result, however I note that >testing '(S) body cavity' and '(S) heart' in the DL Query pane suggests >that there are still some "heart structures" such as coronary arteries >returned as subtypes of body cavity. > >The disjointness suggestion looks very sensible - interested to hear >suggestions about how to do this (both in terms of what to apply, where >to apply it and how to make sense of the result!). > >Kind regards > >Ed > > > >_______________________________________________ >IHTSDO anatomy redesign project group discussion list > >General discussions - anatomy >https://csfe.aceworkspace.net/sf/go/post9160 >To cancel your subscription to this discussion, please e-mail >discuss-anatomy_model_project-unsubscribe@csfe.aceworkspace.net > The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail. | hgoldberg | Wed Oct 30 16:56:43 Z 2013 | forum1410 | post9162 | proj1113 | post9160 | A | topc6225 |
| RE: "Body cavity" content | In the DL query tab you can run the query for the intersection, 'material anatomical entity' and 'immaterial anatomical entity', which returns zero subclasses. Also I made them disjoint via an edit in Protégé (added a "disjoint with" axiom) and ELK handles it just fine. The version 0.3.1 of ELK (and later) supports a fragment of OWL EL that corresponds to the DL EL+⊥ which extends EL with bottom (⊥) and role inclusion axioms. And you get disjointness from bottom. (i.e. X and Y -> bottom is the way to say X and Y are disjoint). --Kent -----Original Message----- From: Goldberg, Howard Seth,M.D. [mailto:HGOLDBERG@PARTNERS.ORG] Sent: Wednesday, October 30, 2013 09:57 To: discuss-anatomy_model_project@csfe.aceworkspace.net Subject: Re: "Body cavity" content I tried the experiment to make 'material anatomic entity' disjoint from 'immaterial anatomic entity'. Of course, that changes the DL expressivity from ALE to S, out of the domain of ELK. Both FACT++ and Hermit fail and complain about a cyclic dependency in the partof object property hierarchy. Howard S. Goldberg, MD Sr Corp Mgr, Enterprise Informatics Infrastructure Services Partners Healthcare System 93 Worcester Rd. Wellesley, MA 02481 On 10/30/13 11:57 AM, "Ed Cheetham (ihtsdo UK)" <ed.cheetham@hscic.gov.uk> wrote: >Thanks Howard > >Interestingly I had tried something similar myself, but I think I only >modified the first axiom you mention. Consequently heart was still >returned as a kind of '(S) body cavity' and I gave up! I've just made >both changes you suggest - with the same result, however I note that >testing '(S) body cavity' and '(S) heart' in the DL Query pane suggests >that there are still some "heart structures" such as coronary arteries >returned as subtypes of body cavity. > >The disjointness suggestion looks very sensible - interested to hear >suggestions about how to do this (both in terms of what to apply, where >to apply it and how to make sense of the result!). > >Kind regards > >Ed > > > >_______________________________________________ >IHTSDO anatomy redesign project group discussion list > >General discussions - anatomy >https://csfe.aceworkspace.net/sf/go/post9160 >To cancel your subscription to this discussion, please e-mail >discuss-anatomy_model_project-unsubscribe@csfe.aceworkspace.net > The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail. _______________________________________________ General discussions - anatomy https://csfe.aceworkspace.net/sf/go/post9162 | kspackman | Wed Oct 30 19:35:15 Z 2013 | forum1410 | post9163 | proj1113 | post9160 | A | topc6225 |
| Re: "Body cavity" content | Thanks Kent, I thought that ELK was just ignoring the disjoint axiom. When you make '(S) immaterial anatomic entity' disjoint from 'material anatomic entity', then ELK will show that a very large number of concepts are inconsistent. Good use of the classifier to make sure that all concepts will be properly remodeled. Howard S. Goldberg, MD Sr Corp Mgr, Enterprise Informatics Infrastructure Services Partners Healthcare System 93 Worcester Rd. Wellesley, MA 02481 On 10/30/13 3:35 PM, "Kent Spackman" <ksp@ihtsdo.org> wrote: >In the DL query tab you can run the query for the intersection, >'material anatomical entity' and 'immaterial anatomical entity', which >returns zero subclasses. >Also I made them disjoint via an edit in Protégé (added a "disjoint >with" axiom) and ELK handles it just fine. >The version 0.3.1 of ELK (and later) supports a fragment of OWL EL that >corresponds to the DL EL+⊥ which extends EL with bottom (⊥) and role >inclusion axioms. >And you get disjointness from bottom. (i.e. X and Y -> bottom is the way >to say X and Y are disjoint). > >--Kent > >-----Original Message----- >From: Goldberg, Howard Seth,M.D. [mailto:HGOLDBERG@PARTNERS.ORG] >Sent: Wednesday, October 30, 2013 09:57 >To: discuss-anatomy_model_project@csfe.aceworkspace.net >Subject: Re: "Body cavity" content > > >I tried the experiment to make 'material anatomic entity' disjoint from >'immaterial anatomic entity'. Of course, that changes the DL expressivity >from ALE to S, out of the domain of ELK. Both FACT++ and Hermit fail and >complain about a cyclic dependency in the partof object property >hierarchy. > >Howard S. Goldberg, MD >Sr Corp Mgr, Enterprise Informatics Infrastructure Services Partners >Healthcare System >93 Worcester Rd. >Wellesley, MA 02481 > > > > > > >On 10/30/13 11:57 AM, "Ed Cheetham (ihtsdo UK)" <ed.cheetham@hscic.gov.uk> >wrote: > >>Thanks Howard >> >>Interestingly I had tried something similar myself, but I think I only >>modified the first axiom you mention. Consequently heart was still >>returned as a kind of '(S) body cavity' and I gave up! I've just made >>both changes you suggest - with the same result, however I note that >>testing '(S) body cavity' and '(S) heart' in the DL Query pane suggests >>that there are still some "heart structures" such as coronary arteries >>returned as subtypes of body cavity. >> >>The disjointness suggestion looks very sensible - interested to hear >>suggestions about how to do this (both in terms of what to apply, where >>to apply it and how to make sense of the result!). >> >>Kind regards >> >>Ed >> >> >> >>_______________________________________________ >>IHTSDO anatomy redesign project group discussion list >> >>General discussions - anatomy >>https://csfe.aceworkspace.net/sf/go/post9160 >>To cancel your subscription to this discussion, please e-mail >>discuss-anatomy_model_project-unsubscribe@csfe.aceworkspace.net >> > > > >The information in this e-mail is intended only for the person to whom it >is addressed. If you believe this e-mail was sent to you in error and the >e-mail contains patient information, please contact the Partners >Compliance HelpLine at http://www.partners.org/complianceline . If the >e-mail was sent to you in error but does not contain patient information, >please contact the sender and properly dispose of the e-mail. > > >_______________________________________________ >General discussions - anatomy >https://csfe.aceworkspace.net/sf/go/post9162 > > > > >_______________________________________________ >IHTSDO anatomy redesign project group discussion list > >General discussions - anatomy >https://csfe.aceworkspace.net/sf/go/post9163 >To cancel your subscription to this discussion, please e-mail >discuss-anatomy_model_project-unsubscribe@csfe.aceworkspace.net | hgoldberg | Thu Oct 31 02:45:51 Z 2013 | forum1410 | post9164 | proj1113 | post9163 | A | topc6225 |
| RE: "Body cavity" content | I've now uploaded a version of the owl file that no longer has anything that is both a material anatomical entity and all-or-part-of an immaterial anatomical entity. I also include a disjointness declaration so that anything made to be a subtype of both will be inconsistent. --Kent -----Original Message----- From: Goldberg, Howard Seth,M.D. [mailto:HGOLDBERG@PARTNERS.ORG] Sent: Wednesday, October 30, 2013 19:46 To: discuss-anatomy_model_project@csfe.aceworkspace.net Subject: Re: "Body cavity" content Thanks Kent, I thought that ELK was just ignoring the disjoint axiom. When you make '(S) immaterial anatomic entity' disjoint from 'material anatomic entity', then ELK will show that a very large number of concepts are inconsistent. Good use of the classifier to make sure that all concepts will be properly remodeled. Howard S. Goldberg, MD Sr Corp Mgr, Enterprise Informatics Infrastructure Services Partners Healthcare System 93 Worcester Rd. Wellesley, MA 02481 On 10/30/13 3:35 PM, "Kent Spackman" <ksp@ihtsdo.org> wrote: >In the DL query tab you can run the query for the intersection, >'material anatomical entity' and 'immaterial anatomical entity', which >returns zero subclasses. >Also I made them disjoint via an edit in Protégé (added a "disjoint >with" axiom) and ELK handles it just fine. >The version 0.3.1 of ELK (and later) supports a fragment of OWL EL that >corresponds to the DL EL+⊥ which extends EL with bottom (⊥) and role >inclusion axioms. >And you get disjointness from bottom. (i.e. X and Y -> bottom is the >way to say X and Y are disjoint). > >--Kent > >-----Original Message----- >From: Goldberg, Howard Seth,M.D. [mailto:HGOLDBERG@PARTNERS.ORG] >Sent: Wednesday, October 30, 2013 09:57 >To: discuss-anatomy_model_project@csfe.aceworkspace.net >Subject: Re: "Body cavity" content > > >I tried the experiment to make 'material anatomic entity' disjoint from >'immaterial anatomic entity'. Of course, that changes the DL >expressivity from ALE to S, out of the domain of ELK. Both FACT++ and >Hermit fail and complain about a cyclic dependency in the partof object >property hierarchy. > >Howard S. Goldberg, MD >Sr Corp Mgr, Enterprise Informatics Infrastructure Services Partners >Healthcare System >93 Worcester Rd. >Wellesley, MA 02481 > > > > > > >On 10/30/13 11:57 AM, "Ed Cheetham (ihtsdo UK)" ><ed.cheetham@hscic.gov.uk> >wrote: > >>Thanks Howard >> >>Interestingly I had tried something similar myself, but I think I only >>modified the first axiom you mention. Consequently heart was still >>returned as a kind of '(S) body cavity' and I gave up! I've just made >>both changes you suggest - with the same result, however I note that >>testing '(S) body cavity' and '(S) heart' in the DL Query pane >>suggests that there are still some "heart structures" such as coronary >>arteries returned as subtypes of body cavity. >> >>The disjointness suggestion looks very sensible - interested to hear >>suggestions about how to do this (both in terms of what to apply, >>where to apply it and how to make sense of the result!). >> >>Kind regards >> >>Ed >> >> >> >>_______________________________________________ >>IHTSDO anatomy redesign project group discussion list >> >>General discussions - anatomy >>https://csfe.aceworkspace.net/sf/go/post9160 >>To cancel your subscription to this discussion, please e-mail >>discuss-anatomy_model_project-unsubscribe@csfe.aceworkspace.net >> > > > >The information in this e-mail is intended only for the person to whom >it is addressed. If you believe this e-mail was sent to you in error >and the e-mail contains patient information, please contact the >Partners Compliance HelpLine at http://www.partners.org/complianceline >. If the e-mail was sent to you in error but does not contain patient >information, please contact the sender and properly dispose of the e-mail. > > >_______________________________________________ >General discussions - anatomy >https://csfe.aceworkspace.net/sf/go/post9162 > > > > >_______________________________________________ >IHTSDO anatomy redesign project group discussion list > >General discussions - anatomy >https://csfe.aceworkspace.net/sf/go/post9163 >To cancel your subscription to this discussion, please e-mail >discuss-anatomy_model_project-unsubscribe@csfe.aceworkspace.net _______________________________________________ General discussions - anatomy https://csfe.aceworkspace.net/sf/go/post9164 | kspackman | Thu Oct 31 06:41:40 Z 2013 | forum1410 | post9165 | proj1113 | post9163 | A | topc6225 |
| Re: RE: "Body cavity" content | Wow, thanks both. I'm not entirely sure what Kent did, but the data is dramatically changed - and as stated there is a much neater set of cavity-type classes beneath '(S) body cavity'. Kind regards Ed | edcheetham | Thu Oct 31 11:55:04 Z 2013 | forum1410 | post9168 | proj1113 | post9165 | A | topc6225 |
...
| Title | Content | Created By | Date Created | Forum ID | ID | Project ID | Reply To ID | Status | Topic ID |
|---|---|---|---|---|---|---|---|---|---|
| allOrPartOf some 'long bone' | querying (Protege / elk classifier) direct subclasses which should return the rParts and cParts of long bones. Returns '(S) bone tissue of long bone' '(S) diaphysis' '(S) epiphysis' '(S) long bone of lower limb' '(S) long bone of upper limb' 'long bone' Would have expected also '(S) metaphysis', which is a definitional issue. The explanation for '(S) long bone of lower limb' is '(S) long bone of lower limb' EquivalentTo allOrPartOf some 'long bone of lower limb' 'long bone of lower limb' SubClassOf 'long bone' and 'bone of lower limb' and (laterality some side) which is correct, but adds noise to the return. Couldn't think of a way around this. | hgoldberg | Tue Oct 15 18:53:32 Z 2013 | forum1410 | post9084 | proj1113 | A | topc6191 | |
| Re: allOrPartOf some 'long bone' | FMA says that the metaphysis is part of the diaphysis. So '(S) metaphysis' is one level lower in the hierarchy and won't show up in the top level there. We can disagree and make the metaphysis a separate region from diaphysis, if that is more useful or clinically correct. Simple to edit. Perhaps not so simple to decide. To get just the parts but not the whole bones, use Query "(PartOf some 'long bone')" I've made a few more edits. FMA says that the ribs, clavicle, metatarsals and metacarpals are also long bones, and I've added them to the long bone hierarchy. Also edited so that all the shafts of long bones are fully defined as "diaphysis and (rPartOf some x)" where x is the bone. Then also defined all the proximal shafts as "'proximal third of bone' and (rPartOf some 'shaft of x')", and likewise for middle and distal. Then the transitive regional part-of will infer that the proximal shaft of x is a regional part of x, (since the shaft of x is a regional part of x, and the proximal shaft is a regional part of the shaft). Let me know if you want to see the revised version - I hesitate to do too frequent updates of the collabnet alpha but could share the interim files another way if you want. | kspackman | Tue Oct 15 22:52:14 Z 2013 | forum1410 | post9086 | proj1113 | post9084 | A | topc6191 |
| Re: allOrPartOf some 'long bone' | > FMA says that the metaphysis is part of the diaphysis. So '(S) metaphysis' is > one level lower in the hierarchy and won't show up in the top level there. We > can disagree and make the metaphysis a separate region from diaphysis, if > that is more useful or clinically correct. Simple to edit. Perhaps not so > simple to decide. Should we acknowledge that Hollinshead (Rosse's text) is a primary source for content in the wiki so users have a source to refer to the origin of these definitions. I cross-reference via the FMA explorer, but I think the semantic consistency, completeness, and correctness of FMA is less than what SCT will need to scale. | hgoldberg | Wed Oct 16 14:42:26 Z 2013 | forum1410 | post9092 | proj1113 | post9086 | A | topc6191 |
...
| Title | Content | Created By | Date Created | Forum ID | ID | Project ID | Reply To ID | Status | Topic ID |
|---|---|---|---|---|---|---|---|---|---|
| Anatomy redesign approach | Dear all I'm not really sure what stage the review of the redesign owl file has reached, so apologies if what follows looks like I'm critiquing "the detail" rather than the principles being demonstrated, but I think they are related. Looking at the background slide deck, and in particular the SEP/modelling & inference diagrams on slides 19-27, it appears that the modeller can assert either IsA or Part_Of relationships between 'entire' classes - a choice resulting in a slightly different set of inferences. My specific search started in the current SNOMED CT data where there is a separation between 69105007 | Carotid artery structure | and most of its constituent parts ('common' is assocated, but 'internal' and 'external' are remote). In the current redesign data 'common' and 'external' are now subtypes, but 'internal' remains remote. However, the associations in the redesign data have been achieved by asserting an IsA relationship between 'entire common carotid artery' and 'entire carotid artery' - and a similar 'IsA' approach has been used to associate 'entire SVC' and 'entire IVC' with 'entire vena cava'. Looking elsewhere I see that 'ascending aorta' and 'descending aorta' are related to 'entire aorta' by 'rPartOf' assertions - an approach which would seem more appropriate for both the earlier vascular examples (the named vessel sections are "parts of" the named whole vessel). Whilst it is clear that there are many 'entireA -> entire B' relationships that are appropriately related by IsA relationships (in particular where the superclass is not a structure in its own right (such as 'aorta' IsA 'large artery'), am I right in suggesting that there is value in revising the above examples from IsA assertions to PartOf (? rPartOf) between the various entire structures? If so, do we need to identify heuristics to help decide which is appropriate in particular circumstances? Kind regards Ed | edcheetham | Tue Sep 17 12:56:56 Z 2013 | forum1410 | post8985 | proj1113 | A | topc6128 | |
| RE: Anatomy redesign approach | We don't need heuristics so much as clarity of the meaning of the terms. What does "carotid artery" actually mean? In other words, what is the anatomical extent of the thing referred to. If "carotid artery" is a synonym of "common carotid artery" (as asserted by FMA, for example), then our current "entire carotid artery" is a duplicate of "entire common carotid artery". If it is a common grouper, meaning "one of the set of arterial trunks: common carotid, external carotid, and internal carotid", then we need to add an isa relationship from the internal carotid (in the redesign) to "carotid artery". There is no "entire carotid artery" anatomically, which the internal, common and external can be part of. The thing they are all part of is a peculiar subtree of the arterial system that includes the common and its two branches. I don't believe that is what most clinicians (or anatomists) would mean by "carotid artery". Likewise there is no "vena cava" for the inferior and superior to be part of. There are two "venae cavae". So is-a is the appropriate relation between "inferior vena cava" and a concept that means "one of the venae cavae". So, in summary, I would agree that there are things that need attention when looking at is-a vs part-of. Based on the examples, it is the meaning of the vascular "grouper" terms that needs attention. | kspackman | Tue Sep 17 17:22:21 Z 2013 | forum1410 | post8987 | proj1113 | post8985 | A | topc6128 |
| RE: Anatomy redesign approach | One more note: The need for an anatomy concept "carotid artery" which subsumes the common, external and internal is derived from grouper level concepts in ICD. Here is what is currently in ICD-11 beta under http://id.who.int/icd/entity/1285223016 Injury of carotid artery Parent(s) Injury of blood vessels at neck level ICD-10 : S15.0 Narrower Terms Carotid artery (common) (external) (internal) * Body Site Carotid artery structure (body structure) The meaning appears to be "injury of one of the following: common, external, or internal carotid artery". | kspackman | Tue Sep 17 18:02:48 Z 2013 | forum1410 | post8988 | proj1113 | post8985 | A | topc6128 |
| Re: RE: Anatomy redesign approach | Thanks Kent. I must confess that I had wrongly thought there was such a thing as a 'carotid artery' (with a common carotid artery part, an internal carotid artery part etc...) but you are quite right. Same goes for "jugular vein" and presumably dozens of other structures - there is no distinct name for the collection of "all jugular veins". By contrast, we do have names for collections such as "cervical spine" - in this case each "numbered entire cervical vertebra" IsA "entire cervical vertebra" and IsPartOf "entire cervical spine". For interest I note that we do already publish a text definition against 'carotid artery structure' which almost perfectly reflects your ICD-11 observation "One of the common carotid, internal carotid, or external carotid arteries". Kind regards Ed | edcheetham | Wed Sep 18 11:11:44 Z 2013 | forum1410 | post8992 | proj1113 | post8988 | A | topc6128 |
...
| Title | Content | Created By | Date Created | Forum ID | ID | Project ID | Reply To ID | Status | Topic ID |
|---|---|---|---|---|---|---|---|---|---|
| FMA relations / laterality | Dear all Does anyone know if the full set of FMA relations is published anywhere? I can find plenty of overview documents as well as subsets of the relations (e.g. in EN 15521, FMA in owl), but I can't find the whole set summarised anywhere. In particular I'm interested in whether FMA has any formal features which would help us increase the sensitivity of detection of *lateralisable* structures. Our SEP-map.txt file doesn't have FMA maps to/from the "... half of X lateral to midsagittal plane..." content but clearly does link to explicitly lateralised structures. I seem to recall discussions which suggested that "lateralisability" wasn't a particularly valuable anatomical notion - just a quirk of symmetry, so I would quite understand if it's not directly represented (but could perhaps be derived from analysis of the named left/right structure pairs). Kind regards Ed | edcheetham | Wed Jul 17 12:36:01 Z 2013 | forum1410 | post8719 | proj1113 | A | topc5974 | |
| Re: FMA relations / laterality | My understanding is that FMA pre-coordinates all left/right versions of a structure, so that anything that is lateralizable will have two subtypes. But there is no relationship that can be used to represent laterality. I think the relationships are documented in the paper by Mejino et al in AMIA 2003 Proceedings pp 450 "Representing Complexity in Part-Whole Relationships within the Foundational Model of Anatomy" | kspackman | Wed Jul 17 23:20:11 Z 2013 | forum1410 | post8720 | proj1113 | post8719 | A | topc5974 |
| Re: FMA relations / laterality | Thanks Kent. Ed | edcheetham | Fri Jul 19 12:01:52 Z 2013 | forum1410 | post8729 | proj1113 | post8720 | A | topc5974 |