The need
Full disclosure: I'm a member of the research group (the Structural Informatics Group at the University of Washington) who carried out a lot of this work, although my work is different from and in parallel to the work done by other members of the group. I will cite a couple of my own papers here, but most of the work I cite will not be research I carried out myself.
You should bear that affiliation in mind when you read this, to evaluate whether--as I should--I have connected the dots for you without positive bias for people I work with, and have a very high regard for.
In Mejino 2004 (Rosse C, Mejino JL Jr. Symbolic modeling of structural relationships in the Foundational Model of Anatomy. In Proceedings, First International Workshop on Formal Biomedical Knowledge Representation (KR-MED 2004), Whistler Mountain, Canada), the authors first review the situation that anatomical knowledge representation is currently in (as of that writing).
The main objective of the terminologies correlated by UMLS is to serve as repositories of terms that can be reused with consistency by a variety of applications.
UMLS is the Unified Medical Language System, developed and maintained by the National Library of Medicine for information-sharing across biomedical disciplines and between individuals and institutions.
In general, most of the current biomedical and educational applications are designed to present hard-coded, didactic information, or they support low-level, look-up functions with no, or at best limited, capabilities for inference. The semantic structure of today's controlled medical terminologies (CMTs) as well as of biomedical ontologies seems adequate for the needs of such contemporary applications. (Mejino 2004)
Inference is the process of reasoning, of developing new information or knowledge based on extending in a sound and valid way what we already know.
Mejino here is contrasting how contemporary computer medical knowledge sources can retrieve answers to questions for which humans have already done the work,
Example: "name the muscles that make up the quadriceps".
(corresponding roughly to the lowest level of Bloom's learning taxonomy--passively recognizing memorized facts. This is the level at which standardized testing, such as the MBLEX or NCETM/NCETMB, evaluate knowledge.)
but are not yet very good at independently figuring out new information that humans have not explicitly put into the knowledge base yet.
Example: "Use the information about the different cell-type lifetimes and replacement rates to figure out an explanation for why cancer patients undergoing radiation often lose their hair and tissue in their esophagus, and explain the reasoning behind your answer".
(corresponding roughly to the higher levels of Bloom's learning taxonomy--application, analysis, synthesis, and evaluation. Standardized testing, such as the MBLEX or NCETM/NCETMB, cannot evaluate knowledge at this level, because it requires original writing from the student, and it requires evaluators who are capable of correctly grading the quality of the student's answers--a very time-intensive process. These are, however, levels of knowledge processing required in clinical practice.)
They contrast the status quo with needs on the horizon.
Next-generation applications, however, will have to incorporate increasing levels of intelligence in order to meet the demands of the evolving environment in education, biomedical research and the practice of the various health professions. Such knowledge-based applications call for the representation of much deeper and richer knowledge than that retrievable from today's CMTs and ontologies. (Mejino 2004)
As they indicate, the current situation is not going to remain sufficient for much longer, given the rapid changes in the fields that need information from these systems. In the very near future, we are going to need software that can provide the user with much more detailed and sound knowledge than this generation of software is able to provide.
Since most of these projects primarily target clinical medicine, they are deficient in basic science concepts necessary to support reasoning. (Mejino 2004)
This point may not be immediately obvious, so let's clarify it.
Star Trek and other popular media depictions aside--with a few exceptions in active research areas--computers are really pretty stupid.
"Garbage in, garbage out" as the saying goes--the quality of the information you get out depends directly on the quality of what you or someone else previously put into the computer.
They're like the Sorcerer's Apprentice, except they're not conscious or self-aware.
The Sorcerer's Apprentice (German: Der Zauberlehrling) is a poem by Goethe, written in 1797. The poem is a ballad in fourteen stanzas.
...
The poem begins as an old sorcerer departs his workshop, leaving his apprentice with chores to perform. Tired of fetching water by pail, the apprentice enchants a broom to do the work for him — using magic in which he is not yet fully trained. The floor is soon awash with water, and the apprentice realizes that he cannot stop the broom because he does not know how.
Not knowing how to control the enchanted broom, the apprentice splits it in two with an axe, but each of the pieces becomes a new broom and takes up a pail and continues fetching water, now at twice the speed. When all seems lost, the old sorcerer returns, quickly breaks the spell and saves the day. The poem finishes with the old sorcerer's statement that powerful spirits should only be called by the master himself. Wikipedia, "The Sorcerer's Apprentice" accessed 13 August 2012
Image may be NSFW.
Clik here to view.
Source: http://upload.wikimedia.org/wikipedia/commons/0/05/Tovenaarsleerling_S_B... accessed 13 August 2012
(I so dearly wanted to use a still here of Mickey Mouse as the Sorcerer's Apprentice from the animated film Fantasia, but Disney is notoriously litigious about use of its intellectual property, and I really don't need the headache they gave this Florida daycare over it. So I didn't.)
These systems are developed to the needs of clinical practitioners, who have long ago finished studying basic sciences. The applications are programmed to respond to what the clinician needs in daily practice, not the basic sciences knowledge base.
Since most of these programs are about storing and retrieving information, rather than reasoning to derive new knowledge, this works for short-term needs. If basic science is needed, the user does the cognitive work involved, and stores the pertinent information in the system.
This won't work for more intelligent systems, though. The computer can't use knowledge that is not explicitly stated, even though humans work with each other like that every day.
In order to make all these systems work on the computer, all the information that we've implicitly communicated with other humans--or thought we were communicating, even if we weren't really--needs to be made explicit.
Moreover, since relationships between concepts constitute an important dimension of knowledge, next-generation knowledge sources must model comprehensively not only the concepts but also the relationships that characterize a particular field of basic science. (Mejino 2004)
"Comprehensively model" means to make all this information explicit, so that all the steps necessary to support reasoning are present and available. They model not only the existence of the entities, but also the relationships among those entities--how they interact with, and relate to, each other.
So comprehensively modeling superficial fascia in a way that promotes sound and valid reasoning about it and other anatomy it's involved with means putting it in the right place in the hierarchy, to make it clear that it's not a sibling of deep fascia, and should not be confused with it.
Therefore, there is a need to generate enabling knowledge sources at least in those domains that generalize to diverse fields of education, biomedical research and clinical practice. Anatomy is such a fundamental domain. (Mejino 2004)
Absolutely right--getting the anatomy correct is just that important.
How well are we doing at it?
As we're about to see, not that great. But some of our problems are getting fixed, even as we speak--the ones Jacobs indicates, and a host of other ones, as well.
UMLS and the other initiatives in biomedical informatics are intended for creating large and interoperable bodies of knowledge that can communicate with each other, and with people.
Although they're not primarily intended to improve anatomy education, they will inevitably do so as a side effect.
Making knowledge explicit, rather than implicit, and ensuring the validity of the knowledge that is modeled--discarding mistaken information and affirming validated information--will lead to supporting our students as they learn this material, to supporting our teachers as they step up to the challenge of increased teaching competencies, and to supporting our clients as we get better at understanding and communicating about the anatomy that underlies what we have to offer them.
Image may be NSFW.
Clik here to view.
Source: http://upload.wikimedia.org/wikipedia/en/7/74/MGM_sorcererhat.jpg accessed 13 August 2012
Coming up next: The scope of the problem