Personalized learning pathway for ankle fractures: Semantic imaging database and knowledge space modeling using a reproducible research methodology.
Marcelo Bordalo
Talitha Yen
José Elvano Moraes
Jacson Barros
Tarcisio Barros Filho
João Ricardo Vissoci
Ricardo Pietrobon
Trauma in general still constitutes one of the main reasons behind disability around the world, but with an even greater importance in developing countries and areas with a poor distribution of healthcare providers. In these locations not only the trauma incidence tends to be higher, but also because professionals with appropriate training for the diagnosis and therapeutic planning of bone fractures might not be present. Given the high degree of specialized knowledge required for these tasks, not only the risk of subsequent disability is increased , but even life threatening fractures might be missed such as in the case of cervical fractures. Despite the importance of better educating healthcare professionals regarding bone fractures, to our knowledge no open source, freely available educational technologies have been developed to address this need.
- imaging signs and fracture classification as the basis for diagnosis and therapy
- how difficult a given sign might be - prior for more sophisticated IRT models and personalized learning
The objective of this study is therefore to first create an ankle fracture imaging repository, semantically encoding imaging signs and classifications validated by a group of experts. Second, these same experts rank each sign according to a difficulty scale and encode this information into a knowledge space model used to establish an ankle fracture learning pathway.
We first describe how the database was created, validated, and then semantically encoded. We then describe the process to model the knowledge space toward the creation of a learning pathway.
- Select a group of 100 ankle fractures - DICOM for radiographs and anonymized . images selected to represent a wide range of severity
- create a checklist of common signs
- select a classification
- have a group of five specialists independently grade 40 images each, so that each image is graded by at least two specialists
R (@rcit)
Apache 2.0 anonymized images released under a CC license scripts under Figshare and Github
@Manual{rcit, title = {R: A Language and Environment for Statistical Computing}, author = {{R Core Team}}, organization = {R Foundation for Statistical Computing}, address = {Vienna, Austria}, year = {2014}, url = {http://www.R-project.org/}, }