EDHER-MED PROJECT

Summary of the Coordinate Project

The EDHER-MED project is multidisciplinary and merges the fields of health and informatics. The consortium submitting the proposal is formed by two teams with an outstanding track record in Natural Language Processing (NLP) research and its application to the medical domain. In fact, we have been actively collaborating for 10 years with very significant results that have allowed us to develop resources (data collections, models and tools) of great utility in the medical field. The complementarity of both groups has been evident both in our history of collaboration and in the shared development of critical tools for the advancement of our research.

Thanks to the experience of both teams in the medical domain we have the collaboration and participation in the project of healthcare institutions interested in finding solutions to important health challenges they face.

A common aspect of these health challenges raised in the project is the importance of early detection of the problem. Based on the data available in various medical documents our hypothesis is that several advanced tools can be developed to help identify indications of the presence of the problems under consideration that can alert physicians. The results of the previous joint research of both groups allow us to propose in this proposal the exploration of more advanced language processing technologies. Specifically, PLN resources include the creation and annotation of different corpora related to the considered use cases. We will also advance in the improvement of information representation models for the biomedical domain. We also propose the enrichment of medical ontologies, in which Spanish is less represented, as well as the automatic generation of clinical argumentation to support or oppose a given hypothesis, the definition and refinement of argument units to facilitate explainability, and the automatic generation of timelines to predict future diagnoses or patient admissions.

The tools and technology jointly developed by the groups will be applied and evaluated on a series of use cases that, for different reasons, have a high social impact. Specifically, we intend to address the study of the following health problems: - The first use case will be developed by both groups and concerns the early detection of mental health problems in children and adolescents, with special attention to suicide. - Early detection of HIV will also be addressed by exploring different techniques to extract key indicators of HIV status. Both statistical techniques and machine learning and deep learning techniques will be explored. - Research will also be carried out to improve the characterization of rare diseases and their effects on the mental health and well-being of children. - Finally, the last use case contemplated in this project corresponds to the automatic discovery of potential risk factors associated with cardiovascular complications (e.g. ischemic stroke or cardiomyopathy) after a first episode.