Connectionists: PhD Studentship, Department of Computer Science, University of Sheffield (UK)

[Aline Villavicencio]

Project: Automatic representation and identification of similarities in multimorbidity for large patient cohorts

Department of Computer Science,  University of Sheffield (UK)
Supervisor: Aline Villavicencio
Deadline for applications: January 29, 2021
More details: https://www.findaphd.com/phds/project/automatic-representation-and-identification-of-similarities-in-multimorbidity-for-large-patient-cohorts/?p128514
About the Project

The increasing availability of Electronic Health Records (EHR) along with advances in natural language processing have contributed to the automatic extraction of information from EHRs. This can also be augmented with additional knowledge from external sources about the characteristics of diseases and clinical terms found in these records, including taxonomic and synonymy information that can help discover similarities among terms. These techniques for detecting similarities among words, sentences, documents have gained particularly in accuracy with the automatic construction of large-scale models for representing knowledge, like BERT (Devlin et al. 2019), along with specific versions developed for the clinical domain like BioBERT (Lee et al., 2019), SciBERT (Beltagy et al., 2019), ClinicalBERT (Alsentzer et al., 2019), BioMedRoBERTa (Gururangan et al., 2020)  and BERT-XML (Zhang et al. 2020). 

This project aims to develop methods for automatically discovering similarities among different morbidities that could lead to a single unified treatment for some multimorbidities. This involves comparing similarities between diseases at a large scale to identify possible candidate combinations, extracting relevant information about diseases from sources including medical taxonomies and encyclopedias and creating a representation for each of the conditions and their properties that allows for their comparison. In addition, techniques for identifying similarities between representations will be used for large-scale comparison of diseases according to their common characteristics, including symptoms, treatments and evolution. This involves combining clinical information, codes and  medical taxonomies descriptions with additional information from external sources, such as publications. We propose to use this enriched information to detect direct and indirect similarities in different morbidities. These will enable the discovery of candidate combinations of diseases that have similar characteristics and could have joint treatments. 

----------------------------------------------------

Aline Villavicencio
Chair in Natural Language Processing
Department of Computer Science, University of Sheffield
https://sites.google.com/view/alinev
https://www.sheffield.ac.uk/dcs/people/academic/aline-villavicencio

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/connectionists/attachments/20210114/2543200c/attachment.html>