Browse Prior Art Database

Disease Relevancy Evaluation for a Corpus of Evidence

IP.com Disclosure Number: IPCOM000234117D
Publication Date: 2014-Jan-13
Document File: 3 page(s) / 33K

Publishing Venue

The IP.com Prior Art Database

Abstract

Disclosed is a process for creating disease-relevance scores for a corpus of evidence.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 37% of the total text.

Page 01 of 3

Disease Relevancy Evaluation for a Corpus of Evidence

In the healthcare field, there is an overwhelming amount of literature being produced, and keeping up with the latest developments in a particular field is nearly impossible. Busy doctors seldom have the time to devote to reading the literature, and some studies have suggested that the average doctor spends less than five hours a week keeping up with the latest studies and trials. Products such as IBM Watson* aim to address this problem by bringing the most current, reliable, and pertinent information to the doctors, at the time that they need it. Their time needs to be made more efficient, and products like IBM's Watson must be able to quickly and reliably find the most relevant articles for a case and for a disease. This technique uses many examples of a healthcare setting, but all of this could be just as easily applied to a different field of expertise.

    One of the techniques being used to solve the problem, and the focus of this disclosure, is the idea of applying disease relevance to evidence sources. An approach has been developed where one can assemble a core set of medical concepts for each type of disease. This can range from very general to very specific, and this approach is flexible in that regard. So if a cardiologist is interested in Hypertrophic Cardiomyopathy, one can start with that anchor concept, and expand to related concepts such as Beta Blockers, Ejection Fraction, Angina, Syncope, Dyspnea, etc. There are different techniques available for this expansion of concepts, and the main one being used here is to exploit medical ontologies like Unified Medical Language System (UMLS) and HealthLine to find closely related terms. These medical ontologies are being used to expand the list of concepts until there is a reasonable size of core set of medical concepts. If desired, this set of core concepts can be reviewed by a subject matter expert. This set of concepts would likely be roughly a hundred or more concepts, all of which are closely related to the disease being analyzed. Special care has been taken to avoid broad terms likes "pain", "disease" or "patient".

    This procedure then uses this core set of concepts as a way to rank evidence. This approach could also be used outside of the medical field, perhaps for financial investments or life insurance. Any field in which there is a large corpus of evidence to process, and a dictionary of related terms, could benefit from this invention. This process uses the corpus of evidence, along with an ontology of concepts in that fields, and with some guidance from a subject matter expert, produces a set of relevance scores for each subtopic.

    The main idea of this disclosure is that once you have this set of core concepts for a medical field, or particular condition or disease (Hypertrophic Cardiomyopathy, Lung Cancer, Hypertension, etc.), you can use this to develop a disease relevance score for any document in your corpu...