Data mining of audiology data for procuring hearing aids and tinnitus maskers

Anwar, Naveed and Oakes, Michael (2011) Data mining of audiology data for procuring hearing aids and tinnitus maskers. In: Fourth York Doctoral Symposium on Computer Science, 20th October 2011, York, UK.

Full text not available from this repository. (Request a copy)
Official URL: http://www.cs.york.ac.uk/ftpdir/reports/2011/YCS/4...

Abstract

This poster describes the data mining of a large set of patient records from the hearing aid clinic at James Cook University Hospital in Middlesbrough, UK. As is typical of medical data in general, these audiology records are heterogeneous, containing the following three different types of data:

– Audiograms (graphs of hearing ability at different frequencies)

– Structured tabular data (such as gender, date of birth and diagnosis)

– Unstructured text (specific observations made about each patient in a free text comments field)

There are two research questions for this research:

– To find factors influencing the selection of ITE (in the ear) opposed to BTE (behind the ear) hearing aids. James Cook University Hospital is unique in that it supplies ITE hearing aids on the National Health Service

– Among those patients diagnosed with tinnitus (ringing in the ear), to find the factors influencing the decision whether or not to prescribe a tinnitus masker (a gentle sound source, worn like a hearing aid, designed to drown out tinnitus)

A range of statistical techniques, including clustering, the chi-squared test, principal component analysis, a Naïve Bayesian approach and multiple logistic regression have been used on this set of audiology data. These enabled us to discover candidate variables (including individual words in the free text fields) which may influence the dichotomies identified above. The findings of this research are compared with the experience of the professional audiologist at James Cook Hospital. The candidate variables are combined using both a Naïve Bayesian approach and multiple logistic regression to produce decision support systems, where unseen patient records are presented to the system, and the relative probability that the patient should be fitted with an ITE as opposed to a BTE aid or a tinnitus masker as opposed to no tinnitus masker will be returned. The advantage of these techniques for the combination of evidence is that it is easy to see which variables contributed to the final decision, facilitating the production of explanation facilities for decision support systems to be used in the real world.

Item Type: Conference or Workshop Item (Poster)
Subjects: G400 Computer Science
Department: Faculties > Engineering and Environment > Mathematics, Physics and Electrical Engineering
Depositing User: Naveed Anwar
Date Deposited: 27 May 2015 12:44
Last Modified: 13 Oct 2019 00:31
URI: http://nrl.northumbria.ac.uk/id/eprint/22623

Actions (login required)

View Item View Item

Downloads

Downloads per month over past year

View more statistics