Decision support system to help choose between an ITE or a BTE hearing aid

Anwar, Naveed and Oakes, Michael (2011) Decision support system to help choose between an ITE or a BTE hearing aid. In: British Society of Audiology (BSA) Conference, 7th - 9th September 2011, Nottingham, UK.

Preview

PDF (Poster) BSA2012_poster.pdf - Published Version Download (453kB) | Preview

Abstract

A decision support system (DSS) is used for analysing a situation and making decisions. The goal of this research is to mine a large set of heterogeneous audiology data and create a DSS to help audiology technicians to choose between an ITE or BTE hearing aid. Although, in many cases such a choice is clear cut, but at other times this system could be used as a second opinion to predict the hearing aid type. A number of data mining techniques, such as clustering of audiograms, association analysis of variables (such as, age, gender, diagnosis, masker, mould and free text keywords) using contingency tables and principal component analysis on audiograms were used to find candidate variables to be combined into a DSS. The DSS was created using the techniques of logistic regression, Naïve Bayesian analysis and Bayesian networks, and these systems were tested and validated on test data to see which of the techniques produced the better results. This DSS takes air and bone conduction frequencies, age, gender, diagnosis, masker, mould and some free text words associated with a patient as input and gives as the output a decision as to whether the patient would be more likely to prefer an ITE or a BTE hearing aid type. The highest agreement between predicted results and actual hearing aid type in the data were obtained using Bayesian networks, with 93 to 94 percent similarity overall, with a precision of 0.91 for ITE and 0.96 for BTE. The reason for this might be that the Bayesian network also considers interaction between variables while the other two techniques (logistic regression and Naïve Bayesian analysis) consider only the individual variables. One of the important features of this DSS is that once the final choice of hearing aid type is predicted, the decision process can be tracked back to see which factors (variables) contributed how much to the final decision. The theoretical upper bound of classifier performance is the inter-annotator agreement (Altman, 1991), in this case the rate at which two expert audiologists would assign the same hearing aid to the same patient. Unfortunately, this type of data was not included in the audiology database.

Actions (login required)

View Item

Downloads