Application of Artificial Neural Network in Analysis of Acoustic Reflectometry Data

M. Hannula, A. Hirvikoski, H. Hinkula, and T. Holma (Finland)

Keywords

Otorhinolaryngology, intelligent data analysis, ear infection, middle ear effusion.

Abstract

Fluid in the middle-ear is an indicator of acute otitis media. In a number of previous studies acoustic reflectometry (AR) has been found to be a potential method to detect fluid in the middle ear. However, it has been found out in many studies that the reliability of AR in detection of the presence of the fluid requires improvement. To develop new tools to improve the reliability, in this study applicability of artificial neural network (ANN) in analysis of AR measurement data was evaluated. In the study a total of 375 AR measurements with five different levels (none, ¼, ½, ¾ and full) of fluid in the ear were performed and analyzed with ANN method. The measurements were done by using a plastic model of ear, including models of ear canal, tympanic membrane, malleus and middle ear. The results showed the ANN based analysis clearly to have potential for reliable detection of the middle ear fluid from AR measurement data; the method yielded in 100% accuracy in analysis of level of fluid in the ear with the five different levels of the fluid with validation data set (62 AR measurements). Repeatability of the ANN analysis was evaluated with 250 AR measurements with good results.

Important Links:



Go Back