The Application of a Genetic Algorithm in the Noninvasive Assessment of Vocal Nodules in Children

Abstract

The application of IT solutions in medicine makes it possible to develop new, more accurate, and noninvasive medical diagnostics. The aim of this study was to propose this kind of solution. It enables the accurate assessment of vocal nodules in children while measuring glottal insufficiency. The input data includes voice and electroglottographic recordings of patients' voices as well as diagnoses made by practitioners. The recordings were parameterized and used to develop a classifier to assess glottal insufficiency of vocal nodules. The classifier was designed with the help of a genetic algorithm. The diagnoses established thanks to the classifier show a 92% agreement with those reached through medical examination. Such effective performance renders the classifier a useful noninvasive screening tool. We compared our method with Deep Neural Network classifier and the Covariance Matrix Adaptation Evolution Strategy (CMA-ES) evolutionary algorithm. The solution that we propose offers a more accurate continuous diagnosis in comparison with the discrete diagnosis of a deep neural network as well as greater accuracy in relation to the CMA-ES algorithm. Another advantage of the proposed solution is the ease with which it can be implemented by healthcare professionals. A Visual Basic for the Applications (VBA) code for LibreOffice macro for the classifier is attached at the end of this paper.