TB or not TB? Acoustic cough analysis for tuberculosis classification

TL;DR

This paper demonstrates that advanced neural network architectures, including BiLSTM and attention mechanisms, improve tuberculosis cough classification and enable interpretability of audio features, with potential for better diagnostic tools.

Contribution

It introduces a novel attention-based architecture with greedy feature selection for TB cough classification, enhancing generalization and interpretability over previous methods.

Findings

BiLSTM improves classification performance.

Attention mechanism identifies important temporal audio regions.

Neural style transfer reveals distinct spectral features for TB and non-TB coughs.

Abstract

In this work, we explore recurrent neural network architectures for tuberculosis (TB) cough classification. In contrast to previous unsuccessful attempts to implement deep architectures in this domain, we show that a basic bidirectional long short-term memory network (BiLSTM) can achieve improved performance. In addition, we show that by performing greedy feature selection in conjunction with a newly-proposed attention-based architecture that learns patient invariant features, substantially better generalisation can be achieved compared to a baseline and other considered architectures. Furthermore, this attention mechanism allows an inspection of the temporal regions of the audio signal considered to be important for classification to be performed. Finally, we develop a neural style transfer technique to infer idealised inputs which can subsequently be analysed. We find distinct differences between the idealised power spectra of TB and non-TB coughs, which provide clues about the origin of the features in the audio signal.