Weighted Box Fusion Ensembling for Lung Disease Detection

Abstract

Lung disease is a kind of infection that cannot be observed by human eyes, but it can be detected by X-ray images. Nevertheless, diagnostic accuracy from the look on these X-ray images is always a challenging task because it frequently relies on the experiences of the doctors. Hence, many computer vision techniques are developed to support medical staffs to perform lung disease detection. So as to encourage different detectors to identify different kinds of lung disease, this paper proposes a novel post-processing technique, a Weighted Box Fusion (WBF) ensembling algorithm, which may improve the efficiency in lung disease detection. This algorithm is applied to an available benchmarking dataset of X-ray chest in literature for justifying its performance.