Automatic Fuzzy Cognitive Maps for Explainable Image-based Pneumonia Detection

Georgia Sovatzidi, Michael Vasilakakis, Dimitris K. Iakovidis

Abstract: Pneumonia has a significant impact on morbidity and mortality worldwide and is associated with serious diseases, such as coronavirus disease 2019 (COVID-19). Pneumonia diagnosis is typically performed by medical experts, trained to evaluate chest x-rays, which is usually a difficult and time-consuming task. To address this problem, in this paper, a novel classification scheme based on a Fuzzy Cognitive Map (FCM) is introduced. The proposed FCM model is applied for the detection of foci of consolidation, which is a common radiographic manifestation of pneumonia, while enabling the explanation of the outcome using linguistic terms. Also, unlike most FCM models, it is automatic, in the sense that it does not require any manual intervention for the construction of the fuzzy graph. Experimental results using publicly available datasets demonstrate the effectiveness of the introduced model.