Convolutional neural networks with approximation of Shapley values for the classification and interpretation of pneumonia in X-ray images

Pneumonia is a lung disease responsible for the highest number of deaths from infection in children and adults. Its diagnosis must be fast and accurate so that procedures are taken as soon as possible to combat the disease. In this work, Convolutional Neural Networks were explored for the classification of chest radiography images in the context of pneumonia diagnosis. Although these models are highly effective, their predictions are difficult to interpret. Therefore, the proposed method additionally aims at presenting an explainable model based on Shapley approximation values to perform the diagnosis of pneumonia with higher robustness. Results show that the model achieves competitive accuracy when compared to other architectures, and overcome them with respect to interpretation abilities.