A New Deep Learning Architecture for Pneumonia Detection in Pediatrics

This paper proposes a new convolutional neuronal network architecture for pneumonia detection in pediatric healthcare. The model was constructed from the scratch and was trained, validated, and tested using a public data set composed of 5863 images classified into two distinguished classes: normal and pneumonia. Indeed, the obtained results show an improvement compared to the existing works which have used the same data set. In this work, our achieved experimentations have reached an accuracy equal to 99.3% with a validation accuracy equal to 95.89% and a test accuracy equal to 97.10%. Thus, obtained results are very encouraging compared to existing models in the literature.