An effective model for the detection of pneumonia from chest X-ray images using inner residual inception

Abstract Pneumonia is a serious disease that can lead to death if it is not diagnosed in an accurate manner. This paper presents three models for diagnosing pneumonia based on Chest X-Ray images. The first proposed model depends on the combination of inception, residual, and dropout. The second model is based on adding a batch normalization layer to the first model. The third model adds inner residual inception. The inner residual inception block has four branches, each of which has a significantly deeper root than any other known inception block, necessitating the use of residual connections between each branch. Inner residual inception blocks eventually consist of 4 distinct ResNet architectures. Each branch has a building block that is repeated three times with residuals, and then a dropout layer is added on top of that. These models used logistic regression and the Adam optimizer. The metrics used to evaluate the models are accuracy, precision, recall, F1-score, AUC, and balanced accuracy. From the results, the third proposed model has achieved the highest accuracy of 96.76%, and the best balance accuracy of 95.08%.