Chapter 6: Implementation of Densenet Deep Learning Mechanism for Thorax Disease Classification in Chest X-Ray Images

One of the most readily available diagnostic methods for chest problems is thorax radiography. Deep learning has been developed drastically these days and its algorithms are steadily more utilized to improve the accuracy of detection of the abnormalities related to the thorax disorders on chest X-ray pictures. It’s crystal clear that the correlation of deep convolutional neural networks (DCNNs) on comparing to single networks helps in outperforming chest X-ray picture categorization performance. Here, we put forward DenseNet-169 to study the features with abnormalities for the classification of thorax disease on the results of the radiography on the upper body. A training technique is also developed to combine the loss contributions of the participating classifiers into a single loss. A blend of the discriminative functions in DenseNet helps to enhance the performance of the thorax disorder class in chest X-ray images. We carry out our trials on the CheXpert dataset that exhibits the productiveness of the proposed technique when compared to its previous proposals.