Convolutional Neural Network for Classifying Thoracic Diseases in Adult Patients

This paper proposes an automatic classifier, based on a convolutional neural network, capable of identifying different pathologies and diseases seen in anterior chest radiographs. The dataset was obtained from the National Institutes of Health (NIH) and Kaggle to be used as training for the classifier. For example, healthy, cardiomegaly, infiltration, effusion, mass, pneumothorax, emphysema, fibrosis, oedema, nodules and others. The average results for the evaluated metrics recall, accuracy and F1 score were 92%, 82% and 87% respectively, showing the flexibility of the proposed model to handle different tasks. This paper proposes an automatic classifier based on a convolutional neural network capable of identifying different pathologies and diseases shown in anterior chest X-ray studies. The data set was obtained from the National Institutes of Health (NIH) and Kaggle to be used as training for the classifier. E.g., Healthy, Cardiomegaly, Infiltration, Effusion, Mass, Pneumo-thorax, Emphysema, Fibrosis Edema, Nodule and others.