Classification of Pneumonia from Chest X-ray Images Using Keras Module TensorFlow

Franky Chandra Satria Arisgraha, S.T., M.T. (Biomedical Engineering Study Program, Department of Physics, Faculty of Science and Technology, Universitas Airlangga, Indonesia)
Riries Rulaningtyas (Biomedical Engineering, Faculty of Science and Technology, Universitas Airlangga, Surabaya, Indonesia)
Miranti Ayudya Kusumawardani (Biomedical Engineering, Faculty of Science and Technology, Universitas Airlangga, Surabaya, Indonesia)

Indonesian Applied Physics Letters
Vol. 4 No. 1 (2023): June

Pneumonia is a respiratory disease caused by bacteria and viruses that attack the alveoli, causing inflammation of the alveoli. This study aims to examine the ability of the Convolutional Neural Network (CNN) model to classify pneumonia and normal x-ray images. The method used in this research is to construct a CNN model from scratch by compiling layers one by one with the help of the Keras TensorFlow module, which consists of a Convolution layer, MaxPooling layer, Flatten layer, Dropout layer, and Dense layer. Data used in this research is from Guangzhou Women and Children Medical Center, Guangzhou, China. The total data used is 200 images divided into 160 test data, 20 training data, and 20 validation data. From the results of the research conducted, the model has the fastest processing speed of 9.6ms/epoch with a total of 20 epochs. The model has the highest accuracy value of 77% in the training process and an accuracy value of 80% in the testing process. The highest sensitivity value is 1.54 in training and 1.6 in testing. The highest specificity value is 0.77 in training and 0.8 in testing. It can be said that the model can do good classification.