Enhancing Disease Detection with Machine and Deep Learning: Analyzing Chest X-rays for Pneumonia and Atelectasis Identification

Machine learning is a useful artificial intelligence (AI) tool that can be used in the medical field to obtain accurate predictions in laboratory results and give healthcare professionals a better understanding of the patient’s conditions before diagnosis. Deep Learning (DL) is another technique that uses advanced learning techniques and algorithms that can help give a better output in certain cases. From using x-rays, CT, MRI, and PET scan images, to using different signal and text datasets, an easier understanding of other diseases and their physical, mental, and social conditions can be well understood using various ML and DL techniques and algorithms, such as k-Nearest Neighbors (kNN), Transfer Learning, Decision Tree methods like AdaBoost, Random Forest, Convolutional Neural Networks (CNN) like AlexNet and DenseNet, Artificial Neural Networks (ANN), etc. In this work, the area of interest lies in using chest x-ray (CXR) images, which depicts two major disease conditions – Pneumonia and Atelectasis. Atelectasis-infected patients would have disproportionate lungs due to the fall of the alveolar sac found in them, making the lungs appear smaller than they are. In Pneumonia-infected patients, white patches are found inside the pulmonary organs in x-rays. Through this work, the Transfer Learning model achieved maximum accuracy of 95.55%, while minimum accuracy of 47.78% was obtained when using CNN model with DenseNet architecture.