Chest X-ray image analysis and classification for COVID-19 pneumonia detection using Deep CNN

Gao, T.; Wang, G.

In order to speed up the discovery of COVID-19 disease mechanisms, this research developed a new diagnosis platform using deep convolutional neural network (CNN) is able to assist radiologists with diagnosis by distinguishing COVID-19 pneumonia from non-COVID-19 pneumonia in patients at Middlemore Hospital based on chest X-rays classification and analysis. Such a tool can save time in interpreting chest X-rays and increase the accuracy and thereby enhance our medical capacity for detection and diagnosis COVID-19. The idea is that a set of X-ray medical lung images (which include normal, infected by bacteria, infected by virus including COVID-19) were used to train a deep CNN which can be able to distinguish between the noise and the useful information and then uses this training to interpret new images by recognizing patterns that indicate certain diseases such as coronavirus infection in the individual images. The supervised learning method is used as the process of learning from the training dataset can be thought of as a doctor supervising the learning process. It becomes more accurate as the number of analyzed images growing. In this way, it imitates the training for a doctor, but the theory is that since it is capable of learning from a far larger set of images than any human, can have the potential of being more accurate.