Title:Applied Computer Vision on 2-Dimensional Lung X-Ray Images for Assisted Medical Diagnosis of Pneumonia

Abstract:This study focuses on the application of a specific subfield of artificial intelligence referred to as computer vision in the analysis of 2-dimensional lung x-ray images for the assisted medical diagnosis of ordinary pneumonia.
A convolutional neural network algorithm was implemented in a Python-coded, Flask-based web application that can analyze x-ray images for the detection of ordinary pneumonia. Since convolutional neural network algorithms rely on machine learning for the identification and detection of patterns, a technique referred to as transfer learning was implemented to train the neural network in the identification and detection of patterns within the dataset. Open-source lung x-ray images were used as training data to create a knowledge base that served as the core element of the web application and the experimental design employed a 5-Trial Confirmatory Test for the validation of the web application.
The results of the 5-Trial Confirmatory Test show the calculation of Diagnostic Precision Percentage per Trial, General Diagnostic Precision Percentage, and General Diagnostic Error Percentage while the Confusion Matrix further shows the relationship between the label and the corresponding diagnosis result of the web application on each test images.
The developed web application can be used by medical practitioners in A.I.-assisted diagnosis of ordinary pneumonia, and by researchers in the fields of computer science and bioinformatics.